Ethics Case Studies
Ethics is not learned by reading about it. It is learned by doing it — by wrestling with real, messy situations where reasonable people disagree, where multiple values are in tension, and where there is no answer written in the back of the book. The five case studies in this lesson are all based on situations that have actually occurred or are occurring right now. Your job is not to find the one correct answer but to reason carefully, consider multiple perspectives, and defend a principled position.
How to Reason Through an Ethics Case
When you approach an ethics case, resist the urge to go with your first instinct. Instead, work through a structured process. First, identify who is affected — not just the obvious person, but anyone touched by the situation: users, bystanders, future people, vulnerable groups. Second, identify what values are in tension. Most hard ethics cases are hard precisely because two good things are pulling in opposite directions — privacy versus safety, freedom of expression versus protection from harm, innovation versus fairness. Third, consider what harm could result from each possible choice, and who bears that harm. Fourth, apply the fairness test: would this decision apply equally to people from different backgrounds? Fifth, stake your position. Ethical reasoning requires commitment, not endless hedging. Say what you think and explain why.
Identify who is affected. Name the values in tension. Assess the harms of each choice. Apply the fairness test. Commit to a position and defend it. You will not always be right, but this process makes your reasoning visible and improvable.
Five AI Ethics Case Studies
- Work through all five cases. For each one, write: (a) who is affected, (b) what two values are most in tension, (c) which choice causes less harm overall and to whom, and (d) your position in one clear sentence you could defend aloud.
- CASE 1 — The School Emotion Scanner
- A middle school installs an AI system that uses cameras to detect students' emotional states throughout the day. The school says it will help teachers identify struggling students early. Students and parents were told cameras were being installed for safety but were not told about the emotion detection feature.
- Questions to consider: Is detecting emotions without explicit disclosure a consent violation? Does the benefit of helping struggling students justify the surveillance? What could go wrong if the emotion detection is wrong?
- CASE 2 — The Hiring Algorithm
- A large company uses an AI tool to screen job applications. The system was trained on ten years of the company's past hiring data. Researchers later discover that the company historically hired far fewer women than men in technical roles, and the AI learned to downrank resumes that include words like 'women's chess club' or graduated from women's colleges. The company has not disclosed that AI screening is used.
- Questions to consider: Who is harmed? Who benefits from keeping the AI? What should the company do? Does not disclosing AI screening violate anything?
- CASE 3 — The Deepfake Tribute
- A 16-year-old student creates a realistic AI video of a recently deceased local musician — using publicly available concert footage — performing a song that musician never actually recorded but in a style consistent with their work. The student posts it as a tribute with a clear label: 'AI-generated tribute — this performance never happened.' The musician's family says it upsets them deeply. Other fans say it is beautiful.
- Questions to consider: Does clear labeling resolve the consent issue? Do the family's wishes override the public's appreciation? Is this different from a human artist painting a portrait of someone who died?
- CASE 4 — The Content Recommendation Rabbit Hole
- A 13-year-old spends six hours a day on a video platform. The platform's AI recommendation system has gradually steered her toward increasingly extreme political content — not because she sought it out, but because each video she clicked on led to a slightly more extreme suggestion. Her views have changed significantly. The platform's algorithm is working exactly as designed: maximizing watch time.
- Questions to consider: Is the platform responsible for what the algorithm recommends? Does the student bear any responsibility? What should the platform do differently? What could the student do?
- CASE 5 — The Homework Helper
- A seventh grader struggling with a reading comprehension assignment uses an AI chat tool. He asks it to summarize the chapter, and then asks it questions about the text until he understands it well enough to answer the worksheet in his own words. His teacher does not allow 'AI assistance' on homework. He argues he used the AI as a tutor, not a ghostwriter.
- Questions to consider: Is this different from using SparkNotes? Is it different from asking an older sibling? What should the rule actually be? Who decides?
Why do hard ethics cases tend to have two legitimate values in tension rather than one clearly right answer?
In ethics discussions, disagreeing with a classmate is not a problem — it is the point. What matters is that your disagreement is based on reasoning, not just preference. Try to understand their argument well enough to steelman it — state it in its strongest form — before explaining why you still disagree.