Scribd
Upload
17 views6 pages

Ethics and Bias in Machine Learning

This literature review discusses the ethical implications and biases present in machine learning (ML) systems, particularly in critical fields like healthcare and criminal justice. It highlights the importance of recognizing different types of biases, implementing ethical guidelines, and developing fairness-aware algorithms to mitigate these issues. The paper emphasizes the need for transparent and accountable AI practices to ensure responsible deployment and address ethical challenges effectively.

Uploaded by

momenfahmid
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
17 views6 pages

Ethics and Bias in Machine Learning

This literature review discusses the ethical implications and biases present in machine learning (ML) systems, particularly in critical fields like healthcare and criminal justice. It highlights the importance of recognizing different types of biases, implementing ethical guidelines, and developing fairness-aware algorithms to mitigate these issues. The paper emphasizes the need for transparent and accountable AI practices to ensure responsible deployment and address ethical challenges effectively.

Uploaded by

momenfahmid
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 6

COM754

Research Methods for Digital Technologies

Ethics and Bias in Machine Learning: A Literature


Review

Student Name:
Student ID:
Word count: 1209
Introduction
Machine Learning systems have increasingly become a core component in industries
like hiring, criminal justice, healthcare and even finance. However they could also be
ethically wrong and biased results. All these elements affect the outcomes that can
adversely impact society based on biases and patterns encoded in algorithms or data.
Ethics and bias consider the implications of these outcomes. If datasets or training data
is not well represented then bias condition occurs which discriminate with certain set of
data. This leads to treating wrongly for example people of certain races or race-
serviceable positions and this has a real impact on the lives of peoples. We need a
transparent and ethically rich vein of AI to head this bureaucratic train wreck off at the
pass. To prevent immoral consequences, data diversification and fairness-aware
algorithms are customization is very important. The only way to reduce this issue is by
knowing the types of bias and it's unethical background, framework or methods. Also,
responsible deployment of machine learning systems is a must in order to avoid future
consequences of biased outcomes.

Types and Sources of Bias in ML


In cancer diagnosis and treatment, machine learning (ML) can produce biased
outcomes. The training data and the system of machine learning both are responsible
for biases. Data bias is very common as it occurs through the training data. When this
data lacks diverse patient information, the system creates diagnostic errors and
treatments based on bias. Another bias which can cause even more issues for the
consumers is the algorithmic bias. A machine learning (ML) model might accidentally
learn from previous biased data and patterns. This takes away the machine's ability to
fairly judge and produce outcomes in some scenarios. On top of these biases,
interpretive bias is seen in complex algorithms. It creates new challenges for the system
while identifying bias thus the system is unable to rectify itself in crucial decisions.
Therefore, it is crucial to create strong ethical guidelines so that the system can make
crucial and reliable decisions excluding bias. (3)

Ethical Implications of Bias


When it comes to gender, race and age, machine learning (ML) models can produce
unfair and biased outcomes. Model bias can strengthen disparities that are already
present in the datasets and affect the choice of the ML model in various fields such as
financial services, employment sectors and justice systems. The cited research worked
on the significance of bias detection in ML models by introducing a method which
changes the potentially biased attributes (PBAs). Training data or certain patterns that
are causing biased outcomes are recognized to make changes in the PBAs.
Researchers closely observed how these new changes and modifications affected the
system results. It is vital to create a ML model which is very transparent and open to
inspections in order to recognize any biases or unfair patterns and rectify them. (1). So,
responsible AI development is vital to steer clear of future ethical concerns regarding
ML models.

Detection and Measurement of Bias


Since ML models make decisions for critical components, it is essential that we have a
thorough way of detecting bias so as to build fairness into these systems. There are
three different types of bias mitigation techniques: pre-processing, in-processing and
post-processing. In-processing is the most frequently used one This method, on the
other hand involves altering your model in a training loop using datasets. In contrast,
post-processing perturbs the outcomes to return a fairer output. A pre-processing
method processes the input datasets prior to training the model in order to reduce any
bias. The reported survey examined both datasets and processes to explore the way in
which models are being assessed. These two concerns discussed specifically include
— how do you make sure the data is reliable and what are groups of data that need to
be checked for potential or existing biases. (4). One is required to find particular
mitigation technologies in order that equity of models is practiced.

Techniques for Mitigating Bias in ML


Despite this, a robust FATE (fairness, accountability, transparency and ethics) principles
grounded framework is necessary to ensure ethical and equitable deployment of
Machine Learning (ML) models. An in-depth examination of FATE for AI further
encourages the need to define with exactitude and devise techniques on how these
ethical dilemmas may be tackled, which is especially critical within complex fields such
as education. Fairness is important, and the way to achieve fairness most easily centers
around quantitative methods that measure how well ML models serve different
populations. Where transparency provides flexibility of interpretation and allows
stakeholders to better understand decision-making, accountability ensures that
developers who release models own up to the outcomes. And of course, this also
includes how AI impacts our most vulnerable populations — a matter which can and
should be approached via both qualitative and quantitative evaluations. That is the
conclusion of a new study that calls for more research into consistent, open-access
ways to formalize FATE principles in ML applications—which helps set the stage for
ethical AI practices. (2)

Current Challenges and Gaps


There are difficulties to implement ethical values throughout the lifecycle of artificial
intelligence (AI). This whole procedure can be credentialed to a series of ethical
standards, including accountability, justice,equity and transparency being the first two.
Nonetheless, since AI often raises different issues that laws struggle to address
satisfactorily in the here and now, so not having a common ethical frame could be a
major impediment. And, without many reliable tools — like an ethical review board or
even a checklist for ethics in AI workflows— it can be hard to actually put that idea of
practicing ethics into practice with tech. The distance between theoretical ethical
principles and applied practices underscores the need for accessible, actionable
strategies rooted in accountability and transparency. These ethical principles should be
given substance in future initiatives to ensure the responsible development and
deployment of AI. (5)

Conclusion
In this paper, we provide a comprehensive study of the ethical challenges and barriers
in ML systems including what FATE means for AI. While providing groundbreaking
assistances for key domains like healthcare, finance and criminal justice; ML systems
have ethical burdens to bear in the form of biases baked into training data or model
design. They stress the importance of systematic and reproducible identification,
quantification and correction methods suitable for bias. The results underscore how
critical it is to develop pre-, in- and post-processes for creating a basis that will allow
bias mitigation, as well as ethical AI deployment. Equally, as opening up new ethical
dilemmas égalités de pointe face à des intégrations simplifiées with theoretical ethics
not relied on so highly to reduce AI inheritants loops in providing moral codes that
seems paltry of some insights into how they would interact. There are not yet practical
tools to preserve the accountability and transparency of AI as it develops (e.g. a review
board or an ethical checklist). Future work should consider what good ethical oversight
practices are and how to foster the formation of honest, answerable machine learning
technologies.
References

1) Alelyani, S. (2021, July 7). Detection and Evaluation of Machine Learning


Bias. MDPI. Retrieved October 27, 2024, from https://www.mdpi.com/2076-
3417/11/14/6271
2) Fraser, S. (2023, June 26). Fairness, Accountability, Transparency, and Ethics
(FATE) in Artificial Intelligence (AI) and higher education: A systematic review.
https://www.sciencedirect.com/science/article/pii/S2666920X23000310
3) Hajjo, R. (2019, January 20). The Ethical Challenges of Applying Machine
Learning and Artificial Intelligence in Cancer Care.
https://ieeexplore.ieee.org/document/8618186/authors#authors
4) Hort, M., Chen, Z., Zhang, J. M., Harman, M., & Sarro, F. (2024, June 20).
Bias Mitigation for Machine Learning Classifiers: A Comprehensive Survey.
https://dl.acm.org/doi/full/10.1145/3631326#core-cited-by
5) Zhou, J., Chen, F., Berry, A., Reed, M., Zhang, S., & Savage, S. (2021,
January 5). A Survey on Ethical Principles of AI and Implementations.
https://ieeexplore.ieee.org/document/9308437/authors#authors

You might also like