AI and ML: What are the

main paradigms?
Educational Computing - MSc Smart EdTech
Chloé Mercier, Inria
 Artificial Intelligence?

Making a machine do typical cognitive functions.

Educational Computing - MSc Smart EdTech - UCA 2

 Artificial Intelligence?

Making a machine do typical cognitive functions.

“the science of making machines do things that would require intelligence if

done by human” (Marvin Minsky, 1968)

“AI system: An AI system is a machine-based system that, for explicit or

implicit objectives, infers, from the input it receives, how to generate outputs
such as predictions, content, recommendations, or decisions that can
influence physical or virtual environments. Different AI systems vary in their
levels of autonomy and adaptiveness after deployment. “

Recommendation of the Council on Artificial Intelligence (OECD, 2019)

Educational Computing - MSc Smart EdTech - UCA 3

 Artificial Intelligence?

Making a machine do typical cognitive functions.

recognizing predicting

“cat”

grouping

deciding

Educational Computing - MSc Smart EdTech - UCA 4

 How are the
ears shaped?
pointy floppy

How is the Are there

Symbolic face shaped? whiskers?
not round yes no
approaches
round

CAT DOG CAT DOG

Educational Computing - MSc Smart EdTech - UCA 5

 How are the
ears shaped?
pointy floppy

How is the Are there

Symbolic face shaped? whiskers?
not round yes no
approaches
round

CAT DOG CAT DOG

Statistical
approaches
(ie Machine Learning)

(animation from Buff ML)

Educational Computing - MSc Smart EdTech - UCA 6

 Symbolic vs Numerical approaches
Rules
(explicit)
Symbolic
approaches
Data Model Answer

Training phase:
Statistical
Data Model
approaches Training
algorithm
(ie Machine Learning)
Rules
After training: (implicit)

Data Model Answer

Educational Computing - MSc Smart EdTech - UCA 7
 Symbolic vs Numerical approaches

Symbolic Declarative
approaches knowledge
(“know that”)

)
lel
ral
(pa
Statistical Procedural
approaches knowledge
(ie Machine Learning) (“know how”)

Educational Computing - MSc Smart EdTech - UCA 8

 Symbolic vs Numerical approaches
Next/Last year

Symbolic Declarative
approaches knowledge
(“know that”)

Statistical Procedural
approaches knowledge
(ie Machine Learning) (“know how”)
Let’s focus on this today!

Educational Computing - MSc Smart EdTech - UCA 9

 Learning paradigms

Supervised Learning

Machine Unsupervised Learning

Learning
Reinforcement Learning

Educational Computing - MSc Smart EdTech - UCA 10

 Learning paradigms

Supervised Learning

Machine Unsupervised Learning

They can all be
“deep” (deep learning)
Learning
Reinforcement Learning

+ others (semi-supervised, generative…

all of these are not necessarily exclusive)

Educational Computing - MSc Smart EdTech - UCA 11

 Supervised Learning
Look: this is a dog!
And that is a cat!

Mmm, ok I see…that one is

probably a cat too then?

image from DatabaseTown

Educational Computing - MSc Smart EdTech - UCA 12
 Unsupervised Learning
Mmm these ones
look similar, I will
put them together

image from DatabaseTown

Educational Computing - MSc Smart EdTech - UCA 13
 Mmm delicious! I
will choose this
Reinforcement Learning
again next time

Mmm I lost another

piece, I shouldn’t
have done that… I
will try another
move next time

image from Data Basecamp

Educational Computing - MSc Smart EdTech - UCA 14
 Learning Paradigms: Recap?
Supervised ML Unsupervised ML Reinforcement ML

Type of data

Criteria to
optimize

Type of
problems

Type of
model (after
training)

Applications

Educational Computing - MSc Smart EdTech - UCA

Quiz time? 15
 Learning Paradigms: Recap
Supervised ML Unsupervised ML Reinforcement ML

Type of data Labelled data Unlabelled data The environment to interact

with + rewards

Criteria to Minimizing the error Maximizing the similarity along Maximizing the reward
optimize between given answer and some feature dimensions
label

Type of Regression, Classification Clustering, Dimensional Reduction Navigation, Planning

problems

Type of Mapping: Mapping: Mapping:

model (after Features → labelled class Features → unlabelled class State → action
(if classification)
training)
Or: Features → numerical
value (if regression)

Applications Image Recognition, Recommendation Systems, Self Driving Cars, Gaming…

Weather Forecast… Anomaly Detection…

Educational Computing - MSc Smart EdTech - UCA 16

 A brief history of AI
2011 onwards:
personal
1973: First assistants (Siri, 2016: AlphaGo
humanoid robot Alexa, etc) and beats Go
Wabot (from Japan) recommendation champion
1986: Multilayer
systems (Netflix)
perceptron + 1996: DeepBlue
1957: First “artificial
backpropagation beats Kasparov
neural network” 1970 1990 2010
(perceptron) at chess 2020
2022 onwards:
Generative AI boost
74- 87- 2014:
8 9
(it w 0: 1st (it i 9: 2n (ChatGPT,
ill n AI w s us d Generative
eve i eles AI win Midjourney, Dall-e…)
r w nter s) ter AI is born
1960 ork 1980 2000 2006:
1966: First )
Humanoid
“chatbot”
1956: Dartmouth robot NAO
ELIZA
conference. “Artificial First expert system
Intelligence” field is (XCON) 2002:
born! Roomba
avoids
obstacles

Mostly numeric (probabilistic/statistical learning, Neuro-

Mostly symbolic (logic, representation and reasoning)
deep learning) symbolic?

Educational Computing - MSc Smart EdTech - UCA To learn more: https://ourworldindata.org/brief-history-of-ai 17

 STUDENT-FOCUSED AIED TEACHER-FOCUSED AIED INSTITUTION-FOCUSED AIED

Intelligent Tutoring *** Automatic Formative *** Plagiarism detection *** Admissions (e.g., student ***
Systems (ITS) Assessment (AFA) /** selection)

AI-assisted Apps (e.g., *** Learning Network *** Smart Curation of Learning *** Course-planning, Scheduling, ***
maths, text-to-speech, Orchestrators /** Materials Timetabling
language learning)

AI-assisted *** Dialogue-based *** Classroom Monitoring *** School Security ***
Simulations (e.g., Tutoring Systems
games-based learning, (DBTS)
VR, AR)

AI to Support Learners *** Exploratory Learning ** Automatic Summative *** Identifying Dropouts and ***
with Disabilities Environments (ELE) Assessment /** Students at risk

Automatic Essay *** AI-assisted Lifelong * AI Teaching Assistant (incl. *** e-Proctoring ***
Writing (AEW) Learning Assistant assessment assistant) /*

Chatbots *** AI Teaching Assistant *** Classroom Orchestration **

/** (incl. assessment /*
assistant)
speculative (*), researched (* *), or commercially available (* * *)
Educational Computing - MSc Smart EdTech - UCA Holmes et al, 2022. State of the art and practice in AI in education 18
 basic architecture of an intelligent tutoring system
Statistical
computing on the
user’s learning
path: which
Performs exercises are the
exercises most suitable?
New
exercises see also ZPDES
Results suggestions
algorithm
(Mazon et al.,
2023)
Recommendation model
(machine learning based, Or another real-life
eg, neural network) example: Lifelong
Monitors
Learning Courses
Recommendation
System to Improve
Students Professional Skills
Using Ontology
and Machine
Curriculum model Learning
(symbolic, eg, knowledge Designs (Urdaneta-Ponte et
graph formalized by an
ontology) al, 2021)
Tutor

Educational Computing - MSc Smart EdTech - UCA 19

 Questions?

Educational Computing - MSc Smart EdTech - UCA 20

 Thank you!
Acknowledgements to:

● Thierry Viéville, Margarida Romero, Saint-Clair Lefèvre

● And the creators of this free Nathaniel Presentation template

Educational Computing - MSc Smart EdTech - UCA 21