Lesson Title: AI Biases & Data Ethics
Lesson Objectives
By the end of this lesson, learners will be able to:
-
Understand what AI bias is and where it originates.
-
Explain the concept of data ethics and its importance in the development and use of AI systems.
-
Analyze real-world examples of AI bias and unethical data practices.
-
Explore strategies for identifying and mitigating bias in AI systems.
-
Discuss their own responsibilities as digital citizens and AI users.
Section 1: Introduction to AI Bias
Definition:
AI bias refers to systematic and unfair discrimination in the outcomes of AI systems due to biased data, algorithms, or design processes.
Where Bias Comes From:
-
Biased Training Data: AI learns from historical or real-world data that may already reflect societal biases.
-
Labeling Bias: Data is labeled by humans, who can introduce subjective opinions or stereotypes.
-
Algorithmic Bias: Models can amplify or reproduce biases during learning.
-
Selection Bias: Data collected may not represent all user groups fairly.
Example:
Facial recognition systems misidentifying people of color at significantly higher rates due to underrepresentation in training datasets.
Class Discussion Prompt:
Have you ever seen or experienced technology behaving unfairly? What do you think caused that?
Section 2: Introduction to Data Ethics
Definition:
Data ethics is the branch of ethics that studies and evaluates moral problems related to data collection, sharing, analysis, and use—especially in automated systems.
Core Principles of Data Ethics:
-
Transparency: How decisions are made by AI should be understandable.
-
Accountability: Developers and companies must be held responsible for AI’s impacts.
-
Fairness: AI should not discriminate against individuals or groups.
-
Privacy: Individuals have the right to control how their data is collected and used.
-
Consent: Data must be collected with informed user consent.
-
Beneficence: AI should contribute to human well-being.
Section 3: Real-World Case Studies
Case Study 1: COMPAS – Predictive Policing Tool
-
Predicted recidivism rates and was found to be biased against Black defendants.
-
Problem: Biased historical data, lack of transparency in scoring algorithm.
Case Study 2: Amazon AI Recruiting Tool
-
The tool downgraded resumes from women applying to technical roles.
-
Problem: Trained on 10 years of biased hiring data where men were overrepresented.
Case Study 3: Google Photos Labeling Error (2015)
-
Labeled African Americans as “gorillas”.
-
Problem: Inadequate and racially biased image training datasets.
Case Study 4: TikTok Algorithm Suppression
-
Allegedly suppressed videos from users with disabilities, LGBTQ+ identities, or “non-standard” appearance.
-
Problem: Moderation policies reflecting societal bias and commercial motivations.
Section 4: Strategies to Mitigate AI Bias
Technical Solutions:
-
Data Auditing: Check datasets for imbalances and discrimination.
-
Bias Testing: Run fairness evaluations before deployment.
-
Algorithm Design: Use interpretable and explainable models.
-
Inclusive Data Collection: Diversify sources and use representative samples.
Policy & Process Solutions:
-
Human Oversight: Ensure AI does not make high-stakes decisions alone.
-
Ethics Committees: Diverse teams reviewing AI use cases and development.
-
Transparent Reporting: Open documentation of AI systems (e.g., Model Cards, Datasheets for Datasets).
Individual & Organizational Responsibility:
-
Developers: Must test and disclose risks.
-
Users: Should ask critical questions about AI tools they use.
-
Institutions: Should set clear ethical standards and policies.
Section 5: Activity – Ethics in Action
Option A: AI Bias Detective
-
Students are given a short news story describing a biased AI outcome.
-
In small groups, they identify:
-
What kind of bias occurred?
-
Where did the bias originate?
-
How could it be prevented?
-
Who should be accountable?
-
Option B: Ethics Debate
-
Motion: “AI should never be used in hiring decisions.”
-
Split into two teams and present arguments for or against, focusing on fairness, efficiency, and societal impacts.
Section 6: Learner Reflection & Discussion
Prompt Questions:
-
What surprised you about how bias can appear in AI?
-
Should companies be punished for deploying biased AI systems?
-
How can you be a more ethical digital citizen when using or developing technology?
Reflection Journal Entry:
Write a short paragraph about one ethical issue in AI you care about most and why.
Section 7: Resources for Ethical AI Learning
Free Courses & Toolkits:
Podcasts & Articles:
-
“In Machines We Trust” – MIT Tech Review Podcast
-
“AI and Ethics” – World Economic Forum Articles
-
“Weapons of Math Destruction” – Book by Cathy O’Neil
Tools for Practice:
-
Google’s What-If Tool for testing AI fairness
-
IBM’s AI Fairness 360 Toolkit
-
Hugging Face Datasets explorer
Assessment / Homework
Project:
-
Create a slide deck or poster that explains a real-world example of AI bias.
-
Include the origin of the bias, who was affected, and a suggested ethical solution.
Quiz Topics:
-
Definitions: AI bias, data ethics, fairness
-
Real-life AI failures and what caused them
-
Principles of ethical AI development
129
