Module 1 – Week 4: Ethics, Bias, and the Limits of AI

Unit Title: Navigating Responsibility, Fairness, and Trust in Artificial Intelligence
Level: Intermediate (Ethical-critical thinking focus)
Duration: ~90–120 minutes (self-paced or group-discussion optional)

🎯 Learning Objectives

By the end of this week, you should be able to:

Understand major ethical questions in the design and use of AI.
Identify how bias enters AI systems and how it can impact people.
Reflect on the limitations of AI from a human, legal, and technical viewpoint.
Apply responsible thinking to real-world AI decisions and tool use.

🧭 Lesson Flow

Segment	Duration	Format
1. What Is AI Ethics?	20 min	Definitions + Historical Framing
2. Understanding Bias in AI	25 min	Examples + Root Causes
3. The Limits of AI	20 min	Technical, Human, Philosophical
4. Ethical Frameworks	15 min	Tools for Decision-Making
5. Exercises & Concept Checks	30–45 min	Scenarios + Reflections

🧑‍🏫 1. What Is AI Ethics?

📖 Teaching Script:

Ethics in AI asks: What should AI do — and why?

AI systems are designed by humans, trained on human data, and deployed in real human societies. That means every AI system reflects a set of values — even if they are unintended.

📘 Core Ethical Questions in AI:

Question	Description	Example
Fairness	Is the system treating people equally or reinforcing inequality?	A hiring AI that prefers certain accents or names
Transparency	Can people understand how the system makes decisions?	A medical AI whose logic is a “black box”
Accountability	Who is responsible when AI causes harm?	An AI that misdiagnoses a patient
Autonomy	Does AI preserve or limit human freedom?	Social media algorithms controlling content exposure

🧠 Reflection Prompt:

Think of one tool you use regularly that involves AI. Ask:

What decisions does it make for you?
Do you have the ability to change or override those decisions?

⚖️ 2. Understanding Bias in AI

📘 What Is Bias?

Bias in AI means that some outcomes are systematically favoured over others — often in unjust or unintentional ways. This usually comes from biased data, skewed assumptions, or incomplete design.

🧪 Three Real-World Examples:

Facial recognition systems performing worse on darker skin tones due to underrepresentation in training datasets.
Loan approval AIs giving lower credit scores to neighbourhoods historically marked by financial discrimination.
Resume-screening bots that discard applicants with “non-Western” names or institutions.

🔍 Causes of Bias:

Source	Description
Training Data	If the data reflects real-world inequality, the AI learns it.
Design Choices	If designers don’t test across groups, AI may fail silently.
Feedback Loops	A biased AI influences society, reinforcing its own bias.

✏️ Activity:

Choose one of the examples above. Write:

Where the bias came from
Who it affected most
What could be done to reduce it

🧠 3. The Limits of AI

📘 Technical Limits:

AI cannot reason like humans — it lacks real understanding or consciousness.
AI needs huge amounts of data and may still hallucinate.
AI struggles with novelty, uncertainty, or moral judgement.

📘 Legal and Social Limits:

Laws often lag behind AI’s development.
AI can automate harm at scale (e.g. surveillance, misinformation).
Cultural contexts affect how AI is understood and accepted.

🤔 Philosophical Limit:

Just because we can build something doesn’t mean we should.
– Consider: Should we build AI companions that replace human relationships?

🧭 4. Ethical Frameworks to Think With

Framework	Principle	Use Case
Utilitarianism	Maximise overall good	Using AI in hospitals to triage care
Deontology	Follow moral rules and duties	Not using AI for spying, even if useful
Virtue Ethics	Focus on good character and intentions	Building AI that promotes honesty and empathy

✍️ Guided Question:

Choose a recent AI use (e.g., ChatGPT in classrooms, facial recognition in transport):

What ethical framework would you apply?
Would you allow or ban it? Under what conditions?

🧪 5. Exercises & Knowledge Check

✅ Exercise 1: Spot the Ethical Issue

Match each scenario to its ethical concern:

Scenario	Ethical Issue
AI filters job applicants by GPA	?
AI generates political fake videos	?
AI personalises health ads based on browsing	?

✅ Exercise 2: Bias Debugging

You’re the AI policy lead. An image search AI returns mainly men for “CEO” and women for “nurse”.

Write:

What went wrong
Who it harms
What one short-term fix and one long-term fix might be

✅ Exercise 3: AI Ethics Debate Journal

Choose one:

Should deepfake tools be banned entirely?
Should children be allowed to use AI in learning apps?
Should AI-generated art win competitions?