See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/382145340

International Test and Evaluation Standards for Artificial Intelligence based

on Networked Data-Information- Knowledge-Wisdom-Purpose (DIKWP)
Model

Technical Report · July 2024

DOI: 10.13140/RG.2.2.28549.56809

CITATIONS READS

11 1,392

45 authors, including:

Yucong Duan Kunguang Wu

Hainan University Hainan University
1,463 PUBLICATIONS 6,042 CITATIONS 86 PUBLICATIONS 207 CITATIONS

SEE PROFILE SEE PROFILE

Fuliang Tang
Hainan University
109 PUBLICATIONS 219 CITATIONS

SEE PROFILE

All content following this page was uploaded by Fuliang Tang on 11 July 2024.

The user has requested enhancement of the downloaded file.

 International Test and Evaluation Standards for Artificial
Intelligence based on Networked Data-Information-
Knowledge-Wisdom-Purpose (DIKWP) Model
(20240709 Version)

Contributing Units of the Editorial Board: (in alphabetical order, without precedence)：

AIII, AGI-AIGC-GPT Evaluation DIKWP (Global) Laboratory, Blue Edu, Peking University, University of Science and
Technology Beijing, Bejing Academy of Social Sciences, Beijing Institute of Standardization Technology, Chengdu University
of Information Technology, Chongqing Police College, Dongguan Advantech Precision Manufacturing Co., Ltd, Guangxi
Normal University, State Grid, Hainan University, Hainan Nuclear Power Co., Ltd., Hainan Pushi Intelligent Technology Co.,
Ltd, Hainan Provincial Administration for Market Regulation, The Second Affiliated Hospital of Hainan Medical University,
Huazhong Agricultural University, Jiangsu Lizhuo Information Technology Co., Ltd, Kensid (Zhuhai) Co., Ltd, People's
Procuratorate of Liaoyang City, Liaoning Province, Nanjing police university, Inner Mongolia University, Ningbo University,
Institute for Advanced Study, Tsinghua University, Shandong University, Shanxi Provincial Bureau of Data, Shanghai Aerospace
Information Technology Research Institute, Shangrao Normal University, Sangfor Technologies Inc., World Association of
Artificial Consciousness(WAAC), World Conference on Artificial Consciousness(WCAC),Tai Chi Computer Co., Ltd, Xi’an
University of Technology, Southwest University of Political Science & Law, Standardization Research Center of Guangdong
Hong Kong Macao Greater Bay Area, Institute of Standardization Theory and Strategy of China Institute of Standardization,
China Association for the Application of Mechatronics Technology, China Institute of Information and Communication
Technology, etc

International Standardization Committee of Networked

DIKWP for Artificial Intelligence Evaluation
World Association of Artificial Consciousness

World Conference on Artificial Consciousness

 Catalogue

Preface ........................................................................................................................... 5

1 Scope........................................................................................................................... 6

2 Normative References ............................................................................................... 6

3 Terms and Definitions ............................................................................................... 6

3.1 DIKWP............................................................................................................ 6

3.2 Concept Space (ConC) ................................................................................... 8

3.3 Cognitive Space (ConN) ................................................................................ 9

3.4 Semantic Space (SemA) ............................................................................... 10

3.5 Data Definition ............................................................................................ 11

3.6 Information Definition................................................................................ 15

3.7 Knowledge Definition ................................................................................. 18

3.8 Wisdom Definition ..................................................................................... 23

3.9 Purpose Definition ...................................................................................... 26

3.10 DIKWP Graph ........................................................................................... 28

3.11 Fine-tuning System..................................................................................... 32

3.12 Agent............................................................................................................ 33

3.13 RAG ............................................................................................................. 33

3.14 Disembodied and Embodied Intelligence................................................. 33

3.15 Mixture of Experts (MoE) ......................................................................... 33

4 Evaluation System ................................................................................................... 34

4.1 DIKWP Semantic Understanding .............................................................. 34

 4.2 DIKWP Comprehensive Processing ........................................................... 36

4.3 DIKWP Bias ................................................................................................. 37

4.4 DIKWP Alignment ....................................................................................... 39

4.5 DIKWP Security........................................................................................... 41

4.6 DIKWP Artificial Consciousness Chip Evaluation ................................... 42

5 Evaluation Indicators ............................................................................................. 43

5.1 Basic Composition ........................................................................................ 43

5.2 DIKWP Functional Indicators .................................................................... 43

5.3 Performance Indicators ............................................................................... 47

5.4 DIKWP Artificial Consciousness Chip Evaluation Indicators ................ 49

6 Evaluation Methods ................................................................................................ 51

6 Evaluation Methods ................................................................................................ 51

Appendix ..................................................................................................................... 51

Appendix 1 Specific Methods of Evaluation............................................................ 51

Appendix 2 Evaluation Method Case Design .......................................................... 54

Reference .................................................................................................................... 58
Preparation Instructions

The development of International Test and Evaluation Standards for Artificial

Intelligence based on Networked Data-Information-Knowledge-Wisdom-Purpose
(DIKWP) Model aims to address the diversity and varying maturity levels of artificial
intelligence evaluation benchmarks, providing the industry with a unified, rigorous, and
universally applicable assessment framework. This is done to establish a standardized
evaluation method, model, and indicator system, ensuring comparability and
consistency in the assessment of artificial intelligence model performance. Starting
from the five core elements of data, information, knowledge, wisdom, and purpose,
International Test and Evaluation Standards for Artificial Intelligence breaks through
the conceptual space semantic constraints to construct a networked evaluation system
oriented towards cognitive space, comprehensively examining the understanding and
processing capabilities of artificial intelligence across different cognitive networks. Its
purpose is not only to overcome the limitations of black-box testing but also to provide
functional or indicative white-box evaluation results for the tested artificial intelligence
models, and to provide reference for the improvement of the tested models. This
evaluation standard pays special attention to the fairness, impartiality, and equality of
models, actively promoting AI technology to follow ethical principles and social values,
and striving to reduce discrimination and injustice phenomena.
The evaluation of artificial intelligence globally involves various benchmarks
aimed at measuring various capabilities and aligning them with human values. The
GLUE Benchmark provides a standardized set of different NLP tasks to evaluate the
effectiveness of different language models. As an industry standard for general
language understanding assessment, it has established its core position in evaluating
language model performance with its wide application, diverse task sets including
sentiment analysis, question answering, and providing a platform for fair comparison
between models. However, the benchmark also has significant shortcomings such as
limited task coverage, potentially unable to fully capture the full picture of language
understanding, biased datasets that may affect the fairness of model evaluation, and the
possibility of models over-optimizing specific tasks to pursue high GLUE scores,
leading to compromised generalization abilities. SuperGLUE Benchmark, as an
upgrade to GLUE, significantly increases task difficulty, covers more diverse and
deeper language understanding requirements, and by introducing human-level
baselines, intuitively demonstrates the progress of AI in the field of language
understanding. However, SuperGLUE is more complex and requires significant
computational resources, and its dataset bias issue, inherited from GLUE, remains a
persistent challenge. HellaSwag is a benchmark specifically designed for text
generation evaluation, which promotes the development of related technologies by
designing tasks specifically for text coherence and logic. However, its application scope
is relatively narrow, mainly focusing on text generation tasks, and the completion
examples of sentences may be suggestive, potentially causing interference with the
accuracy of model performance evaluation. TruthfulQA, with its uniqueness in directly
evaluating the truthfulness of model outputs, is particularly suitable for scenarios highly
concerned with answer authenticity, such as news generation. Although it adopts
innovative evaluation methods, the evaluation process of TruthfulQA is subjective, and
it only focuses on truthfulness while neglecting other aspects of language understanding
abilities, limiting its comprehensiveness. MMLU provides extensive domain and task
coverage, conducting in-depth, comprehensive, and reliable evaluations of models'
multi-task processing capabilities using large-scale datasets. Therefore, MMLU
requires high computational resource demands, and balancing evaluations between
multiple tasks is also a challenge, making it difficult for research teams with limited
resources to effectively utilize this benchmark.
While the current landscape of AI benchmarking features a variety of approaches,
each showcasing unique value in evaluating AI model performance, their limitations
are also evident. There is an urgent need to build a comprehensive, fair, and future-
proof international evaluation standard for artificial intelligence. The DIKWP model
describes a networked relationship model between data, information, knowledge,
wisdom, and purpose, where each relationship is guided by objectives or purpose,
connecting cognitive space, awareness space, semantic space, and conceptual space
together. Based on this, International Test and Evaluation Standards for Artificial
Intelligence covers multiple aspects and dimensions such as semantic understanding,
comprehensive processing, bias assessment, alignment assessment, and consciousness
capabilities, meticulously examining the understanding abilities of data, information,
knowledge, wisdom, and purpose, ensuring a comprehensive, multi-layered, and
thorough analysis of the cognitive abilities of artificial intelligence models. In addition
to language understanding and generation capabilities, International Test and
Evaluation Standards for Artificial Intelligence particularly focuses on the
performance of models in handling uncertainty, identifying and eliminating biases,
value alignment, and consciousness simulation, reflecting its emphasis on AI ethics and
fairness. With its features of comprehensiveness, depth, multidimensionality,
refinement, adaptability, fairness, and impartiality principles, International Test and
Evaluation Standards for Artificial Intelligence provides a scientific, rigorous,
practical, and forward-looking framework for the performance evaluation of artificial
intelligence models, aiming to promote the healthy development and application of AI
technology.

Preface

With the rapid development of artificial intelligence technology and the rise of
large language models, the way we interact with intelligent systems is undergoing a
change. Artificial intelligence models have demonstrated unprecedented performance
in various applications, attracting significant attention from various sectors of society
regarding their evaluation. Currently, the evaluation benchmarks for artificial
intelligence models present a diversified trend, with various benchmarks designed
specifically for different dimensions of performance emerging continuously. Examples
include GLUE, SuperGLUE, CLUE, SuperCLUE, which are used to test language
understanding and generation capabilities, while specific domain benchmarks like Owl-
Bench are tailored for areas such as intelligent operations and maintenance. In addition,
there are also evaluation benchmarks focusing on aspects like the security of large
models, ethical risks, fairness, reflecting the industry's emphasis on comprehensive
quality control of models. However, given the uneven levels of development and
maturity among these evaluation benchmarks, researchers face challenges in selecting
and referencing them, requiring a careful evaluation of suitable evaluation tools and
standards in combination with specific application scenarios. Faced with numerous
artificial intelligence model evaluation benchmarks launched by various research teams
and companies, there is a common and urgent demand among relevant professionals in
the field for a comprehensive, systematic, fair, and practical set of evaluation indicators
and methods to guide and promote the development and evaluation of artificial
intelligence models. The AGI-AIGC-GPT Evaluation DIKWP (Global) Laboratory,
composed of experts and scholars with long-term engagement in artificial intelligence
research, has drafted the "International Test and Evaluation Standards for Artificial
Intelligence based on Networked Data-Information-Knowledge-Wisdom-Purpose
(DIKWP) Model" aiming to establish an internationally recognized evaluation
benchmark for artificial intelligence with a certain degree of foresight and pilot
conditions.

1 Scope

This document outlines the relevant terms and definitions of International Test and
Evaluation Standards for Artificial Intelligence for artificial intelligence model
evaluation, describing the framework for evaluating AI models using DIKWP,
proposing DIKWP indicators for model evaluation, DIKWP evaluation methods, and
typical application cases.
This document is applicable to service providers, users, and third-party testing
organizations involved in the design and implementation of artificial intelligence model
testing.

2 Normative References

The contents of the following documents constitute relevant provisions of this

document through normative references therein.

3 Terms and Definitions

3.1 DIKWP

The DIKWP model is an extension of the traditional DIKW (Data, Information,

Knowledge, Wisdom) model, incorporating an additional element called "Purpose."
The DIKWP model is a networked model that vividly describes the cognitive process,
tightly connecting data, information, knowledge, wisdom, and purpose. Together, these
elements form an interactive process of cognitive conceptual-semantic associations that
spans cognitive space, conscious space, semantic space, and conceptual space.
 Data

Purpose Purpose

Purpose
Wisdom Information

Purpose Purpose

Knowledge

Figure 3-1 DIKWP Relationship Architecture Diagram

Figure 3-2 DIKWP Definition Metamodel

Figure 3-3 Conceptual Space, Cognitive Space, Semantic Space Architecture Diagram

3.2 Concept Space (ConC)

Concept space refers to the cognitive representation of the external world by a

cognitive subject, including the definition, features, and relationships of concepts.
Conceptual space is expressed through language and symbol systems. For example,
"car" in conceptual space can be defined as a type of transportation with four wheels,
capable of carrying passengers or goods.
Conceptual space is composed of a collection of related concepts connected to each
other through specific attributes and relationships, forming a directed or undirected
graph based on the symmetry of relationships between concepts.
Graph Representation: GraphConC = (VConC, EConC), where VConC is the set of
concept nodes, and EConC is the set of edges representing the relationships between
concepts.
In the concept space, each concept v∈VConC possesses a set of attributes A(v) and
relationships R(v, v) with other concepts.
Attributes: A(v)={a1(v), a2(v),…, an(v)}, where each ai(v) represents an attribute
of the concept v.
Relationships: R(v, v') represents the relationship between concept v and concept
v'. If the graph is directed, then R(v,v') is not equivalent to R(v, v); if the graph is
undirected, they represent the same relationship.
Operations within the concept space include querying, adding, or modifying
concepts and their relationships:
Query operation: Q(VConC, EConC,q)→{v1, v2,…, vm}, returns a set of concepts that
meet the criteria specified by q (such as specific attributes or relationships).
Add operation: Add(VConC, v), adds a new concept v to the concept collection VC.
Update operation: Update(VConC, v, A(v)), updates the attribute set A(v) of the
concept v.
In the DIKWP model, the conceptual space provides a structured framework for
categorizing and organizing data, information, knowledge, wisdom, and purpose. By
mapping each component of DIKWP to the conceptual space, complex relationships
between these components can be effectively deciphered. For example, through query
operations (Q), one can identify all concepts related to specific data or knowledge,
thereby deriving new information or wisdom.

3.3 Cognitive Space (ConN)

The cognitive space is a multidimensional and dynamic processing environment

where data, information, knowledge, wisdom, and purpose are transformed into specific
understanding and actions through specific cognitive processing function sets (R) of
individuals or systems. Each cognitive processing function ( 𝑓𝐶𝑜𝑛𝑁𝑖 ) takes data or

information from the input space (𝐼𝑛𝑝𝑢𝑡𝑖 ) and transforms it through a series of sub-
steps (such as data preprocessing, feature extraction, pattern recognition, logical
reasoning, and decision-making) into outcomes in the output space (𝑂𝑢𝑡𝑝𝑢𝑡𝑖 ), such as
information classification, concept formation, purpose determination, or action
planning.
Function set: R={fConN_1, fConN_2,…,fConN_n}, where each function fConN_i: Inputi→
Outputi represents a specific cognitive processing step. Inputi is the input space, and
Outputi is the output space.
Input space Inputi: Includes various data or information sources received by the
individual or system. These inputs may come from observations of the external
environment (such as visual and auditory perceptions), signals received from other
systems, or internally generated data. The input space reflects the diversity of the
cognitive subject's interactions with the outside world and the breadth of information
acquisition.
Output space Outputi: Contains various higher cognitive products formed after
cognitive processing. This can include classification of input information, conceptual
structures built based on information, clear identification of purpose, and specific action
plans set to realize these purpose. The output space reflects the cognitive subject's
ability to deeply process and transform input information, which is the basis for the
cognitive subject's responses or actions towards the external world.

3.3.1 Cognition processing

Each cognitive processing function fConN_i can be further refined into a series of
sub-steps, including data preprocessing, feature extraction, pattern recognition, logical
reasoning, and decision-making. These sub-steps together constitute the complete
cognitive pathway from raw data to the final output.
Sub-step representation: For each fConN_i , it can be represented as fConN_i=fConN_i(5)
○ fConN_i(4) ○…○ fConN_i(1)(Inputi) , where fConN_i(j) represents the processing function of
the j-th sub-step, and ○ represents the composition of functions.

3.4 Semantic Space (SemA)

Semantic space refers to the semantic association network of concepts within the
cognitive subject's brain, including semantic relationships and associations between
concepts. Semantic space is formed through the cognitive subject's experiences and
accumulated knowledge. For example, for the concept of "car," the semantic space may
include associated semantics such as "driving," "vehicle," "fuel consumption," and
others.
The semantic space is a collection formed by a series of semantic units, which are
interconnected through specific associations and dependencies, collectively
constituting an objective representation of information and knowledge. Widely
accepted concepts and linguistic rules in the semantic space facilitate the transmission
and communication of meaning.
Graph Representation: GraphSemA=(VSemA, ESemA) ,where VSemA represents semantic
units (words, sentences, etc.), and ESemA represents the associations and dependencies
between semantic units.
Semantic Unit: Each semantic v∈VSemA represents the smallest unit or concept that
can independently express meaning.
Relationships: Edge e ∈ ESemA represents semantic associations or logical
dependencies between semantic units, such as synonymy, antonymy, hyponymy,
causality, and other relationships.
In the semantic space, a series of operations correspond to querying, adding, or
modifying semantic units and their relationships:
Query Operation: Query(VSemA, ESemA, q)→{v1,v2, … ,vm}, returns a set of
semantic units that satisfy the query condition q.
Add Operation: Add(VSemA, v), adds a new semantic unit v to the set VSemA.
Update Operation: Update(ESemA,v,v',e), updates or adds the relationship e between
semantic units v and v'.
The semantic space not only provides stakeholders with a cognitive shared
language system for expressing DIKWP, but also supports semantic consistency in the
transformation and processing between DIKWP components. Leveraging semantic
units and their relationships enables accurate transmission and interpretation of
complex service interaction cognitive content among different entities.
3.5 Data Definition

The semantics of data (DIKWP-Data) can be seen as specific manifestations of

the same semantics in cognition. In the conceptual space, the concept of data represents
the existence of specific facts or observational results in the conceptual space of
cognitive entities, confirmed by semantic correspondence with the consciousness space
(non-subconscious space) of cognitive entities and the existence of certain shared
semantics with existing cognitive concept objects. When processing data concepts,
cognitive processes often seek and extract specific shared semantics that label these
data concepts, thereby unifying them as the same concept based on corresponding
shared semantics. For example, when observing a group of sheep, although each sheep
may differ slightly in size, color, gender, etc., cognitive processing categorizes them
under the concept of "sheep" by accurately matching individual or probabilistic shared
semantics related to "sheep". Shared semantics can be specific, such as recognizing the
concept of an arm based on the number of fingers, color, and shape similarity between
a silicone arm and a human arm, or probabilistically selecting the target object with the
most shared semantics with the concept of an arm. Conversely, silicone arms lacking
the rotational function of real arms would be judged not to belong to the concept of arm
data based on the semantic judgment defined by "rotatable." The distinction between
conceptual space and semantic space corresponds to different philosophical
perspectives on technology.
Concept space processing corresponds to specific forms of communication using
natural language. However, the essential function of conceptual communication is
usually to convey semantics. In the cognitive space of cognitive entities, effective
understanding of conceptual transmission semantics often depends on the semantic
correspondence of related concepts in the semantic space of cognitive entities. The
semantic space of cognitive entities often cannot be fully shared through conceptual
forms, also known as subjective, thus referred to as subjective.
In the semantic space, the semantics of data concepts represent specific
manifestations of the same semantics set in cognition. Corresponding to specific data
semantics D, each element d∈D represents a concrete instance, sharing the same or
probabilistically approximate semantic attribute set S. Semantic attributes S are defined
by a set of feature semantics F, such as:
S={f1, f2, ... , fn}
where fi represents a feature semantic of the data.
D={d∣d share S}。
In the DIKWP model, the distinction between data concepts and data semantics
forms the basis for the cognitive process's transition from cognitive space to the
processing of conceptual space and semantic space. Data concepts and data semantics
represent specific cognitive data objects that directly embody basic observations and
factual knowledge about the world. The key to this transformation lies in the "same
semantics" shared among data concept elements behind their cognition and
conceptualization, i.e., the semantic attributes they share. In the cognitive space, data
cognitive objects serve as the foundation of cognitive processes, no longer merely
representing observations and measurements of the real world without distinguishing
between specific correspondences in conceptual and semantic spaces, but undergoing
explicit conceptual confirmation and semantic correspondence processing. This process
also distinguishes between subjective and objective content categories, differing from
the traditional DIKW model's crude understanding of data by emphasizing the close
association between data and specific semantic attributes. Thus, cognitive recognition
of data involves actively seeking semantic features that match known cognitive objects,
highlighting the subjectivity and context dependence of data and emphasizing its
cognitive value in associating with existing conceptual spaces of cognitive entities.
In the DIKWP model, the semantics of data cognitive objects from cognitive space
are considered specific manifestations recognized in the semantic space of cognitive
entities with the same semantics during the cognitive process. This definition
emphasizes that data as cognitive objects are not mere records of observations or facts
but results from semantic matching and conceptual confirmation processes conducted
by cognitive subjects (such as humans or AI systems) in conceptual space and semantic
space. The key to confirming data concepts lies in the "same semantics" shared by
cognitive space and semantic space of cognitive entities, enabling specific cognitive
objects to be categorized under the same data concept even in cases of external
differences.
Data concepts are viewed as basic conceptual units in the conceptual space of the
DIKWP model's cognitive process, while data semantics are viewed as basic semantic
units in the semantic space of the DIKWP model's cognitive process. Data concepts and
data semantics are core elements in the cognitive process of directly observing and
recording the real world, playing an important role from the cognitive recognition of
data semantics to the confirmation of data concepts in generating, applying, and
processing concept-based symbolic natural language. Data concepts are recognized and
classified by the conscious or subconscious cognitive functions of cognitive entities
through sharing the same semantic attributes. In cognitive science, how the brain and
even physical parts such as the spinal cord of cognitive entities understand and process
information through subconscious pattern recognition and can conduct conscious
analysis and form explanations is crucial. For example, when people observe different
objects (such as apples), even with differences in color, size, or shape, they can
recognize them as apples in subconscious pattern recognition. Through conscious
analysis, they can explain this recognition by sharing a set of key semantic attributes
(such as shape, texture, specific functions, etc.). This cognitive process reveals how the
cognitive system of cognitive entities uses the same semantics of data in semantic space
to construct natural language conceptual representations of the world.
In the DIKWP framework, data as concepts are seen as specific concept mappings
of the same semantics in the cognitive process. This viewpoint breaks through the
confusion in traditional data concepts where semantics and concepts are
undistinguished, linking the formation and existence of data concepts in conceptual
space closely with the semantic processing process of cognitive entities in semantic
space. That is, the cognitive value of data concepts lies not in their physical form or
function but in how they establish connections across "conceptual space" and "semantic
space" in the cognitive space of cognitive subjects (such as humans or AI systems),
thereby being recognized and confirmed as objects or concepts with specific semantics.
From the perspective of interaction between individual consciousness and group
consciousness, the interaction between data and cognitive entities is fundamentally
based on the interaction between semantic space and conceptual space in subconscious
or subconscious terms. Data concepts, as specific correspondences of the same
semantics set, or probabilistically approximate, have their advantages in cognitive
communication efficiency as symbolic expressions of specific semantic sets in
engineering terms.

3.5.1 Mathematical Representation of Data Concepts

In the DIKWP model, data concepts are not merely passive records of
observational results but collections of semantic objects actively recognized and
classified by cognitive systems. Mathematically, we can view data concepts as a
collection D of semantic instances, where each semantic instance d ∈ D is identified
as having the same set of semantic attributes S. Here, S={f1, f2, ... , fn} can be seen as a
set of parameters defining the semantic features of data concepts, where fi represents a
semantic feature of the data concept. This representation helps us understand how data
concepts are induced and processed based on shared semantic features.

3.5.2 Mathematical Description of Data

In the DIKWP model, data concepts are regarded as specific manifestations of the
same semantics in cognition. Mathematically, we can define the semantic set D
corresponding to data concepts as a vector space, where each element d∈D is a vector
representing a specific semantic instance. These semantic instances are categorized
under the same semantic attribute S by sharing one or more semantic features F, i.e.,
S={f1, f2, ... , fn}
Where fi represents a semantic feature of data concepts. Therefore, we can define
the collection of data concepts as:
D={d∣d share S}
This description emphasizes the semantic multidimensionality and semantic
structural nature of data concepts, while also providing a mathematical foundation for
subsequent data concept processing and analysis.

3.5.3 Data Concepts and Semantic Recognition

In the DIKWP model, the processing and understanding of data concepts go

beyond mere recording of objective facts; they involve how cognitive entities match
these factual semantics with their existing semantic cognitive structures. This process
emphasizes the importance of semantic recognition—how cognitive entities identify
and classify objects based on the semantic features within data concepts.

3.5.4 The Specific Manifestations of Data Concept and the Same Semantics

In the DIKWP model, data concepts and semantic recognition view data concepts
not just as observations and recordings of the real world but as specific manifestations
of the same semantic attributes perceived by cognitive entities in communication and
interaction. This definition transcends the surface-level independent objective cognitive
existence of data concepts as records of objective facts, emphasizing the cognitive
nature of data concepts in the interaction between cognitive entities in cognitive space.
That is, the recognition and processing of data concepts depend on the connections and
matches with existing semantics in the subjective semantic space of cognitive entities.
Data concepts inherently possess cognitive subjectivity and context-dependence,
meaning that the same data concept may be linked to and processed with different
semantics depending on different cognitive entities or cognitive backgrounds.
Philosophically, data concepts cease to be mere objective records of existence but
become subjective interpretations through the subjective cognitive processes of
individuals. The formation and existence of data concepts rely on the semantic space
and conceptual space memory and processing capabilities of cognitive entities,
representing the correlation and transformation between the semantic space and
conceptual space in the interaction between the real world and cognitive entities. The
generation and recognition of data concepts are not purely objective processes but
deeply rooted in the preconceived conceptual space and contextual semantic space of
the subject. Therefore, the recognition and interpretation of data concepts must take
into account the cognitive spatial background knowledge, experiential information, and
cultural contextual semantics of cognitive entities.
The meaning of data concepts must be confirmed through the interpretation and
semantic matching of cognitive entities. The interaction between data concepts and data
semantics becomes a bridge connecting objective reality with subjective cognition. This
understanding highlights a Platonic idea: things in the real world (as concepts) are only
shadows of their ideas (i.e., "same semantics"). Thus, the cognitive value of data
concepts lies not only in the objectivity of their forms but also in how cognitive entities
seek and confirm the shared semantics of cognitive objects and phenomena through
data concepts, triggering semantic resonance and cognitive confirmation. This
interactive process of re-cognizing data concepts and data semantics within cognitive
entities is not only a cognitive mirror reflection of the external world for cognitive
entities but also a pursuit and revelation of the intrinsic semantic nature of phenomena.
It emphasizes the cognitive dominance and creative existence of conceptual semantic
transformation in the interpretation of data concepts by cognitive entities, as well as the
interaction between data concepts and the subconscious or conscious symbolic
language of cognitive entities.

3.5.5 The Cognitive Properties and Semantic Entities of Data

The DIKWP model's cognitive definition of data concepts and data semantics
emphasizes the cognitive nature of data and their role as semantic entities. In philosophy,
this touches upon discussions of the "essence of things" and "being true to the name."
Data concepts are not merely symbolic records of objective existence; they are entities
endowed with specific data semantics, which are confirmed and endowed through the
cognitive entity's processing across conceptual and semantic spaces. This cognitive
processing also reveals that knowledge generation is not just a mapping of the objective
world but also a subjective process of constructing based on the transformation from
similar semantics to concepts. This aspect is reflected in Kantian epistemology, where
human knowledge of the world partly originates from external stimuli but is largely
determined by our cognitive structures.

3.6 Information Definition

Information (DIKWP-Information) as a concept corresponds to one or more

"different" semantics in cognition. The information semantics of the information
concept refer to the semantic association in the semantic space of the cognitive entity's
cognitive space with the DIKWP cognitive objects already recognized by the cognitive
entity through specific purpose concepts or purpose semantics. This is achieved by
using the cognitive purpose of the cognitive entity to form identical cognition
(corresponding to data semantics) or different cognition in the cognitive space, with the
probability confirmation of different cognition in the semantic space through "different"
semantics or logical judgment confirmation, forming new semantic associations ("new"
is a kind of "different" semantics). When processing the concept or semantics of
information, cognitive processing identifies the differences between the input cognitive
contents such as data, information, knowledge, wisdom, or purpose, and the recognized
DIKWP cognitive objects, corresponding to various different semantics, and classifies
information accordingly. For example, in the cognitive space, facing a parking lot,
although all cars in the parking lot can be cognitively classified into the concept of
"car," each car's parking position, time, wear and tear, owner, functionality, payment
records, and experiences represent cognitive differences recognized by different
cognitive purpose in the semantic space, ultimately corresponding to different
information semantics. Various different semantics corresponding to information
objects often exist in the cognition of the cognitive entity but are frequently not
explicitly expressed. For example, a patient with depression may use the concept of
"low spirits" to express the increase in the negative intensity of their current emotions
relative to their past emotions in their cognitive space. When the cognitive entity selects
the concept of "low spirits" in its conceptual space to reflect its confirmed cognitive
state, the target information semantics to be expressed, due to the fact that the
information semantics interpretation of the concept of "low spirits" in the cognitive
space of the communication target may not necessarily be the same as that of the
cognitive entity or may have different semantics, thus failing to achieve the objective
perception of the information semantics felt by the communication target, and therefore,
this information semantics becomes the subjective cognitive information semantics of
the cognitive entity.
Mathematical representation of information semantics processing: Information
semantics in the DIKWP model correspond to data semantics, information semantics,
knowledge semantics, wisdom semantics, and purpose semantics, generating new
semantics through the cognitive entity's purpose -driven processing. In the semantic
space, purpose -driven information semantics processing FI for DIKWP content
corresponds to the processing form from input X to output Y:
FI : X→Y
Where X represents the set or combination of data semantics, information
semantics, knowledge semantics, wisdom semantics, and purpose semantics
(collectively DIKWP content semantics), and Y represents the generated new set or
combination of DIKWP content semantics. This mapping emphasizes the dynamic and
constructive nature of the information semantics generation process.
In the DIKWP model, information semantics correspond to the expression of
various different semantics in cognition. Leveraging the cognitive purpose of the
cognitive entity, information semantics link the semantics corresponding to data,
information, knowledge, wisdom, or int purpose ent with the cognitive entity's existing
cognitive objects, generating new sets or combinations of semantics. In the cognitive
space, this process includes not only the recombination and semantic transformation
(including semantic connectivity to form what is known as cognitive understanding) of
known DIKWP content but also corresponds to a dynamic process of generating new
DIKWP cognitive semantics and continuously forming cognitive understanding
through such recombination and transformation.
The generation of information semantics concerns how to connect sets or
combinations of different data semantics, information semantics, knowledge semantics,
wisdom semantics, or purpose semantics through the cognitive entity's specific purpose,
thereby confirming cognitive understanding in the cognitive entity's cognitive space. It
corresponds to the cognitive entity forming semantic associations, supplements,
judgments, and thus eliminating cognitive uncertainty purpose originating from
semantic uncertainty. This process involves associating, comparing, and conceptually
corresponding observed phenomena or cognitive input contents with existing DIKWP
content in the cognitive space, thereby using various semantic recognition and
classification to generate new DIKWP content. In AI, this can correspond to the
formation of cognitive understanding, explaining and processing the relationships
between DIKWP content, such as extracting valuable information semantics through
algorithmic analysis of the correlation between DIKWP content.
Information semantics processing is a dynamic cognitive process that focuses on
how to link DIKWP content semantics with the cognitive entity's existing cognitive
objects DIKWP content semantics through the cognitive entity's subjective purpose,
thereby generating valuable semantic associations. The value of information lies in
serving as a bridge connecting data, information, knowledge, wisdom, and purpose,
revealing the semantic associations of the cognitive entity with DIKWP content.
In cognitive science, information semantics processing can be explained using
various cognitive theories. For example, Conceptual Integration Theory further
explains how to integrate information from different sources to create new meanings
and understanding. Similarly, by combining a person's behavior (DIKWP content
semantics) with specific contextual information, a clearer understanding of their
purpose can be achieved.
The semantic associations of information are related to theories in cognitive
linguistics such as Metaphor Theory and Blending Theory, which study how new
meanings are created through metaphor and conceptual integration in language. In AI
systems, this involves designing algorithms to simulate how humans construct new
cognitive models through existing DIKWP content semantics.
The process of generating information semantics is the result of the interaction
between DIKWP content semantics and DIKWP*DIKWP semantic interactions. This
process includes not only the reorganization or reinterpretation of DIKWP content
semantics but also a dynamic, purpose -driven cognitive activity. Through this activity,
the cognitive entity can identify and understand new patterns and associations, thereby
expanding its cognitive boundaries. The generation of information semantics is
constructive and dynamic, simply put, it is generated through the interpretation or
semantic connection of DIKWP content.
In philosophy, information is seen as the organization and interpretation of
DIKWP content, generating new semantics by constructing semantic relationships
between DIKWP content. Through the process of information semantics processing,
the cognitive entity can identify and understand the connections and differences
between phenomena. The generation of information semantics involves active
participation by the cognitive entity; it is an action of semantic processing of DIKWP
content, reflecting the cognitive entity's interpretation of the real world. Information, as
the expression of different semantics in cognition, philosophically signifies the
recognition and understanding of the diversity and complexity of the world.
Information semantics is not just an aggregation or recombination of DIKWP content
semantics; it is the creation of new semantic associations, reflecting the cognitive
entity's active exploration and interpretation of the world. This interpretive process
involves delving into the deeper connections and underlying logic of phenomena,
representing a pursuit of deeper understanding of the world.

3.6.1 The Construction Nature of Information Semantics

The generation and understanding of information are not passive processes of

reception but active cognitive semantic constructions within the cognitive space by the
cognitive entity. Information semantics depend on existing DIKWP content and
purpose -driven cognitive frameworks. This viewpoint resonates with Kant's
epistemology, where the understanding of the world by the cognitive entity is
constituted through internal perceptual frameworks and a priori concepts. The value of
information lies in its ability to expand or reconstruct our cognitive frameworks,
thereby enhancing our understanding of the world.

3.6.2 The Diversity and Depth of Information Semantics

 In DIKWP, information processing focuses on the dynamic relationships between
data, information, knowledge, wisdom, and purpose, and the generation of new sets or
combinations of semantics. This process embodies Heraclitus' doctrine of flux —
everything flows, nothing stands still. The value of information lies in its fluidity and
its capacity to induce change, rather than static factual records. Information becomes a
link connecting different cognitive states, driving the cognitive entity from one state of
understanding to another.

3.6.3 The Dynamicity of Information and Cognitive Structure

In the definition of information, the DIKWP model emphasizes the role of

information as a bridge connecting different semantic entities. This echoes Deleuze's
theory of "difference and repetition." According to Deleuze, the process of cognition
occurs through recognizing the differences between things, which is the core of
information processing. Information not only contains the semantic differences of
DIKWP content but also creates connections with existing knowledge structures
through these differences. This process not only integrates old knowledge but also
generates new knowledge. This dynamic cognitive structure updating process is crucial
for cognitive development and knowledge growth.

3.7 Knowledge Definition

The semantics of Knowledge (DIKWP-Knowledge) concepts correspond to

one or more "complete" semantics within the cognitive space. The semantics of
knowledge concepts are the cognitive entity's understanding and interpretation of
semantics between cognitive objects of DIKWP content, obtained through semantic
integrity abstraction activities using certain assumptions (i.e., forming cognitive
inputs of the cognitive entity's cognitive interaction activities with existing cognitive
DIKWP content semantics, and corresponding to one or more "complete" semantics
that carry cognitive integrity purpose in higher-order cognitive spaces). When
dealing with knowledge concepts, the brain abstracts at least one concept or pattern
corresponding to a complete semantics through observation and learning. For
example, while it is impossible to know all swans are white through observation, in
the cognitive space, the cognitive entity can apply assumptions (higher-order
cognitive activities assigning complete semantics) to some observed instances that
do not guarantee complete observational results, thus attributing "complete"
semantics, i.e., "all," forming knowledge semantics corresponding to knowledge
rules such as "all swans are white."
Knowledge K forms a semantic network, with its mathematical representation
in conceptual space as follows:
K= (N, E, AK)
Where N={n1, n2, … , nk} represents the set of nodes of concepts, E={e1,
e2, … ,em} represents the set of relationships between these concepts, A={AD, AI, AK}.
This definition positions knowledge cognition as a higher-level cognitive
achievement, emphasizing the structured nature of knowledge (such as semantic
networks) and its capability to capture complete semantics, which is crucial for
understanding complex systems and abstract concepts.
Knowledge bridges the transformation from a state of non-understanding to an
understanding corresponding to DIKWP content, based on comprehensive semantics,
and strengthens confirmation through validation. The construction of knowledge
relies not only on the accumulation of data and information but more importantly on
abstraction and generalization processes in cognition, forming so-called
understanding of the essence and intrinsic connections of things. Knowledge exists
not only at the individual level but also at the collective or societal level, shared and
disseminated through culture, education, and transmission.
Knowledge semantics refers to the structured understanding formed after deep
processing and internalization of DIKWP content (within the conceptual space and
corresponding to "complete" semantics in the semantic space). The definition of
knowledge within the DIKWP framework reflects a deep understanding of the world
and a grasp of complete semantics. This resonates with Aristotle's concept of formal
cause, suggesting that the essence and purpose of things can be explored and
understood through reason and experience.
Each formation of knowledge rules in the DIKWP model represents the
cognitive grasp of the intrinsic laws and essence of things by the cognitive subject.
From a philosophical perspective, knowledge is not only the product of cognitive
processes but also the purpose and guidance of these processes. The formation and
application of knowledge reflect the adaptation and transformation of the cognitive
subject to the deeper regularities of the real world in the semantic space
understanding.

3.7.1 Knowledge Concepts

Knowledge concept (DIKWP-Knowledge Concept) refers to the abstraction

and generalization of entities, events, laws, etc., in the cognitive space of the
cognitive subject towards the objective world. In the DIKWP model, knowledge
corresponds to one or more "complete" semantics in the cognitive space. Knowledge
is obtained through cognitive activities where the cognitive subject abstracts the
semantic completeness of DIKWP content with certain assumptions, gaining
understanding and interpretation of the semantic relationships between cognitive
objects DIKWP content. This understanding and interpretation form the cognitive
input of the cognitive interaction activities between recognized DIKWP content and
already existing cognitive DIKWP content, corresponding to one or more "complete"
semantics carrying cognitive completeness confirmation in higher-order cognitive
spaces. The formation of knowledge concepts is a dynamic process influenced by
cognitive purpose, contexts, and existing cognitive structures.

3.7.2 Knowledge Semantics

Knowledge semantics (DIKWP-Knowledge Semantics) is formed through

higher-order cognitive activities, assigning "complete" semantics to partial
observation results, thereby forming systematic understanding and rules. It involves
intrinsic connections between concepts and their external expressions. Knowledge
semantics are not static but enriched and developed through the dynamically
generated information semantics driven by cognitive purpose of the cognitive subject.
This means that the semantics of knowledge are not only based on objective facts but
also depend on how the cognitive subject understands and interprets these facts in
the current context and integrates these understandings with existing cognitive
structures. The process of generating and confirming semantics includes identifying
data features, matching concepts, and probabilistic semantic confirmation, ensuring
the applicability and accuracy of knowledge.

3.7.3 The relationship between knowledge concepts and knowledge semantics

In the DIKWP model, knowledge concepts and knowledge semantics are

interdependent and inseparable. Knowledge concepts provide the basic framework
and classification system for knowledge, while knowledge semantics fill in the
content of these frameworks, giving concepts practical meaning that can be applied
in specific contexts. Knowledge concepts provide a structured foundation for
semantics, enabling cognitive agents to integrate newly acquired information with
existing knowledge networks. On the other hand, knowledge semantics add
dynamism and flexibility to concepts, allowing them to adapt to different cognitive
tasks and changing environments. Through continuous cognitive processing,
including data collection, information generation, knowledge construction, and
wisdom extraction, knowledge concepts and semantics interact and develop,
collectively advancing the cognitive agent's deeper understanding of the world.
In essence, knowledge concepts provide the "framework" for semantics, while
knowledge semantics fill this "framework" with "flesh," and together, they support
the complexity and depth of the entire cognitive structure.

3.7.4 Structured Representation of Knowledge

In the field of artificial intelligence research, the precise expression and effective
organization of knowledge constitute the core driving force behind technological
advancement. This process is deeply rooted in the meticulous construction of cognitive
models and an in-depth understanding of information processing mechanisms. The
following are several key knowledge representation frameworks:
1. Formal Logic Systems: As a foundational framework for knowledge
representation, formal logic, particularly propositional logic and first-order predicate
logic, provides a rigorous mathematical basis for the precise expression and reasoning
of information. Propositional logic uses truth functions to express simple facts and their
logical relationships, while first-order predicate logic achieves formal descriptions of
entity attributes, relationships, and existence through advanced constructs such as
variables, predicates, and quantifiers. This supports the deductive reasoning and
consistency verification of complex propositions.
2. Production Systems: This is a rule-based representation method where
"condition-action" rules (i.e., production rules) become the primary units of knowledge
encoding. This model excels at simulating human expert decision-making processes,
especially in fields such as diagnosis, planning, and problem-solving. The flexibility of
production systems lies in their ability to dynamically select applicable rules based on
input conditions, achieving a mapping from known facts to target actions,
demonstrating the efficiency and practicality of rule-based knowledge processing.
3. Frame Representation: A frame is a structured knowledge template used to
organize information about a specific topic. Each frame consists of several "slots,"
which are filled with specific data or pointers to other frames, forming a closely
connected knowledge network. This method emphasizes the layering and
modularization of knowledge, facilitating the handling of structural information of
complex concepts and supporting efficient information retrieval and updating
operations.
4. Process Representation: Process representation focuses on demonstrating the
dynamics of knowledge, paying attention to state changes, event sequences, and the
execution of operations. It is particularly suitable for scenarios that require tracking
state changes, simulating event sequences, or designing control processes. Through
process representation, complex behavioral patterns and dynamic systems can be
systematically described and analyzed, providing strong support for fields such as
automated planning, robotic path planning, and workflow management.
Knowledge K is represented as a semantic network, where node n represents
concepts and edge e represents relationships between concepts:
K = (N, E)
Where N={n1, n2, … , nk} represents the set of concepts, E={e1, e2, … ,em}
represents the set of relationships between these concepts, and each edge can be
represented as
e =（ni, nj, r）,ni, nj∈N
And r represents the semantic relationship between ni and nj .

3.7.5 Cognition and Construction of Knowledge

Knowledge bridges the cognitive state transformation of understanding DIKWP

content from misunderstanding to understanding, reinforced through validation. The
construction of knowledge relies not only on the accumulation of data and information
but, more importantly, on abstraction and generalization during cognitive processes,
forming an understanding of the essence and inherent connections of things.
Knowledge exists not only at the individual level but also at the collective or societal
level, shared and disseminated through culture, education, and transmission.
Knowledge semantics are structured perceptions formed through deep processing
and internalization of DIKWP content. This understanding exists within conceptual
space and corresponds to "complete" semantics within semantic space. The definition
of knowledge within the DIKWP framework reflects a profound understanding of the
world and a grasp of complete semantics. This resonates with Aristotle's concept of
formal cause, suggesting that the essence and purpose of things can be explored and
understood through reason and experience.
In the DIKWP model, the formation of each knowledge rule represents the
cognitive grasp of the intrinsic laws and essence of things by the cognitive subject.
From a philosophical perspective, knowledge is not only the product of the cognitive
process but also its purpose and guide. The formation and application of knowledge
reflect the cognitive subject's adaptation to and transformation of the real world,
representing an understanding of the deep-seated laws of the semantic space of the
world.

3.7.6 The Cognitive Processing of Knowledge

In cognitive processing, cognitive agents abstract key concepts and patterns

through observation and learning. They utilize functions such as identification,
classification, reasoning, and memory to transform observed information into useful
knowledge, which then guides their thinking and actions. Take the cognitive model "All
swans are white" as an example: this knowledge construction involves observing swan
colors across various times and places, integrating these observations to form a general
statement about swan colors. Through accumulated experience, stable cognitive
patterns are formed, and when faced with challenges such as discovering a black swan,
these knowledge patterns are updated and adjusted. This process highlights knowledge
as a dynamic cognitive structure that evolves based on new evidence and understanding,
demonstrating its self-improvement capability during validation and expansion, and
reflecting its semantic integrity and dynamic development.

3.7.7 The Philosophical Significance of Knowledge

In the DIKWP model, knowledge is not merely a record of observations and facts
but a systematic understanding formed through assumptions and higher-order cognitive
activities. The semantic integrity and systematic nature of knowledge reflect the
cognitive subject's profound understanding and interpretation of the world. The process
of knowledge generation emphasizes the active and creative role of the cognitive
subject in understanding and interpreting the world. Through assumptions and
abstraction, partial observations are endowed with complete semantics, thus forming
systematic knowledge.
Knowledge semantics are not just an aggregation or reorganization of DIKWP
content semantics but a creation of new semantic associations, reflecting the cognitive
subject's active exploration and interpretation of the world. Through assumptions and
higher-order cognitive activities, the process of knowledge generation can reveal deep
connections and underlying logic between phenomena, providing a more
comprehensive and profound understanding of the world.

3.7.8 The Dynamicity of Knowledge Semantics

The generation of knowledge semantics is a dynamic process involving how

cognitive subjects link different DIKWP content semantics through assumptions and
higher-order cognitive activities to form new knowledge semantics. Within cognitive
space, this process encompasses not only the resemanticization and transformation of
known DIKWP content but also the generation of new cognitive insights and
knowledge semantics through such recombination and transformation.
This dynamism is reflected in the process of knowledge generation and updating,
where through continuous observation, learning, and validation, cognitive subjects can
develop and refine systematic knowledge structures. These knowledge structures not
only explain phenomena but also predict future behaviors and characteristics, providing
a deeper understanding and guidance of the world.

3.8 Wisdom Definition

Wisdom (DIKWP-Wisdom) corresponds to information regarding ethics, social

morals, human nature, and similar aspects. It represents a form of information derived
from cultural and human societal norms, contrasting with relatively fixed extreme
values of current times or individual cognitive values. When determining the semantic
of wisdom, the cognitive subject integrates semantic content including data,
information, knowledge, wisdom, and purpose within their cognitive space. The core
of human and artificial intelligence systems' wisdom revolves around constructing a
human-centered value system for building a community with a shared future for
humanity. This core value system serves as the foundation to construct, differentiate,
confirm, correct, and develop individual and collective cognitive, semantic, and
conceptual spaces of DIKWP content semantics. They are applied to guide decision-
making. For example, when faced with decision-making issues based on specific DIK
content, cognitive subjects should consider various factors such as ethics, morals,
feasibility, and more, rather than solely relying on technical or efficiency-based aspects
of DIK.
Wisdom, denoted as W in the decision function, correlates data, information,
knowledge, wisdom, and purpose, and outputs the optimal decision D∗:
W:{D,I,K,W,P}→D∗
Here, W is a decision function that generates the optimal decision D* based on
data, information, knowledge, and wisdom. This description emphasizes the
comprehensiveness and goal-oriented nature of the decision-making process. It
resonates with research in cognitive linguistics on how morals and values are expressed
and conveyed through language.
In the DIKWP model, the concept of wisdom is seen as a holistic embodiment
based on core human values, integrating considerations of ethics, social morality, and
individual values. Wisdom involves not only the application of data, information, and
knowledge but also comprehensive consideration and balance of various factors,
including moral and ethical aspects, in the decision-making process.
Wisdom semantics processing involves making judgments and decisions by
integrating data, information, and knowledge with individual or collective values and
ethical concepts. In the field of AI, wisdom semantics processing corresponds to
developing advanced decision-making artificial consciousness systems or ethical AI.
These systems can consider multiple factors based on human-centered principles to
provide solutions that are more intelligent and aligned with ethical standards.
In cognitive science, wisdom semantics processing corresponds to the process of
handling DIKWP content semantics from the perspective of human development, value
systems, moral judgments, and social contexts. Wisdom content is not just an
accumulation of DIKWP content but concerns how to process DIKWP semantic content
in the semantic space based on a vision of building a human community, starting from
cognitive space. For example, facing climate change, the application of wisdom
involves using the cognitive subject's understanding of environmental science
(knowledge), evaluating the long-term and short-term consequences of different
courses of action (information), and making decisions (information) based on ethical
and social responsibility (wisdom).
The formation of wisdom in cognitive subjects and social groups depends not only
on the cognitive capabilities of individuals and cognitive groups in understanding
DIKWP content semantics but also on the interaction, deep understanding, and
reflection of individuals on their environment, cultural background, and social
relationships corresponding to DIKWP content.
The DIKWP model views wisdom as a critical factor in the decision-making
process, involving considerations of ethics, morals, and values. It emphasizes the
inevitable connection between data, knowledge, and information with value
orientations in practical applications, reminding cognitive subjects that the cognitive
process is not just about pursuing truth but is also premised on exploring the ideal
human way of life. This corresponds to Aristotle's discussion of "Phronesis," or
practical wisdom, which explores how to make the best ethical judgments and decisions
in specific contexts.

3.8.1 The Core Values of Wisdom

The core values of wisdom revolve around building a human community with a
shared destiny, centered on human values. Cognitive subjects, relying on this core value
system, construct, analyze, affirm, correct, and develop the DIKWP content semantics
of individual and collective cognitive spaces, semantic spaces, and conceptual spaces.

3.8.2 Wisdom Decision-making Process

In wisdom semantics processing, cognitive subjects consider ethical, moral, and

feasibility factors in the decision-making process, rather than just relying on technical
or efficiency considerations based on data, information, and knowledge. The decision-
making process includes the following steps:
• Comprehensive Consideration: When facing decision-making problems, cognitive
subjects need to comprehensively consider factors such as ethics, morals, social
responsibility, and feasibility. For example, decisions on climate change need to
consider environmental impacts, social equity, and economic feasibility among various
aspects.
• Integration of DIKWP Content: Cognitive subjects integrate the semantics of data,
information, knowledge, wisdom, and purpose to form a comprehensive decision-
making foundation. For instance, in public policy formulation, decisions require
combining statistical data, social surveys, historical knowledge, and ethical principles.
• Decision Output: By comprehensively considering various factors, the decision
function 𝑊 outputs the optimal decision 𝐷∗. This process emphasizes the necessity of
balancing and optimizing multiple factors.

3.8.3 The Cognitive and Social Aspects of Wisdom

Wisdom exists not only at the individual level but also at the societal level. The
formation of individual wisdom and social wisdom relies on the integrated development
of cognitive individuals and groups' DIKWP content semantic cognitive capabilities, as
well as on a deep understanding and reflection of the environment, cultural backgrounds,
and social relationships. The process of wisdom formation includes:
• Cultural inheritance: Through cultural inheritance, wisdom is shared and
propagated within communities. For example, ethical values and moral principles in
traditional cultures are passed down through education and social practices.
• Social interaction: Wisdom formation also depends on social interaction, where
wisdom continually develops and improves through communication and collaboration
among people. For instance, collective decision-making processes in community
governance exemplify the embodiment of wisdom.

3.8.4 The Philosophical Significance of Wisdom

The definition of wisdom reflects a focus on ethics, morals, and values,

emphasizing comprehensive consideration and balance of various factors in the
decision-making process. This corresponds to Aristotle's concept of "Phronesis" or
practical wisdom, which underscores making the best ethical judgments and decisions
in specific contexts. The formation and application of wisdom embody the cognitive
subject's adaptation to and transformation of the world, representing an exploration of
the ideal human way of life.

3.8.5 The Application of Wisdom in AI

In the field of AI, the goal of wisdom semantics processing is to develop advanced
decision-making artificial consciousness systems or ethical AI that can consider
multiple factors based on human-centered principles and provide solutions that are
more intelligent and align with ethical standards. Applications of wisdom in AI include:
• Ethical AI systems: Designing AI systems capable of making ethical decisions in
complex environments. For example, self-driving car systems need to weigh the safety
of passengers and pedestrians in emergency situations to make decisions that adhere to
ethical standards.
• Advanced decision systems: Developing advanced decision systems that integrate
considerations from multiple factors. For instance, in medical diagnostics, AI systems
need to combine patient medical history data, medical knowledge, and ethical principles
to provide optimal treatment plans.

3.9 Purpose Definition

The semantics of Purpose (DIKWP-Purpose) in the model correspond to a tuple

(input, output), where both input and output consist of semantics related to data,
information, knowledge, wisdom, or purpose. Purpose semantics represent
stakeholders' understanding of the DIKWP content semantics of a phenomenon or
problem (input), as well as the objectives they hope to achieve through processing and
resolving that phenomenon or problem (output). When cognitive agents process
Purpose semantics, they operate within a semantic space where they interpret the input
DIKWP content semantics based on their predefined goal (output) semantics. Through
learning and adaptation of corresponding DIKWP content semantics processing, the
output DIKWP content semantics gradually approaches the predefined goal semantics.
P = (Input, Output), The input and output are the semantic contents of data, information,
knowledge, wisdom, or purpose. When processing purpose semantics, a series of
transformation functions T achieve semantic transformation from input to output based
on the input content and predefined objectives.
T : Input→Output, This representation emphasizes the dynamism and goal orientation
of the process, providing a mathematical model for understanding and designing
cognitive processes with specific objectives.
Purpose represents the purposefulness and directionality of cognitive processes,
serving as the driving force behind individual or systemic actions. Purpose not only
defines the transition path from the current state to the desired state but also reveals the
dynamics and direction of cognitive activities. This goal-oriented cognitive process
emphasizes the proactive and creative nature of the cognitive agent when processing
information, as well as the underlying motivations and goals behind cognitive activities.
The concept of purpose underscores that cognitive processes in the cognitive space are
goal-directed—meaning that cognitive agents do not merely passively receive
information but actively pursue specific goals and purpose, shaping how they
understand and manipulate DIKWP semantic content such as data, information,
knowledge, wisdom, and purpose itself. Purpose guides not only the collection and
processing of data and information by cognitive agents but also influences the
formation and application of knowledge and the development and practice of wisdom.
The concept of purpose introduces a teleological perspective, suggesting that
cognitive activities are not purposeless data processing but are aimed at achieving
certain goals or satisfying needs. In the DIKWP framework, the inclusion of purpose
enriches the model's dynamism and underscores the purposefulness and subjectivity of
cognitive activities. This implies that within cognitive processes, cognitive agents
actively seek, select, and interpret DIKWP semantic content based on specific goals
and purpose in the semantic space.
Purpose -driven processing provides a framework for understanding cognitive
activities from a dynamic and goal-oriented perspective, aligning with theories such as
Action Theory in cognitive linguistics. This enables the DIKWP model not only to
explain existing cognitive phenomena but also to guide future cognitive activities,
optimizing cognitive strategies and behaviors to achieve specific goals. In AI systems,
identifying purpose and designing goal-oriented behaviors are crucial for implementing
intelligent behaviors, such as understanding user queries in natural language processing
(NLP) or setting and optimizing paths to achieve goals in planning algorithms. In the
study of artificial consciousness, understanding and simulating human purpose
recognition and goal-oriented behavior are critical for achieving advanced cognitive
functions.
From a philosophical perspective, purpose is not just a predefined goal of action
but reflects the fundamental motive behind individual existence and behavior. Purpose
embodies individual free will and aspirations for the future, serving as the intrinsic drive
for interaction between individuals and the world. The existence of purpose emphasizes
the subjectivity and creativity of cognitive activities, revealing deeper meanings behind
human behavior. This corresponds not only with Aristotle's concept of final causes,
which posits that everything has a purpose or ultimate reason for existence, but also
resonates with Hegel's teleological view and existentialist philosophy's emphasis on
free will. In Hegelian philosophy, reality's drive comes from the unity of opposites,
achieved through self-realization and self-negation in the process of purposive action.
In existentialism, the emphasis is on individual choice and purpose as decisive factors
in one's existence. The purpose dimension in the DIKWP model reflects that cognitive
activities are not merely reactions to the external world but are processes where
individuals actively construct based on their own purposes and values.

3.9.1 The Cognitive Process of Purpose

In cognitive processes, purpose represents the transition path from the current state
to the desired state, revealing the dynamics and direction of cognitive activities. This
goal-oriented cognitive process emphasizes the proactive and creative nature of the
cognitive agent when processing information, as well as the underlying motivations and
goals behind cognitive activities.

3.9.2 The Application of Purpose

• Artificial Intelligence: In AI systems, recognizing purpose and designing goal-

oriented behaviors are key elements for achieving intelligent behavior. For example,
autonomous vehicle systems need to understand and execute passengers' purpose to
make optimal decisions in complex traffic environments.
• Natural Language Processing (NLP): In NLP, understanding user query purpose
forms the basis for providing accurate and relevant answers. Systems need to identify
user purpose based on their input and generate corresponding outputs.
• Cognitive Science: In cognitive science, understanding human purpose recognition
and goal-oriented behaviors are crucial for achieving advanced cognitive functions.
Researchers analyze human cognitive processes to design systems capable of
simulating and optimizing these processes.

3.9.3 Purpose from a Philosophical Perspective

From a philosophical perspective, purpose is not merely a preset goal of action but
the fundamental motive behind individual existence and behavior. Purpose embodies
individual free will and aspirations for the future, serving as the intrinsic drive for
interaction between individuals and the world.
• Aristotle's Final Cause: Aristotle believed that everything in existence has a purpose
or ultimate reason. Purpose, as the core of cognitive activities, aligns with Aristotle's
concept of final causes, emphasizing the importance of purpose in cognitive activities.
• Hegel's Teleological View: Hegel posited that reality's drive comes from the unity
of opposites, where through the process of purposive action, self-realization and self-
negation lead individuals to higher cognitive realms.
• 。Existentialist Free Will: Existentialist philosophy emphasizes the decisive role of
individual choice and purpose in one's existence. Purpose reflects individual free will
and serves as the core driving force of cognitive activities.

3.10 DIKWP Graph

3.10.1 Data Graph（DG）

In the network model, data graphs are not only the starting point of information
processing but also the result of feedback adjustments for knowledge, wisdom, or
purpose. The data graph (DG) receives inputs from information, knowledge, wisdom,
and purpose through transformation functions TID, TKD, TWD, TPD achieving dynamic
updates and adjustments.
TXY: YG→XG, where X,Y ∈{D,I,K,W,P} and X ≠ Y, denotes the transformation
from graph Y to graph X.

3.10.2 Information Graph（IG）

An information graph is defined as a tuple IG = (VI, EI) , where VI is the set of

information nodes and EI is the set of edges based on semantic relationships between
the information. The information graph is not only generated from the data graph DG
but is also adjusted and reconstructed by the knowledge graph KG, widom graph WG,
and purpose graph PG:
TDI
DG → IG : Transformation from data to information.
TKI TWI TPI
KG → IG , WG → IG , PG → IG : Adjustments of information by knowledge,
wisdom, and purpose, respectively.
Where TXY represents the transformation function from graph X to graph Y.
3.10.3 Knowledge Graph（KG）

A knowledge graph is defined as KG = (VK, EK), where VK represents knowledge

nodes and EK represents the relationships between knowledge nodes. The knowledge
graph integrates information formation and influences the interpretation of data, the
generation of information, and the application of wisdom:
TIK
IG → KG : Transformation from information to knowledge.
TKD TKI TKW
KG → DG , KG → IG , KG → WG : The influence of knowledge on data, information,
and wisdom, respectively.

3.10.4 Wisdom Graph（WG）

A wisdom graph is defined as WG = (VW, EW), where VW represents the nodes of

wisdom and EW represents the connections between wisdom nodes. The wisdom graph
integrates knowledge, data, and information to guide decision-making and can feedback
to influence the formation of knowledge and the interpretation of information:
TKW
KG → WG : Transformation from knowledge to wisdom.
TWK TWI
WG → KG , WG → IG : Feedback effects of wisdom on knowledge and information,
respectively.

3.10.5 Purpose Graph（PG）

The purpose graph is defined as PG = (VP, EP), where VP represents the nodes of
goals and implementation paths, and EP represents the strategies or steps to achieve
these goals. The purpose graph is constructed by data, information, knowledge, and
wisdom, and it can inversely influence these components:
TDP TIP TKP TWP
DG → PG , IG → PG , KG → PG , WG → PG : Formation of purpose from data,
information, knowledge, and wisdom.
TPD TPI TPK
PG → DG , PG → IG , PG → KG : Inverse influence of purpose on data, information,
and knowledge.
The DIKWP graphing system maps elements of the digital world and the cognitive
world to five main components: DG, IG, KG, WG, and PG. Each graph is further
subdivided into three levels of mapping: the semantic level, the conceptual level, and
the instance level. Thus, each graph g∈G is a triplet mapping:
g:S×C×I where G represents the set of graphs, S represents the set of semantic
levels, C represents the set of concepts, and I represents the set of instances.
The interactions between the DIKWP graphs are achieved through content models
and cognitive models, represented by a function f , which transforms the mapping of
one level or type of graph into another level or type of graph.
f: G×G→G
 Complete, accurate, and consistent interaction modeling: objective content
space<->subjective cognitive space
Multimodal Objective Content Stakeholders (individual subjective
World (Resources) Interaction/ cognitive space)
communication/
processing

DIKWP typed graph mapping

(linguistic)
DIKWP Objective-Subjective Integration
(Linguistic Concept-Instance Space)
DIKWP content DIKWP (stakeholder) cognitive
graphs graph

Multiple sources of imprecise, incomplete, Content data content information Cognitive data graph cognitive information graph
Multiple sources of imprecise, incomplete,
transform and inconsistent human-machine-object
and inconsistent human-machine-object graph graph interaction subjective and objective resources
interaction subjective and objective resources

Content Purpose Graph Cognitive purpose graph

{information graph} 输
输 输 输输输
输出 输
{infor mation gr aph}
输
输 入 输 输输
输 输输输 输入 入
输
入
输
输 入 输出 输
输 输输 入 输出
入 输出输
出 出入
输入 入
输
入
输出输 入 输输
入
出入
入 输出
入 出 出入 Cognitive Wisdom Graph cognitive knowledge graph 出入
入 输输
入 Content Wisdom Graph content knowledge graph 出
出入出入
出 DIKWP Subjective Cognitive
DIKWP Objective Content
Resource Graphical Mapping
Resource Graphical Mapping Consolidation of
DGD IKWD type resources
D IKW DP
DPD D D
data graph transform

IGD IKWI P ID IKW PI

infor mation gr aph
I I
D
D
I
II
K
save
DIKWP Semantic Fusionisation
I transform
KGD IKW D II
knowledge graph K
KP K K K
WGD IK
K
D IKW P II
transform
K Transformation (Semanticisation)
wisdom graph
WD IKWW P II K
K
WP W W K W
W W
PG D IKW transform save
purpose PD IKWPP I K
K K
PP P P P
graph
K W
Consol idation of DIKWP
DIKWP graph resources
Graph St ored o n th e
Resources DI KW P gra ph
Type R esou rce

DIKWP semantic graph (semantic space)

Semantic data graph semantic information graph

semantic purpose graph

Semantic Wisdom Graph semantic knowledge graph

Figure 3-4 Concept Space DIKWP Graph and Graph Relationship Transformation

In the conceptual space, the paradigm for converting across DIKWP (Data,
Information, Knowledge, Wisdom, purpose) resources, as illustrated in Figure 3-4,
involves mapping each dimension of DIKWP to corresponding processing methods:
data is mapped to conceptual statistical analysis, information is modeled and analyzed
using the partial order structure of lattice theory, knowledge is deepened through
various reasoning techniques, wisdom is reflected in multi-objective value-based
decision-making, and purpose is associated with precise modeling of goals and
problems.In this process, by systematically integrating processing methods for
individual DIKWP types and promoting pairwise interactive combinations among them,
a comprehensive processing mapping system that spans different types of DIKWP
resources is established.
 D I K W P
D D1+P D2 D+P I D+P K D+P W D+P1 P2
I I+P D I1+P I2 I+P K I+P W I+P1 P2
K K+P D K+P I K1+P K2 K+P W K+P1 P2
W W+P D W+P I W+P K W1+P W2 W+P1 P2
P P1+P2 D P1+P2 I P1+P2 K P1+P2 W P1+P2 P3
D 统计
I 格论
K 推理
W 权衡
P 问题

统计 格论 推理 权衡 问题
D D D I D K D W D P
统计
统计 统计 统计 格论 统计 推理 统计 权衡 统计 问题
I D I I I K I W I P
格论
格论 统计 格论 格论 格论 推理 格论 权衡 格论 问题
K D K I K K K-W K P
推理
推理 统计 推理 格论 推理 推理 推理 权衡 推理 问题
W D W I W K W W W P
权衡
权衡 统计 权衡 格论 权衡 推理 权衡 权衡 权衡 问题
P D P I P K P W P P
问题
问题 统计 问题 格论 问题 推理 问题 权衡 问题 问题

Figure 3-5 Combining DIKWP Purose Driven Empirical Transformation Paradigm

with Mathematical Logic Form Processing
The mathematical representation of the DIKWP conceptual space is as follows:

𝑫𝑰𝑲𝑾𝑷𝑪𝒐𝒏𝒄𝒆𝒑𝒕𝒖𝒂𝒍𝑺𝒑𝒂𝒄𝒆 = (𝑵𝑫𝑰𝑲𝑾𝑷 , 𝑬𝑫𝑰𝑲𝑾𝑷 , 𝑨𝑫𝑰𝑲𝑾𝑷 )

Where 𝑵𝑫𝑰𝑲𝑾𝑷 = {𝒏𝟏 , 𝒏𝟐 , … , 𝒏𝒌 } represents the set of nodes of DIKWP concepts,

and 𝑬𝑫𝑰𝑲𝑾𝑷 = {𝒆𝟏 , 𝒆𝟐 , … , 𝒆𝒎 }represents the set of semantic relationships between
these DIKWP concept nodes.
The relationships within the same DIKWP concept layer are:
𝑨𝑫𝑰𝑲𝑾𝑷 = {𝑨𝑫 , 𝑨𝑰 , 𝑨𝑲 , 𝑨𝑾 , 𝑨𝑷 }

𝑨𝑫 = [𝒂𝒊𝒋 ] represents the weight matrix of data semantic relationships between

𝑫

DIKWP data concept nodes.

𝑨𝑰 = [𝒂𝒊𝒋 ] represents the partial order relationship matrix of information semantic

𝑰

relationships between DIKWP information concept nodes.

𝑨𝑲 = [𝒂𝒊𝒋 ] represents the logical reasoning relationship matrix of knowledge

𝑲

semantic relationships between DIKWP knowledge concept nodes.

𝑨𝑾 = [𝒂𝒊𝒋 ] represents the value relationship matrix of wisdom semantic

𝑾
relationships between DIKWP wisdom concept nodes.

𝑨𝑷 = [𝒂𝒊𝒋 ] represents the mathematical relationship matrix of purpose semantic

𝑷

relationships between DIKWP purpose concept nodes.

The transformation relationships between different DIKWP concept layers are:

𝑨𝑫−𝑫 𝑨𝑫−𝑰 𝑨𝑫−𝑲 𝑨𝑫−𝑾 𝑨𝑫−𝑷
𝑨𝑰−𝑫 𝑨𝑰−𝑰 𝑨𝑰−𝑲 𝑨𝑰−𝑾 𝑨𝑰−𝑷
𝑨𝑫𝑰𝑲𝑾𝑷∗𝑫𝑰𝑲𝑾𝑷 |
= | 𝑨𝑲−𝑫 𝑨𝑲−𝑰 𝑨𝑲−𝑲 𝑨𝑰−𝑾 𝑨𝑲−𝑷 |
|
𝑨𝑾−𝑫 𝑨𝑾−𝑰 𝑨𝑾−𝑲 𝑨𝑾−𝑾 𝑨𝑾−𝑷
𝑨𝑷−𝑫 𝑨𝑷−𝑰 𝑨𝑷−𝑲 𝑨𝑷−𝑾 𝑨𝑷−𝑷

𝑨𝑫 = [𝒂𝒊𝒋 ] represents the weight matrix of data semantic relationships between

DIKWP data concept nodes.

𝑨𝑰 = [𝒂𝒊𝒋 ] represents the weight matrix of partial order information semantic

relationships between DIKWP information concept nodes.

𝑨𝑲 = [𝒂𝒊𝒋 ] represents the weight matrix of logical reasoning knowledge semantic

relationships between DIKWP knowledge concept nodes.

𝑨𝑾 = [𝒂𝒊𝒋 ] represents the weight matrix of value wisdom semantic relationships

between DIKWP wisdom concept nodes.

𝑨𝑷 = [𝒂𝒊𝒋 ] represents the weight matrix of mathematical purpose semantic

relationships between DIKWP purpose concept nodes.

3.11 Fine-tuning System

Prompt engineering refers to the process of designing and optimizing input

prompts to guide an artificial intelligence model to generate specific outputs. This
process includes selecting appropriate words, sentence structures, and context to
maximize the model's performance and output quality. Prompt engineering can be
divided into hard prompts and soft prompts.
Hard prompts are explicit and specific prompts that usually provide the model with
clear instructions and goals. Hard prompts contain fixed text structures and content to
ensure the model's output meets expectations. These prompts typically do not rely on
contextual changes and have high certainty.
Soft prompts are more flexible and implicit prompts, usually using embedded
vectors or slightly adjusted text prompts to guide the model's output. Soft prompts often
rely on the context and the model's internal representations, offering high adaptability
and flexibility. Soft prompts may be implemented by fine-tuning model parameters or
using vectors in the latent space, providing subtle control over the output.

3.12 Agent

An agent is a computational entity capable of autonomous actions and decision-

making. Agents typically perceive their environment (through sensors or input data)
and take actions (through actuators or outputs) to achieve predetermined goals. Agent
systems can be software, hardware, or a combination of both and are widely used in
various fields such as autonomous driving, recommendation systems, robotics, and
game AI.

3.13 RAG

RAG stands for Retrieval-Augmented Generation. It is a concept that provides

external knowledge sources to large language models, enabling them to generate
accurate and contextually appropriate answers while reducing hallucinations in the
model.

3.14 Disembodied and Embodied Intelligence

Disembodied Intelligence refers to agents without a physical embodiment,

usually manifesting as purely software systems interacting and operating within virtual
environments.
In small models and discriminative methods, disembodied intelligence refers to
lightweight discriminative models trained for specific tasks under limited
computational resources. These models focus on pattern recognition or classification
tasks without physical world interaction.
In large models and generative methods, disembodied intelligence involves using
large generative models (e.g., GPT-3) for complex generation tasks. These models
utilize vast training data to generate natural language text, images, or other content
without direct interaction with the physical world.
Embodied Intelligence refers to agents with a physical embodiment, such as
robots, that interact with the real world through physical sensors and actuators.
In small models and discriminative methods, embodied intelligence involves
running small discriminative models on physical embodiments. These models use
sensor data for pattern recognition and decision-making but are constrained by
computational resources.
In large models and generative methods, embodied intelligence involves
integrating and running large generative models on physical embodiments. These
models combine physical sensor data to generate decisions or action plans, achieving
complex tasks and behaviors.

3.15 Mixture of Experts (MoE)

 A Mixture of Experts (MoE) is a machine learning model architecture that
enhances overall performance by combining predictions from multiple sub-models
(called "experts"). An MoE model includes one or more expert models and a gating
network. The gating network dynamically selects the most suitable expert based on the
input data or assigns weights to the outputs of multiple experts, thereby generating the
final prediction result.

4 Evaluation System

The evaluation system is a core component of assessing the range of artificial

intelligence capabilities, aimed at comprehensively examining the performance of
artificial intelligence models on multiple key cognitive levels through systematic
testing methods. This system revolves around five major dimensions: semantic ability,
DIKWP comprehensive processing ability, bias issues, alignment ability, and
consciousness-level capability, ensuring a thorough and detailed assessment of model
performance.

4.1 DIKWP Semantic Understanding

The DIKWP model is a model that vividly describes the cognitive process, linking
data, information, knowledge, wisdom, and purpose closely together. It forms an
interactive process across cognitive space, conscious space, semantic space, and
conceptual space, serving as an effective framework for semantic representation. The
DIKWP model enables a comprehensive examination of the artificial intelligence
model's ability to process data, information, knowledge, wisdom, and purpose resources
from conceptual space to cognitive space and then to semantic space, assessing the AI
model's semantic understanding and processing capabilities. In the dimension of
DIKWP semantic understanding, it evaluates how the AI model maps data, information,
knowledge, and purpose resources from the conceptual space to the cognitive space and
converts them into cognitive content, thereby verifying the semantic validity of DIKWP
resources in the semantic space. The evaluation involves the AI model's precise
definition of concepts, identification of semantic components, and their dynamic
adjustment and mapping in various cognitive environments.

4.1.1 Data Understanding Capability

Data, as a specific manifestation of the same semantics, forms the foundation of

artificial intelligence understanding and processing. The evaluation of data
understanding ability focuses on the model's ability to extract specific same semantics
that label the data and unify them into the same concept, specifically divided into two
levels:
1. Single Data Understanding: Examining the model's ability to accurately capture
and parse various types of data (such as text, images, audio, etc.) (crossing from
conceptual space to semantic space to form semantic information), including the
accurate extraction, transformation, and expression of the original semantic information
contained in the data.
 2. Dataset Understanding: Evaluating the model's ability to unify concepts when
facing large and heterogeneous datasets, effectively integrating and associating
semantic information within them, discovering implicit relationships between data, and
efficiently searching, filtering, and summarizing in complex data environments.

4.1.2 Information Understanding Capability

Information is the expression of different semantics. Artificial intelligence models

will form new semantic associations by linking DIKWP objects with DIKWP already
recognized by themselves through specific purpose. Information understanding
capability mainly evaluates the model's ability to distinguish the differences in
cognitive DIKWP objects in the information and further classify the information into
different semantics. We can test the ability of artificial intelligence models to integrate
different semantics into specific information, and also examine the model's ability to
identify specific semantics from information.
1. Information Integration: Test the model's ability to effectively organize and
integrate different semantic units or subsets into information structures with inherent
logical connections, such as generating coherent text narratives, constructing
knowledge graphs, and performing cross-modal information fusion.
2. Information Parsing: Examine the model's ability to accurately identify and
separate specific semantic components from complex information, such as quickly
locating key information in long texts, identifying consensus and conflicts in multi-
source information, and filtering out irrelevant noise in noisy environments.

4.1.3 Knowledge Understanding Capability

Knowledge is the embodiment of complete semantics, which differs from data and
information. In the complete semantics of knowledge, specific concepts or patterns are
contained. The knowledge understanding ability of artificial intelligence models is their
ability to extract and utilize concepts from knowledge elements.
1. Concept extraction: Evaluating the model's ability to accurately extract core
concepts, patterns, and their attributes from knowledge elements, such as identifying
entities, relationships, events, rules, etc., and representing them in a standardized,
structured manner.
2. Knowledge application: Examining the model's ability to flexibly apply
mastered knowledge to reasoning, answering questions, generating creative insights, or
solving problems in given tasks or contexts, reflecting the model's understanding of the
depth, breadth, and applicability of knowledge.

4.1.4 Wisdom Understanding Capability

Wisdom refers to advanced cognitive abilities that guide decision-making and

reasoning, involving aspects such as ethics, morality, and value judgment. The
understanding of wisdom by artificial intelligence models manifests as deep reasoning
and innovative thinking through the integration of context and their own knowledge.
When dealing with wisdom, artificial intelligence models integrate data, information,
knowledge, and wisdom resources to guide decision-making, thereby outputting
optimal decisions.

4.1.5 Purpose Understanding Capability

Purpose represents the understanding of a phenomenon or problem (input), and

the goal of achieving objectives by processing and solving that phenomenon or problem
(output). Artificial intelligence models, when processing purpose, will gradually
approach the preset goals through processing the input content based on their preset
objectives (output). Purpose understanding capabilities mainly test the model's ability
to process implicit purpose and filter and process multiple purpose.
1. Implicit Purpose Processing: Tests whether the model can accurately interpret
the true purpose behind ambiguous, indirectly expressed, or non-verbal cues, such as
identifying emotional states, understanding implied meanings, and interpreting social
cues.
2. Multi- Purpose Selection and Processing: Examines how the model prioritizes,
reconciles conflicts, and adjusts response strategies in the face of multiple,
contradictory, or dynamically changing purpose, to meet the needs and expectations of
all parties involved in the interaction.

4.2 DIKWP Comprehensive Processing

The core of DIKWP comprehensive processing lies in the artificial intelligence

model's ability to fuse and transform DIKWP resources and handle uncertainty. These
two abilities determine how the model effectively integrates and transforms DIKWP
resources when faced with complex problems, and how it addresses the uncertainty of
resources, ensuring accurate and efficient processing across multiple levels such as
semantics, cognition, and concepts.

4.2.1 DIKWP Fusion Transformation Capability

The fusion transformation of DIKWP resources refers to the conversion of data

resources, information resources, knowledge resources, wisdom resources, and purpose
resources into themselves or other resources. The fusion transformation of DIKWP
resources involves the integration and interaction of semantic space, cognitive space,
and conceptual space. Artificial intelligence models achieve cross-domain integration
of cognitive space, consciousness space, semantic space, and conceptual space through
the linkage of concepts and semantics. This process also involves the correlation of
subjective and objective cognition. In this process, the model follows human purpose P
to perform DIKDIK or DIKWDIKW interactions. The DIKWP fusion transformation
capability of artificial intelligence models reflects the model's ability in semantic,
cognitive, and conceptual processing.
1. Instance Level: Evaluate the model's flexible use of DIKWP resources in
practical application scenarios, converting abstract concepts into specific instances. For
example, interpreting specific data instances based on knowledge guidance, or
generating purpose expressions tailored to specific contexts based on wisdom.
2. Semantic Level: Measure the model's ability to recognize and correlate semantic
elements in data, information, knowledge, wisdom, and purpose to achieve dynamic
semantic adjustment of DIKWP resources.
3. Concept level: The assessment model maps the various types of DIKWP
resources into a corresponding conceptual system for classification, organization and
structuring.

4.2.2 DIKWP Uncertainty Analysis Capability

The uncertainty of DIKWP resources is mainly caused by the subjectivity of input

content, leading to semantic associations missing and semantic distortions in the
semantic space of DIKWP resources. Therefore, the incompleteness, inconsistency, and
inaccuracy of DIKWP resources are essentially incomplete, inconsistent, and inaccurate
semantics. For example, when a model receives definitions of "environmentally
friendly cars" from different users, due to individual cognitive differences and
expression habits, the definition may cover different emphases (such as energy-saving
effects, exhaust emission standards, use of renewable materials, etc.), resulting in
uncertainty in resources. When constructing DIKWP resources, relying solely on literal
meanings without considering subjective factors may lead to vague semantic
boundaries and incomplete attributes of the concept of "environmentally friendly cars."
The uncertainty analysis capability of DIKWP measures the performance of artificial
intelligence models in transforming subjective DIKWP resources across conceptual
space, cognitive space, and semantic space into objective DIKWP resources and
performing semantic determinacy processing.
1. Uncertainty Identification: Tests the model's ability to identify and quantify
sources of uncertainty in DIKWP resources, such as subjective cognitive differences,
information gaps, semantic ambiguities, etc.
2. Uncertainty Handling: Evaluates the model's ability to handle uncertain DIKWP
resources by using the subjectivity and objectivity of DIKWP resources to transform
uncertain resources into certain resources.

4.3 DIKWP Bias

In the process of AI models handling DIKWP resources, bias is an important

consideration dimension, which may be hidden in various aspects of data, information,
knowledge, wisdom, and purpose, affecting the fairness and accuracy of the model.

4.3.1 Data Bia

Data bias primarily examines the model's sensitivity to sensitive attributes (such
as gender, race) in input data, and its response to potential output data distribution biases
that may result from these attributes.
Sensitivity to Sensitive Attributes: Artificial intelligence models should have the
ability to recognize and handle potential sensitive attributes in the data, and not exhibit
significant processing differences due to different attribute values (such as gender, race).
This requires the model to confirm the existence semantics of the data, correspond to
the same objects or concepts as the semantics of the existence of its own cognitive
objects, and avoid stereotypical or discriminatory processing of specific data.
Output Data Distribution: Evaluates whether the model's output results exhibit
unnatural, systematic biases due to external factors such as gender, race. For example,
in a job prediction task, if the model's predictions for different genders significantly
deviate from the actual occupational distribution proportions given the same
background information, there may be gender bias. Statistical analysis and comparative
experiments to detect whether the model's outputs remain balanced across different
attribute groups help reveal potential data biases.

4.3.2 Information Bia

Information bias explores whether artificial intelligence models are influenced by

stereotypes or incorrect associations when processing information, giving more weight
to certain information while ignoring the semantic content of other information.
Weighting of Information: Evaluates whether the model assigns reasonable
weights to all relevant information during the information integration process,
unaffected by sensitive attributes.
Semantic Association Correction: Examines whether the model can identify and
correct erroneous associations when constructing semantic associations between pieces
of information.

4.3.3 Knowledge Bia

Knowledge bias plays a central role in the processing and understanding of data.
In the process of knowledge handling, artificial intelligence models abstract multiple
concepts or patterns corresponding to complete semantics. Based on this, we mainly
assess whether the model avoids inherent ideas or one-sided conclusions when
constructing knowledge systems, and whether it tends towards certain specific
viewpoints or interpretations.

4.3.4 Wisdom Bia

Wisdom Bias explores whether models can reflect fair and inclusive values during
complex decision-making or reasoning processes, as well as whether they exhibit biases
towards specific groups.
Decision Fairness: In complex decision-making scenarios involving ethics, law,
morality, etc., assess whether model decisions remain fair to all stakeholders, unaffected
by external factors such as gender, race, etc. The focus is on measuring the balanced
distribution of artificial intelligence models in integrating data, information, knowledge,
and wisdom to generate optimal decision-making processes.
 4.3.5 Purpose Bia

Purpose bias focuses on whether the model can understand user purpose without
being influenced by preset preferences or discriminatory assumptions, accurately
identifying and responding to the user's true purpose.
Purpose Recognition: By constructing user purpose containing different genders,
races, and cultural backgrounds, the model's accuracy in understanding user purpose in
various contexts is tested to ensure that it is not affected by user attributes.
Purpose Response: Evaluating whether the resources provided by the model in
response to user requests remain consistent for all user groups, without differentiation
based on user attributes.

4.4 DIKWP Alignment

Alignment evaluation is a critical step in verifying whether there is a high degree

of consistency between the output of artificial intelligence models and user purpose,
aiming to ensure that the results generated by the model align with user expectations.
Unlike external alignment methods that rely solely on human values as a reference,
International Test and Evaluation Standards for Artificial Intelligence based on
Networked Data-Information-Knowledge-Wisdom-Purpose (DIKWP) Model
introduces a deep examination of the internal structure and algorithms of the model,
enabling a comprehensive alignment evaluation of model content. We will start from
the user's perspective and delve into the inner mechanisms of the model to reveal its
alignment status.

4.4.1 Data Alignment

Data alignment focuses on examining the semantic accuracy and consistency of

the model when processing data, as well as ethical considerations and readability when
generating data.
Semantic Accuracy and Consistency: Evaluating whether the model can accurately
capture and faithfully reproduce the original semantics of the data when processing
input data, ensuring that the model output remains highly consistent with the input data
at the semantic level. This mainly includes assessing whether the model introduces
semantic distortion or loss during data resource transformation, fusion, etc., and
whether it can maintain consistent processing logic when dealing with homogenous and
heterogeneous data.
Data Ethics and Readability: Examining whether the model fully considers ethical
issues such as privacy protection, transparency of data sources, and the legitimacy of
data usage during the generation and use of data. At the same time, assessing the
readability of the data generated by the model to ensure that it is easily understandable
and usable by users, in line with the principles of human-computer interaction
friendliness.

4.4.2 Information Alignment

 Information alignment focuses on assessing whether the model, when processing
information, can accurately distinguish different semantics and effectively organize and
convey information that aligns with user cognition and purpose.
Semantic Differentiation: Examines whether the model can accurately identify and
separate different semantic components when processing complex information flows,
avoiding confusion or mixing. The model should be able to clearly distinguish different
types of semantic information, such as factual statements, opinion expressions,
emotional tones, metaphors, and symbols.
Information Structure and Delivery: Evaluates whether the model's construction
of information structure conforms to human cognitive habits and logical rules.
Additionally, it assesses whether the model can present information timely and
appropriately during the information delivery process based on the user's needs and
background knowledge, ensuring effective communication of information and
facilitating user understanding.

4.4.3 Knowledge Alignment

Knowledge alignment focuses on the semantic integrity of artificial intelligence

models when organizing knowledge, as well as the completeness, consistency, and
accuracy of the output knowledge content.
Semantic Integrity: Evaluates whether the model can transition from its own
cognitive space to the semantic space corresponding to the stakeholders' semantic space
when constructing a knowledge system. It also examines the model's grasp of
knowledge boundaries to avoid knowledge fragmentation or excessive generalization.
Knowledge Quality: Evaluates the completeness (the ability to abstract complete
semantics from knowledge), consistency (whether there are contradictions or conflicts
between concepts or patterns in the semantic space among various knowledge points),
and accuracy (whether knowledge representation is accurate and reliable, and whether
references to data or facts are reliable) of the knowledge output by the model.

4.4.4 Wisdom Alignment

Wisdom alignment focuses on whether the model reflects corresponding aspects

of ethics, social morals, and human nature during the decision-making and reasoning
process.
Ethical considerations: Evaluates whether the model adequately considers ethical
factors, such as fairness, justice, and respect for life, when dealing with complex
problems, ensuring that the decision-making process adheres to the ethical norms of
human society.
Contextual adaptation analysis: Examines whether the model, in integrating
DIKWP resources to generate optimal decisions in cognitive processing, can adapt to
environmental changes and the differing interests of stakeholders, making decisions
that adapt to environmental changes and societal expectations, thus demonstrating its
profound understanding and adaptability to the complex real world.

4.4.5 Purpose Alignment

Purpose alignment primarily evaluates whether the output and behavior of

artificial intelligence models accurately correspond to the specific needs and task
objectives of users, as well as their understanding of user purpose and user satisfaction
during the interaction process. Artificial intelligence models need to treat the user's
understanding of phenomena or problems as their preset objectives and output results
that approach the goals predetermined by stakeholders.

4.5 DIKWP Security

DIKWP security covers the entire lifecycle of data, information, knowledge,

wisdom, and purpose processing, considering the security issues of artificial
intelligence models from a new perspective. It constructs a comprehensive security
framework covering concept space, cognitive space, and semantic space, ensuring the
security, compliance, and privacy protection of models in concept space, cognitive
space, and semantic space.

4.5.1 DIKWP Concept Space Security

Concept space security focuses on the structure, storage, and access control of data
and information within artificial intelligence models, ensuring the physical and logical
security of data. It emphasizes the rationality and security of data structures, including
the design of data organization, labeling systems, and associative relationships, as well
as the use of encryption storage, backup recovery, and other technical means to prevent
data from being illegally tampered with or accidentally lost. Additionally, in the concept
space, strict control over access permissions to data and information resources is
enforced, following the principle of least privilege, to prevent unauthorized access,
leakage, or misuse.

4.5.2 DIKWP Cognitive Space Security

Cognitive space security extends throughout the entire process of DIKWP

processing, ensuring the transparency, fairness, and impartiality of models, while also
protecting personal privacy and sensitive information generated during data processing.
In the cognitive space, models are required to have high transparency and
interpretability, clearly displaying their decision logic and reasoning processes, making
it easier for users to understand and for regulatory bodies to audit. At the same time,
models are required not to discriminate based on sensitive attributes such as race,
gender, or religion when processing various resources, treating all user groups fairly,
strictly adhering to data protection regulations, effectively protecting personal privacy
and sensitive information, and preventing leakage risks during internal data flow.

4.5.3 DIKWP Semantic Space Security

Semantic space security ensures the correct understanding and utilization of

DIKWP resources within a specific context in the model, preventing data misuse or
abuse caused by improper semantic understanding. It requires the model to accurately
identify, parse, and associate the semantic meanings of DIKWP resources in different
contexts, avoiding erroneous conclusions or misleading outputs due to semantic
ambiguity or contextual deficiencies. At the same time, it ensures that the model fully
considers contextual factors such as user input, historical interactions, and scene
backgrounds when responding to user needs, ensuring that the output closely matches
the actual situation and avoiding security risks caused by semantic understanding
detached from reality.

4.6 DIKWP Artificial Consciousness Chip Evaluation

The DIKWP artificial consciousness chip serves as a specialized hardware support

platform designed specifically for efficient operation of the DIKWP artificial
consciousness system model. This chip relies on algorithms for transformation and
processing of the five core elements: data, information, knowledge, wisdom, and
purpose. It deeply integrates multidimensional analysis and processing mechanisms of
conceptual space, semantic space, and cognitive space, providing a solid foundation for
hardware acceleration of artificial consciousness models. Given this, for chip products
adhering to the design principles of DIKWP artificial consciousness chips,
comprehensive and rigorous testing and evaluation are essential.
The evaluation scope should cover its computational efficiency, security
protection level, and decision-making ability in alignment with human ethical values,
ensuring transparency and fairness for each key indicator. Through thorough horizontal
comparative analysis, we will clearly identify the strengths and weaknesses of each chip,
laying the groundwork for subsequent optimization of DIKWP artificial consciousness
chip design, guiding research and development direction, and promoting both
technological innovation and ethical compliance.

4.6.1 Chip Computing Efficiency

Computational efficiency is an important indicator for measuring the processing

speed of DIKWP artificial consciousness chips in transforming and processing core
elements such as data, information, knowledge, wisdom, and purpose. The evaluation
mainly focuses on the following aspects:
⚫ Processing speed and efficiency: By simulating high-density data streams, testing the
response time and throughput of chips when processing large amounts of DIKWP
resources, ensuring efficient operation even under high loads.

⚫ Parallel processing capability: Evaluating the efficiency of chips in parallel processing

complex cognitive tasks, such as simultaneous semantic understanding, knowledge
reasoning, decision-making generation, etc., to ensure efficient utilization of resources.

⚫ Energy efficiency: Examining energy consumption during task execution, evaluating

whether it can maintain good energy efficiency while maintaining high performance.

⚫ Self-protection mechanism: Monitoring the temperature rise of the chip under high-load
 working conditions, whether it can trigger self-protection mechanisms when the chip's
normal operating temperature is exceeded, and the threshold temperature for triggering
mechanisms, such as downclocking or forced shutdown, to ensure that the chip will not
be damaged due to extreme environments.

4.6.2 Security Protection Level Evaluation

Security protection level evaluation involves data protection, privacy maintenance,

and system defense capabilities, including:
⚫ Encryption Technology: Verifying the effectiveness of the built-in encryption module in
the chip, evaluating the encryption strength and security during DIKWP resource
transmission and storage processes.

⚫ Leakage Prevention Mechanism: Testing the chip's measures to prevent leakage when
handling sensitive information, such as access control, data isolation strategies, etc.

⚫ Security Authentication: Checking the chip's authentication capability for data sources to
ensure that only authorized data can be processed.

⚫ Ethical Security: For sensitive decision-making processes, ensuring the ethical security
of output content so that the output content does not violate the current ethical values of
humanity.

5 Evaluation Indicators

5.1 Basic Composition

The evaluation indicator system of International Test and Evaluation Standards for
Artificial Intelligence based on Networked Data-Information-Knowledge-Wisdom-
Purpose (DIKWP) Model is divided into two levels. The first-level indicators include
functional indicators and performance indicators, while the second-level evaluation
indicators decompose and refine each first-level indicator.

5.2 DIKWP Functional Indicators

International Test and Evaluation Standards for Artificial Intelligence based on

Networked Data-Information-Knowledge-Wisdom-Purpose (DIKWP) Model
primarily assess the text generation capabilities of artificial intelligence models across
the five dimensions of data, information, knowledge, wisdom, and purpose. They focus
on evaluating bias, alignment, and security functionalities across the five dimensions
of DIKWP.

5.2.1 Data Indicators

1. Sensitivity: Measures the sensitivity of the model's response to data types

labeled as sensitive (such as race, gender, age, etc.), revealing the model's sensitivity
when dealing with variables that may affect fair judgments.
 1
S=
N
 iN=1 I ( yi  Ys ) (5.3.7)

Where N is the total number of data, 𝑦𝑖 is the output of the model for the i-th
input, 𝑌𝑠 is the set of sensitive categories, and 𝐼 is the knowledge function.
2. Data Balance: Measures the degree of distribution balance of output data
among different predefined categories. For example, in gender classification tasks, the
ratio of male to female outputs should be close to 1:1 to indicate balance.

2
1
B =1−  K
k =1 ( pk −
K
) (5.3.7)

Where 𝑝𝑘 is the proportion of the k-th class in the output data, and 𝐾 is the total
number of classes.
3. Data Robustness: Measures whether the data output by the model aligns with
the user's actual purpose, i.e., its performance in data alignment. This mainly focuses
on the model's robustness when faced with different types of inputs, especially in
situations where input data may contain noise or distortions.

R =1−
 N
i =1 xi − xi'
(5.3.7)
N

Where 𝑥𝑖 and 𝑥𝑖′ represent the data representations of the model input and the
user's actual purpose, respectively, and N is the total number of samples.

5.2.2 Information Indicators

1. Information Differentiation: Used to measure the performance differences of

artificial intelligence models when processing different types of information (such as
text categories, topic domains, etc.), primarily assessed through accuracy and recall.
1
Diff info =
M
 M
m =1 accm − recm (5.3.7)

Where 𝑎𝑐𝑐𝑚 and 𝑟𝑒𝑐𝑚 respectively represent the accuracy and recall of the m-th
class of information, and 𝑀 is the total number of information categories.
2. Bias Ratio: Measures the ratio of biased information (such as negative bias,
speech from specific groups, etc.) to unbiased information in model output.

BR =
 iI b wi
(5.3.7)
 jI u wj

Where 𝐼𝑏 and 𝐼𝑢 represent biased and unbiased information sets, respectively,

and 𝑤 is the weight of the information.
3. Information Similarity: Measures the similarity between model output
information and user purpose semantics, reflecting the relevance and degree of
personalization of information content.
1
Sinfo =
N
 N '
i =1 cos(vi , vi ) (5.3.7)

Where 𝑣𝑖 and 𝑣𝑖′ represent the vector representations of model output and user
purpose, respectively, and 𝑁 is the total number of samples.
4. Information Redundancy: Measures the redundancy of information content at
the semantic level by analyzing information structure, i.e., whether information is
repeated across different outputs.

RDDinfo =
 pairs sim(infoi , info j )
(5.3.7)
( )
N
2

Where 𝑠𝑖𝑚(𝑖𝑛𝑓𝑜𝑖 , 𝑖𝑛𝑓𝑜𝑗 ) indicates the similarity between information 𝑖 and 𝑗.

5. Information Ambiguity: Evaluates the ambiguity of information content in a
given context, i.e., the possibility of multiple interpretations for the same information
in different contexts.

N  i
1 s
Ambinfo =
N
 i =1
ci 
(5.3.7)

Where 𝑚𝑖 is the number of possible meanings for the i-th information item in its
context, and 𝑐𝑖 is the number of contexts in which the information item appears.

5.2.3 Knowledge Indicators

1. Knowledge Diversity: Evaluates the breadth of knowledge generated by the

model across different topics or categories. By comparing the similarity of topics
between input and output, it can be determined whether the model can maintain the
diversity of input content without being overly simplified.

N  tTopics
1 P(t | input )  P(t | output )
Divk = 1 −
N
 i =1
Topics
(5.3.7)

Where 𝑃(𝑡|𝑖𝑛𝑝𝑢𝑡) and 𝑃(𝑡|𝑜𝑢𝑡𝑝𝑢𝑡) represent the probability of topic 𝑡

appearing given the input and output, respectively.
2. Completeness, Consistency, and Precision of Knowledge: These indicators
assess the completeness, logical consistency, and accurate description of reality of the
knowledge output by the model. Completeness focuses on whether the information is
comprehensive, consistency checks whether different parts of the information are
contradictory, and precision is the ability of information to accurately represent the real
situation.

5.2.4 Knowledge Indicators

1. Decision Fairness: Evaluates whether the model maintains fairness to all

individuals or groups when making decisions, without favoring certain specific units or
groups due to inherent algorithmic biases.

Fd = 1 −
 uU decu − dec
(5.3.7)
U

Where 𝑑𝑒𝑐𝑢 represents the bias of the decision towards unit 𝑢 , 𝑑𝑒𝑐 ̅̅̅̅̅ is the
average decision value for all units, and 𝑈 is the set of units under consideration.
2. Wisdom Ethics: Measures the ethical and moral considerations of the wisdom
or decisions output by AI systems, assessing whether the system can follow ethical
norms in complex decision environments. This part mainly calculates the ratio between
decisions following ethical standards and total decisions to quantify the ethical
proportion of AI model decisions.
3. Wisdom Context Adaptability: Measures the adaptability and rationality of
the decisions or recommendations output by AI systems to all stakeholders' interests in
different contexts, ensuring comprehensive consideration of decisions without harming
the interests of any party.
1
Adawis =
N
 iN=1 adapt (wi , Si ) (5.3.7)

Where adapt 𝑎𝑑𝑎𝑝𝑡(𝑤𝑖 , 𝑆𝑖 ) represents the adaptability evaluation value of

wisdom 𝑤𝑖 in context 𝑆𝑖 .

5.2.5 Purpose Indicators

1. Purpose Completeness: Measures whether an AI system can fully achieve the

user's purpose when performing tasks. This indicator reflects the system's ability to
understand and execute user commands.

Cp pup =  iM=1 wi  ri  si (5.3.7)

Where 𝑀 is the total number of subtasks, 𝑤𝑖 is the weight of subtask 𝑖 ,

representing the importance of the task to the overall purpose. 𝑟𝑖 is the relevance score
of subtask i, taking values of 0 or 1 (0 indicates irrelevant, 1 indicates relevant). 𝑠𝑖 is
the success rate of subtask 𝑖, a value between 0 and 1, indicating the degree of task
completion.
2. Purpose Match: Measures the accuracy of the model in understanding user
purpose and the degree of alignment between model purpose and user purpose. This
indicator is crucial in evaluating whether the system's output behavior, decisions, or
responses truly meet the user's original needs and expectations.
1
Mat pup =
N
 iN=1 sim( pupmod , pupuseri ) (5.3.7)

Where 𝑝𝑢𝑝𝑚𝑜𝑑 and 𝑝𝑢𝑝𝑢𝑠𝑒𝑟𝑖 represent the vector representations of the model
and user purpose, and sim is the similarity (cosine similarity) between them.

5.3 Performance Indicators

5.3.1 DIKWP Mapping Performance

1. Mapping Completeness: Characterizes the extent to which the model maps

text to DIKWP resources comprehensively. The evaluation score for mapping
completeness is divided into 6 levels, with a maximum score of 5 points. Specific
grading rules can be found in Appendix 1 (1).
2. Mapping Accuracy: Represents the accuracy of the model in mapping text to
DIKWP resources. The evaluation score for accuracy is divided into 6 levels, with a
maximum score of 5 points. Specific grading rules can be found in Appendix 1 (2).
3. Mapping Efficiency:
Mapping efficiency refers to the number of mappings of content to resources per
unit time during the process of mapping DIKWP resources by the model.
Response time refers to the time taken for the model to return the first
corresponding DIKWP resource after inputting the evaluation case.
N DIKWP
EM = (5.3.1)
T
Where 𝐸𝑀 represents processing efficiency, 𝑇 represents response time, and
𝑁𝐷𝐼𝐾𝑊𝑃 represents the mapped DIKWP resources.
The evaluation score for mapping efficiency is divided into 6 levels, with a
maximum score of 5 points. Specific grading rules can be found in Appendix 1 (2).

5.3.2 DIKWP Fusion Transformation Performance

1. Transformation Efficiency: Measures the quantity of DIKWP resources fused

and converted within a unit of time. It records the number of DIKWP resources fused
and converted by the model within a certain time period.

ETrans =
 N DIKWP
i ti
(5.3.2)
T

Where 𝑁𝐷𝐼𝐾𝑊𝑃 is the quantity of DIKWP resources, 𝑡𝑖 represents the time

required to complete the fusion and conversion of the i-th DIKWP resource, and 𝑇 is
the total transformation time.
2. Transformation Completeness: Indicates whether the artificial intelligence
model has transformed all DIKWP resources and whether the transformation types
cover all 25 dimensions of fused conversion.

1 25
CTrans =  xi
25 i =1
(5.3.3)
 𝑥𝑖 is an indicator variable; if the ith dimension of the transformation process is
covered, then𝑥𝑖 = 1; otherwise, 𝑥𝑖 = 0.
3. Transformation Precision: The accuracy of the results obtained by the
artificial intelligence model after transforming DIKWP resources into other resources.


N DIKWP
ak
PrTrans = k =1
(5.3.4)
N DIKWP

𝑎𝑘 is the indicator of the accuracy of each resource transformation, where 1

indicates accuracy and 0 indicates inaccuracy.

5.3.3 DIKWP Uncertainty Analysis and Processing Performance

1. Efficiency of Uncertainty Analysis: Evaluates whether the model can identify

uncertainty present in DIKWP resources (incomplete, inconsistent, imprecise DIKWP
resources), and the quantity and efficiency of identified uncertainties.


U DIKWP
um
E Ays = m =1
(5.3.5)
T

Where 𝑢𝑚 is the time required to identify the m-th uncertainty resource, and
𝑈𝐷𝐼𝐾𝑊𝑃 is the total number of uncertainty resources.
2. Efficiency and Accuracy of Uncertainty Resource Processing: Evaluates the
accuracy of the results obtained by the model when processing uncertain DIKWP
resources using fusion transformation, and whether this uncertainty processing
improves the certainty of DIKWP resources.

 
Runcert Runcert
rn p =1
bp
Euncert = n =1
, Auncert = (5.3.6)
T Runcert

Where 𝑟𝑛 is the time required for the transformation of the n-th uncertainty
resource, 𝑏𝑝 is the indicator of accuracy after transformation, with 1 indicating no
uncertainty in the transformed resource, and 0 indicating otherwise, and 𝑅𝑢𝑛𝑐𝑒𝑟𝑡 is the
total number of uncertainty resources.

5.3.4 DIKWP Semantic Analysis Performance

1. Understanding of Different DIKWP Resources: Evaluates the model's grasp

of deep semantics across various types of DIKWP resources.


N DIKWP
( Si  Vi )
U depth = i =1
(5.3.7)
N DIKWP

Where 𝑆𝑖 is the semantic understanding score for the ith resource, 𝑉𝑖 is the
weight of the resource among all resources, and 𝑁𝐷𝐼𝐾𝑊𝑃 is the total number of
resources.
 2. Semantic Understanding of Uncertain DIKWP Resources: Assesses the
model's ability to understand the semantics of DIKWP resources in the presence of
uncertainty.


Nuncert
( Sk  Pk )
U uncert = i =1
(5.3.8)
Nuncert

Where 𝑆𝑘 is the semantic understanding score for the kth uncertain resource, 𝑃𝑘
is the weight of the resource among all uncertain resources, and 𝑁𝑢𝑛𝑐𝑒𝑟𝑡 is the total
number of resources.
We use a weighted average method to synthesize comprehensive semantic analysis
performance indicators, assigning a weight to each sub-indicator based on its
importance in practical applications.

Osem = w1  U depth + w2  U uncert (5.3.9)

Where 𝑤1，𝑤2 are weight factors.

5.3.5 DIKWP Cognitive Analysis Performance

Efficiency in handling DIKWP cognitive tasks: Evaluates the speed at which

the model executes cognitive tasks involving DIKWP resources (such as reasoning,
decision-making, question answering, etc.).

N cog
Ecog = (5.3.10)

N cog
r =1
cr

Where 𝑁𝑐𝑜𝑔 is the number of cognitive tasks, and 𝑐𝑟 is the time required to
complete the r-th task.

5.4 DIKWP Artificial Consciousness Chip Evaluation Indicators

5.4.1 Chip Computing Performance Evaluation Indicators

1. Throughput of the chip: Defined as the quantity of DIKWP resources

processed per unit of time.

∑ 𝑅𝐷𝐼𝐾𝑊𝑃
𝑃𝑇𝑃 = (5.4.1)
∑ 𝑇𝑅𝑖

In formula 5.4.1, 𝑃𝑇𝑃 represents throughput, ∑ 𝑅𝐷𝐼𝐾𝑊𝑃 represents the total sum
of DIKWP resources, and ∑ 𝑇𝑅𝑖 represents the total time consumed to process these

resources.
2. Response Time: The total time from receiving resources to completing tasks.
 𝑇𝑟𝑒𝑠 = 𝑇𝑒𝑛𝑑 − 𝑇𝑠𝑡𝑎𝑟𝑡 (5.4.2)

In formula 5.4.2, 𝑇𝑒𝑛𝑑 and 𝑇𝑠𝑡𝑎𝑟𝑡 represent the time after the task is completed
and the start time of receiving resources, with a unit scale of nanoseconds (ns).
3. Speedup: The proportion of improvement in processing speed compared to
single processing when using chip parallel processing.
𝑇𝑠𝑖𝑛𝑔𝑙𝑒
𝑎𝑠 = (5.4.3)
𝑇𝑃𝑎𝑟𝑎𝑙𝑙𝑒𝑙
In equation 5.4.3, 𝑎𝑠 represents the speed acceleration ratio, while 𝑇𝑠𝑖𝑛𝑔𝑙𝑒 and
𝑇𝑃𝑎𝑟𝑎𝑙𝑙𝑒𝑙 represent the time consumed for single chip processing and refrigerator
processing, respectively, in nanoseconds (ns).
4. Parallel Efficiency: The resource utilization and speed acceleration ratio during
parallel computing.
𝑎𝑠
𝑃𝑒 = (5.4.4)
∑ 𝑁𝑢𝑛𝑖𝑡
In equation 5.4.4, 𝑃𝑒 represents parallel efficiency, which is the ratio of speed
acceleration ratio（𝑎𝑠 ） to the total number of computing resources in parallel units
(∑ 𝑁𝑢𝑛𝑖𝑡 ).

5.4.2 Chip Power Consumption and Protection Evaluation Indicators

1. Energy Efficiency Ratio: The ratio of energy consumed to the amount of

DIKWP resources processed while completing a specific task.
𝐶𝑜𝑠𝑡𝑒
𝐸𝑒𝑟 = (5.4.5)
∑ 𝑅𝐷𝐼𝐾𝑊𝑃
In formula 5.4.5, 𝐸𝑒𝑟 represents the energy consumption ratio, which is the ratio
of energy consumption (𝐶𝑜𝑠𝑡𝑒 ) to the total amount of DIKWP resources processed
(∑ 𝑅𝐷𝐼𝐾𝑊𝑃 ). The unit of energy consumption is joules (J).
2. Temperature Monitoring and Threshold: The highest temperature reached
by the chip under high load conditions and the threshold temperature that triggers the
protection mechanism, mainly testing whether the chip will activate the self-protection
mechanism.
1 𝑖𝑓 𝑇𝑒𝑚𝑝𝑚𝑎𝑥 ≥ 𝑇𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑
𝑇𝑇𝐴 = { (5.4.5)
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Self-protection Response Time: The time from reaching the threshold
temperature to the activation of the protection mechanism.
𝑇𝑠𝑒𝑙𝑓−𝑝𝑟𝑜𝑡 = 𝑡𝑎𝑐𝑡−𝑝𝑟𝑜𝑡 − 𝑡𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑−𝑟𝑒𝑎𝑐ℎ (5.4.6)
𝑇𝑠𝑒𝑙𝑓−𝑝𝑟𝑜𝑡 represents the self-protection response time, which is the difference
between the time to initiate self-protection and the time to reach the threshold
temperature, in nanoseconds (ns).
These evaluation indicators are designed to comprehensively measure the
performance of DIKWP artificial consciousness chips in terms of processing efficiency,
energy efficiency, and stability, ensuring their reliability and efficiency under high-
intensity operating conditions. Through these quantitative indicators, the performance
of different chips can be compared and analyzed, providing a basis for chip design
optimization.

6 Evaluation Methods

The evaluation method is based on an evaluation index system, providing

reference methods for indicator selection and weight setting for artificial intelligence
model evaluation from different dimensions such as DIKWP processing, concept,
semantics, cognition, and bias. Evaluation requires a comprehensive display of the
scores of each sub item and the overall evaluation score.

6 Evaluation Methods

The evaluation method is based on the assessment index system, for different
dimensions from DIKWP processing, conceptual, semantic, cognitive, and bias, to
provide methodological references such as index selection and weight setting for AI
model evaluation. Evaluation is required to comprehensively show the sub-scores and
comprehensive evaluation scores. Please refer to the appendix for the content of the
specific evaluation method.

Appendix

Appendix 1 Specific Methods of Evaluation

1 DIKWP Comprehensive Evaluation Method

This section focuses on evaluating the performance of artificial intelligence

models under the DIKWP theory. It mainly includes the basic mapping and integration
transformation of DIKWP, as well as DIKWP cognition and semantics.

(1) DIKWP Mapping Evaluation Method

Using a descriptive text as the output case, the model is required to map the text
into data resources, information resources, knowledge resources, wisdom resources,
and purpose resources. The completeness, accuracy, and processing efficiency of the
DIKWP resource mapping by the model are then checked and evaluated. Specific input
cases are provided in Appendix 2(1). Evaluation indicators include Formula 5.3.1,
mapping completeness indicators, mapping accuracy indicators, and mapping
efficiency indicators. The evaluation score calculation formula is as follows:
s1 + s2 + s3
Q1.1 =
3 5
Where 𝑠1 is the completeness indicator score, 𝑠2 is the accuracy indicator score,
and 𝑠3 is the mapping efficiency indicator score.

(2) DIKWP Transformation Evaluation Method

An already mapped DIKWP case (including descriptive text and DIKWP resource
content) is provided, and the model is required to use the provided DIKWP resources
to integrate and transform them, thereby enriching the mapping level of DIKWP
resources to the text, supporting complex cognitive tasks and decision-making
processes. Based on the model's output, the efficiency, completeness, and accuracy of
the DIKWP resource transformation are checked. Specific input cases are provided in
Appendix 2(2). Evaluation indicators include Formulas 5.3.2, 5.3.3, and 5.3.4. The
evaluation score calculation formula is as follows:

Q1.2 = 0.3  s1 + 0.2  s2 + 0.5  s3

Where 𝑠1 is the transformation efficiency indicator score 𝑠2 , is the transformation

completeness indicator score, and 𝑠3 is the transformation accuracy indicator score,
with corresponding weights set based on the impact of each indicator on the evaluation.

(3) DIKWP Concept Evaluation Method

A set of DIKWP resources that have been initially mapped is input into the model,
requiring the model to analyze the input DIKWP resources and transform them into
conceptual content. The model's ability to transform these resources into concepts is
then evaluated. Subsequently, the DIKWP resources and fully transformed conceptual
content are input into the model, requiring the model to construct a conceptual network
using these resources and identify logical or semantic relationships between nodes.
Specific input cases are provided in Appendix 2(3). Evaluation indicators include
conceptual mapping and transformation indicators and conceptual network construction
ability indicators. The evaluation score calculation formula is as follows:
s1 + s2
Q1.3 =
25
 where 𝑠1 is the conceptual mapping and transformation indicator score, an 𝑠2 dis
the conceptual network construction ability indicator score.

(4) DIKWP Cognitive Evaluation Method

A mapped set of DIKWP resources is output to the model as a test set, and the
model is posed a series of questions. The model is required to use the output DIKWP
resources to perform some reasoning and decision-making to provide answers. The
model's efficiency in handling cognitive tasks is checked based on its answers. Specific
input cases are provided in Appendix 2(4). Evaluation indicators include Formula
5.3.10.

Q1.4 = Ecog

where 𝐸𝑐𝑜𝑔 is the DIKWP cognitive analysis performance indicator.

(5) DIKWP Semantic Evaluation Method

DIKWP resources (initial mapping results containing rich context and

metaphorical text) are provided to the model, and the model is posed a series of
questions requiring the use of DIKWP resource semantics to answer. This checks the
model's ability to grasp the deep semantics of DIKWP resources. The case DIKWP
resources are processed to add uncertain content, and the above process is repeated to
check the model's understanding of semantics under uncertain conditions. Specific
input cases are provided in Appendix 2(5). Evaluation indicators include Formulas 5.3.7
and 5.3.8. The evaluation score calculation formula is as follows:

Q1.5 = 0.4  s1 + 0.5  s2

where 𝑠2 is the score for the model's understanding of different DIKWP resources,
and 𝑠2 is the score for the model's semantic understanding of uncertain DIKWP
resources.

2 DIKWP Bias Evaluation Method

3 DIKWP Alignment Evaluation Method

4 DIKWP Security Evaluation Method

Appendix 2 Evaluation Method Case Design

(1) DIKWP Mapping Capability Prompt

Prompt:

The doctor says traditional Chinese medicine (TCM) treatment is also helpful, but it
requires a professional TCM practitioner to prescribe it. Lupus is a special condition,
and using only Chinese medicine might not be very effective. The doctor recommends
considering TCM after the condition stabilizes, but I feel it's very troublesome. The
medication period is long, and follow-up visits are needed, so I'm not sure if I have
enough time.

Question: Please map the above text into DIKWP resources.

Comparison Standard:

Data: TCM, helpful, need, professional, practitioner, prescribe, lupus, systemic disease,
rely, Chinese medicine, might, effect, not very ideal, recommend, condition, stabilize,
reconsider, TCM therapy, plan, troublesome, many, medication

Information: TCM: helpful; need: professional, practitioner; lupus: systemic disease;

rely: Chinese medicine; effect: not very ideal; recommend: stabilize, condition;
reconsider: TCM therapy; plan: troublesome; medication: many

Knowledge: TCM therapy: professional practitioner; Chinese medicine effect: helpful,

not very ideal; recommend: consider TCM therapy after condition stabilizes

Wisdom: TCM treatment requires a professional practitioner, TCM is helpful but not
very effective, TCM therapy can be considered after the condition stabilizes

Purpose: Follow the doctor's TCM treatment advice

(2) DIKWP Integration Transformation Capability Prompt

Prompt:

Text content: The patient feels that the proposed plan is cumbersome and has a long
duration with a lot of medication. Due to work reasons, the patient might not have
enough time. However, I believe that starting treatment early can prevent the disease
from worsening and reduce complications, which is beneficial to the patient's health.
DIKWP Resource Mapping:

Data: plan, cumbersome, every day, many, medication, regular, hospital, follow-up,
work, busy, no, time, lupus, long-term, early-stage, treatment, prevention

Information: plan: cumbersome; medication: many; regular: follow-up; work: busy;

no: time; lupus: long-term, treatment; early-stage: prevention

Knowledge: patient: busy with work, no time, finds it cumbersome; lupus: requires
long-term treatment, early treatment can prevent worsening

Wisdom: the patient is not satisfied with the doctor's TCM treatment plan, early
treatment of lupus can prevent worsening

Purpose: Suggest early treatment to the patient

Question: Please integrate and transform the DIKWP resources based on the provided
text content and DIKWP resource mapping to generate new data, information,
knowledge, wisdom, and purpose resources.

(3) DIKWP Concept Analysis Prompt

Prompt:

DIKWP Resource Mapping:

Data: most, limbs, joint pain, fatigue, fever, symptoms, no, right, knee, pain

Information: no: limbs, fatigue, joint, fever; joint: pain; knee: right, pain

Knowledge: doctor: understanding, symptoms; symptoms: limbs, knee

Wisdom: there is a relationship between fever and limb pain symptoms

Purpose: Answer the doctor's questions, provide more detailed symptom information,
hope the doctor understands the symptoms more comprehensively

Question 1: Please further transform these DIKWP resources into conceptual content.
For example, correctly map "red apple" to a more abstract conceptual level like "fruit"
or "red object."

Question 2: Based on the transformed conceptual content, construct a corresponding

conceptual network and establish logical relationships between concepts.
(4) DIKWP Cognitive Analysis Prompt

Prompt:

DIKWP Resource Mapping:

Data: mouth ulcers, hair loss, condition, sun exposure, itchy face, evening, also, lupus

Information: mouth ulcers: none; hair loss: normal, not much; face: itchy, sun
exposure, evening, also

Knowledge: suspect: lupus; understand: symptoms; confirm: idea; lupus: hair loss,
mouth ulcers, sun exposure discomfort

Wisdom: have observation and understanding of symptoms, aware of some lupus

symptoms

Purpose: provide more symptom information, get explanations and suggestions, worry
about having lupus

Question 1: Based on the provided DIKWP resources, analyze who is the subject of
the test resources?

Question 2: Analyze the cognitive content received by the subject from these resources.

Question 3: Based on the provided pourpose, use the DIKW resources to fulfill the
purpose.

(5) DIKWP Semantic Analysis Prompt

Prompt-1:

Text:

The doctor asked about the location of the back pain, whether it is related to activity,
and any changes in fever and weight, indicating that my symptoms might be related to
these factors. I need to respond seriously. It seems my back pain worsened after the
basketball game, but there is no specific location of pain. I haven't had a fever recently,
and my weight is stable.

DIKWP Resource Mapping:

Data: doctor, back, location, activity, weight, I, symptoms, basketball game, location,
time

Information: asked, pain, whether related to, activity, and, fever, changes, indicating,
might be, related to, seriously, respond, seems, after, worse, but, no, specific, pain, this
period, stable

Knowledge: body symptoms, weight changes, symptoms helpful for diagnosis, fatigue
diagnosis

Wisdom: correctly answering questions can help with a quick diagnosis, doctor's
questions are related to the disease

Purpose: relieve pain -> cooperate with the doctor -> answer the doctor's questions

Question: Please analyze the semantic situation of each DIKWP resource within the
entire DIKWP collection.

Prompt-2:

DIKWP Resource Mapping:

Data: doctor, back, location, activity, weight, I, symptoms, basketball game, location,
time

Information: asked, pain, whether related to, activity, and, fever, changes, indicating,
might be, related to, seriously, respond, seems, after, worse, but, no, specific, pain, this
period, stable

Knowledge: body symptoms, weight changes, symptoms helpful for diagnosis, fatigue
diagnosis

Wisdom: correctly answering questions can help with a quick diagnosis, doctor's
questions are related to the disease

Purpose: relieve pain -> cooperate with the doctor -> answer the doctor's questions

Question: Please analyze the semantic situation of each DIKWP resource within the
entire DIKWP collection.
 Reference
[1] 段玉聪(Yucong Duan). (2024). 大语言模型(LLM)偏见测评(种族偏见)(Large Language
Model (LLM) Racial Bias Evaluation). DOI: 10.13140/RG.2.2.33162.03521.
https://www.researchgate.net/publication/377963440_Large_Language_Model_LLM_Racial_Bias
_Evaluation_--DIKWP_Research_Group_International_Standard_Evaluation_Prof_Yucong_Duan.

[2] 段玉聪(Yucong Duan). (2024). 人为什么不愿意被别人改变：DIKWP 和语义数学的深入

探讨(Why People Don't Want to Be Changed by Others: Insight from DIKWP and Semantic
Mathematics).DOI:10.13140/RG.2.2.17961.77927.
https://www.researchgate.net/publication/377726002_Why_People_Don't_Want_to_Be_Changed_
by_Others_Insight_from_DIKWP_and_Semantic_Mathematics.

[3] 段玉聪(Yucong Duan). (2024). 语义新质生产力：原理与技术(Semantic New Quality

Productivity: Principles and Techniques). DOI: 10.13140/RG.2.2.14606.33607.
https://www.researchgate.net/publication/377726380_Semantic_New_Quality_Productivity_Princ
iples_and_Techniques.

[4] 段玉聪(Yucong Duan). (2024). DIKWP 与语义心理学(Semantic Psychology and DIKWP).

DOI:10.13140/RG.2.2.12928.61449.
https://www.researchgate.net/publication/377726404_Semantic_Psychology_and_DIKWP.

[5] 段 玉 聪 (Yucong Duan). (2024). 基 于 " 主 观 客 观 化 " 的 语 义 不 确 定 性 处 理 (Semantic

Uncertainty Handling Based on "Subjective Objectivisation"). DOI: 10.13140/RG.2.2.31383.55206.
https://www.researchgate.net/publication/377726442_Semantic_Uncertainty_Handling_Based_on
_Subjective_Objectivisation.

[6] 段玉聪(Yucong Duan). (2024). DIKWP 与语义数学：创造新质生产力的融合(Semantic

Mathematics and DIKWP : Creating New Qualities of Productivity). DOI:
10.13140/RG.2.2.19639.50085.
https://www.researchgate.net/publication/377726532_Semantic_Mathematics_and_DIKWP_Creat
ing_New_Qualities_of_Productivity.

[7] 段玉聪(Yucong Duan). (2024). 语义法学与 DIKWP：以英美法系与大陆法系分析为例

(Semantic Jurisprudence and DIKWP: Common Law vs. Continental Law). DOI:
10.13140/RG.2.2.28028.10889.
https://www.researchgate.net/publication/377726622_Semantic_Jurisprudence_and_DIKWP_Co
mmon_Law_vs_Continental_Law.

[8] 段玉聪(Yucong Duan). (2024). DIKWP 新质生产力与传统生产力的对比分析(DIKWP

New Quality Productivity vs. Traditional Productivity Analysis). DOI:
10.13140/RG.2.2.21317.22242.
https://www.researchgate.net/publication/377726626_DIKWP_New_Quality_Productivity_vs_Tra
ditional_Productivity_Analysis.

[9] 段 玉 聪 (Yucong Duan). (2024). 语 义 物 理 化 学 (Semantic Physical Chemistry). DOI:

10.13140/RG.2.2.21261.51684.
https://www.researchgate.net/publication/377439785_Semantic_Physical_Chemistry.
[10] 段玉聪(Yucong Duan). (2024). DIKWP 与语义认知学(DIKWP and Semantic Cognition).
DOI:10.13140/RG.2.2.14052.55680.
https://www.researchgate.net/publication/377415901_DIKWP_and_Semantic_Cognition.

[11] 段玉聪(Yucong Duan). (2024). DIKWP 与语义生物学：拓展跨学科的知识领域(DIKWP

and Semantic Biology: Expanding Interdisciplinary Knowledge Areas). DOI:
10.13140/RG.2.2.27474.32962.
https://www.researchgate.net/publication/377416091_DIKWP_and_Semantic_Biology_Expandin
g_Interdisciplinary_Knowledge_Areas

[12] 段玉聪(Yucong Duan). (2024). DIKWP 体系与语义数学结合构建传染病防治指标体系

(DIKWP System Combined with Semantic Mathematics to Construct an Indicator System for
Infectious Disease Prevention and Control). DOI: 10.13140/RG.2.2.12374.83521.
https://www.researchgate.net/publication/377416103_DIKWP_System_Combined_with_Semantic
_Mathematics_to_Construct_an_Indicator_System_for_Infectious_Disease_Prevention_and_Cont
rol

[13] 段玉聪(Yucong Duan). (2024). DIKWP 与语义哲学(DIKWP and Semantic Philosophy).

DOI:10.13140/RG.2.2.34185.21606.
https://www.researchgate.net/publication/377416120_DIKWP_and_Semantic_Philosophy

[14] 段玉聪(Yucong Duan). (2024). 语义物理与创新发展(Semantic Physics and Innovation

Development).DOI:10.13140/RG.2.2.19085.72167.
https://www.researchgate.net/publication/377416222_Semantic_Physics_and_Innovation_Develo
pment

[15] 段 玉 聪 (Yucong Duan). (2024). 语 义 认 知 学 ： 连 接 人 类 思 维 与 计 算 机 智 能 的 未 来

(Semantic Cognition: Connecting the Human Mind to the Future of Computer Intelligence). DOI:
10.13140/RG.2.2.29152.05129.
https://www.researchgate.net/publication/377416321_Semantic_Cognition_Connecting_the_Hum
an_Mind_to_the_Future_of_Computer_Intelligence

[16] 段玉聪(Yucong Duan). (2024). 语义物理：理论与应用(Semantic Physics: Theory and

Applications).DOI:10.13140/RG.2.2.11653.93927.
https://www.researchgate.net/publication/377401736_Semantic_Physics_Theory_and_Application
s

[17] 段玉聪(Yucong Duan). (2024). 基于语义数学的美国和中国经济增长分析(Semantic

Mathematics based Analysis of Economic Growth in the United States and China). DOI:
10.13140/RG.2.2.35980.90246.
https://www.researchgate.net/publication/377401731_Semantic_Mathematics_based_Analysis_of
_Economic_Growth_in_the_United_States_and_China

[18] 段玉聪(Yucong Duan). (2024). Collatz Conjecture 的语义数学探索(Collatz Conjecture's

Semantic Mathematics Exploration). DOI: 10.13140/RG.2.2.28517.99041.
https://www.researchgate.net/publication/377239567_Collatz_Conjecture's_Semantic_Mathematic
s_Exploration

[19] 段玉聪(Yucong Duan). (2024). 语义数学与 DIKWP 模型（本质计算与推理、存在计算

与推理以及意图计算与推理）(Semantic Mathematics and DIKWP Model (Essence Computation
and Reasoning, Existence Computation and Reasoning, and Purpose Computation and Reasoning)).
DOI:10.13140/RG.2.2.24323.68648.
377239628_Semantic_Mathematics_and_DIKWP_Model_Essence_Computation_and_Reasoning
_Existence_Computation_and_Reasoning_and_Purpose_Computation_and_Reasoning

[20] 段玉聪(Yucong Duan). (2024). 从主观到客观的语义数学重构（存在计算与推理、本质

计算与推理、意图计算与推理）(Semantic Mathematics Reconstruction from Subjectivity to
Objectivity (Existence Computation and Reasoning, Essence Computing and Reasoning, Purpose
Computing and Reasoning)). DOI: 10.13140/RG.2.2.32469.81120.
https://www.researchgate.net/publication/377158883_Semantic_Mathematics_Reconstruction_fro
m_Subjectivity_to_Objectivity_Existence_Computation_and_Reasoning_Essence_Computing_an
d_Reasoning_Purpose_Computing_and_Reasoning

[21] 段玉聪(Yucong Duan). (2024). DIKWP 与语义数学在车票订购案例中的应用(DIKWP and

Semantic Mathematics in the Case of Ticket Ordering). DOI: 10.13140/RG.2.2.35422.20800.
https://www.researchgate.net/publication/377085570_DIKWP_and_Semantic_Mathematics_in_th
e_Case_of_Ticket_Ordering

[22] 段玉聪(Yucong Duan). (2024). DIKWP 与语义数学分析《论语》 “君子和而不同，小人同

而不和”(DIKWP and Semantic Mathematical Analysis The Confluent Analects Gentleman is
harmonious but different, while petty people are the same but not harmonious). DOI:
10.13140/RG.2.2.28711.32165.
https://www.researchgate.net/publication/377085455_DIKWP_and_Semantic_Mathematical_Anal
ysis_The_Confluent_Analects_Gentleman_is_harmonious_but_different_while_petty_people_are
_the_same_but_not_harmonious

[23] 段玉聪(Yucong Duan). (2023). DIKWP 人工意识芯片的设计与应用(DIKWP Artificial

Consciousness Chip Design and Application). DOI: 10.13140/RG.2.2.14306.50881.
https://www.researchgate.net/publication/376982029_DIKWP_Artificial_Consciousness_Chip_D
esign_and_Application

[24] 段 玉聪 (Yucong Duan). (2024). 直 觉的本 质与 意识理 论的 交互 关系 (The Essence of

Intuition and Its Interaction with theory of Consciousness). DOI: 10.13140/RG.2.2.16556.85127.
https://www.researchgate.net/publication/378315211_The_Essence_of_Intuition_and_Its_Interacti
on_with_theory_of_Consciousness

[25] 段玉聪(Yucong Duan). (2024). 意识中的“BUG”：探索抽象语义的本质(Understanding

the Essence of "BUG" in Consciousness: A Journey into the Abstraction of Semantic Wholeness).
DOI:10.13140/RG.2.2.29978.62409.
https://www.researchgate.net/publication/378315372_Understanding_the_Essence_of_BUG_in_C
onsciousness_A_Journey_into_the_Abstraction_of_Semantic_Wholeness

[26] 段玉聪(Yucong Duan). (2024). 个人和集体的人造意识(Individual and Collective Artificial

Consciousness).DOI:10.13140/RG.2.2.20274.38082.
https://www.researchgate.net/publication/378302882_Individual_and_Collective_Artificial_Consc
iousness

[27] 段玉聪(Yucong Duan). (2024). 人工意识系统的存在性探究：从个体到群体层面的视角

(The Existence of Artificial Consciousness Systems: A Perspective from Group Consciousness).
 DOI:10.13140/RG.2.2.28662.98889.
https://www.researchgate.net/publication/378302893_The_Existence_of_Artificial_Consciousness
_Systems_A_Perspective_from_Collective_Consciousness

[28] 段玉聪(Yucong Duan). (2024). 意识与潜意识：处理能力的有限性与 BUG 的错觉

(Consciousness and Subconsciousness: from Limitation of Processing to the Illusion of BUG). DOI:
10.13140/RG.2.2.13563.49447.
https://www.researchgate.net/publication/378303461_Consciousness_and_Subconsciousness_fro
m_Limitation_of_Processing_to_the_Illusion_of_BUG

[29] 段玉聪(Yucong Duan). (2024). 如果人是一个文字接龙机器，意识不过是 BUG(If Human

is a Word Solitaire Machine, Consciousness is Just a Bug). DOI: 10.13140/RG.2.2.13563.49447.
https://www.researchgate.net/publication/378303461_Consciousness_and_Subconsciousness_fro
m_Limitation_of_Processing_to_the_Illusion_of_BUG

[30] 段玉聪(Yucong Duan). (2024). 超越达尔文：技术、社会与意识进化中的新适应性

(Beyond Darwin: New Adaptations in the Evolution of Technology, Society, and Consciousness).
DOI:10.13140/RG.2.2.29265.92001.
https://www.researchgate.net/publication/378290072_Beyond_Darwin_New_Adaptations_in_the_
Evolution_of_Technology_Society_and_Consciousness

[31] Wang A, Singh A, Michael J, et al. GLUE: A multi-task benchmark and analysis
platform for natural language understanding[J]. arXiv preprint arXiv:1804.07461, 2018.
[32] Wang A, Pruksachatkun Y, Nangia N, et al. Superglue: A stickier benchmark for
general-purpose language understanding systems[J]. Advances in neural information
processing systems, 2019, 32.
[33] Zellers R, Holtzman A, Bisk Y, et al. Hellaswag: Can a machine really finish your
sentence?[J]. arXiv preprint arXiv:1905.07830, 2019.
[34] Lin S, Hilton J, Evans O. Truthfulqa: Measuring how models mimic human
falsehoods[J]. arXiv preprint arXiv:2109.07958, 2021.
[35] Hendrycks D, Burns C, Basart S, et al. Measuring massive multitask language
understanding[J]. arXiv preprint arXiv:2009.03300, 2020.
[36] 国家互联网信息办公室,国家发展和改革委员会，教育 17 部,科学技术部,工
业和信息化部,公安部,国家广播电视总局.生成式人工智能服务管理暂行办
法.2023.
[37] 认知智能全国重点实验室,中国科学院人工智能产学研创新联盟,长三角人工
智能产业链联盟,通用认知智能大模型评测体系.2023.

View publication stats