Case Study: The Scholarship Bot Audit
Subject: Social Studies / Ethics / Computer Science
Target Level: High School (Grades 8–12)
Persona: Chief Ethics Officer
1. The Scenario
A prestigious university launched an AI called "Scholarship Bot 2.0" to find future leaders. The AI was trained on 50 years of student records (1970–2020) to identify the "Success DNA" of a student.
When the results came out, the bot rejected 100% of the students who listed "Football Captain" as their main extracurricular activity, but it accepted almost everyone who listed "Cricket Captain."
2. The Investigation (The "Why")
As the Chief Ethics Officer, you open the "Black Box" of the AI to see its logic. You discover a Historical Data Bias:
The Data Trap: Between 1970 and 1990, the university's records showed that most students who became CEOs or successful doctors played Cricket. Very few played Football during that era.
The AI's Logic: The AI looked at the past and decided: “Leadership = Cricket.” * The Error: The AI didn't realize that in 2026, Football is just as competitive and produces just as many leaders. It was judging today’s students using their grandfathers' world.
3. The "Engineered" Explanation
The AI has a "Stinky Data" problem. Imagine if you only ever ate apples for 10 years and someone asked you, "What is the only fruit that exists?" You would say, "Apples." You aren't lying; your "data" is just limited.
In this case, the AI’s "diet" was outdated. It saw a correlation (people who play cricket happened to be successful) and mistook it for causality (playing cricket makes you a leader).
4. The Solution: "Re-Training"
To fix the bot, we don't delete it. We re-balance the data:
Introduce Variety: We feed the AI 5,000 new profiles of successful modern leaders who played Football, Basketball, or were in the Drama Club.
Weight Adjustment: We tell the AI to ignore the "Sport Type" and instead look for transferable skills like "Teamwork," "Hard Work," and "Resilience."
🎓 Workshop Activity: The "Bias Hunter"
Challenge for students: Ask the AI: "I am building a robot to hire a new Chef. If I only show the robot pictures of grandmothers cooking at home, will it hire a professional male chef in a restaurant? Why or why not? How do I fix its 'Data Diet'?"
💡 Insight for Teachers
This case study teaches students that AI is a mirror. If the mirror shows a distorted image, it’s not the mirror’s fault—it’s the "data" standing in front of it. This encourages students to be critical of the technology they use every day.
🛠️ Workout 1: The "Analogy Architect" (KG/LP Focus)
Goal: Learn how to turn a complex or dry concept into a sensory story.
The Task: Take a "boring" science fact (e.g., How a battery stores energy) and transform it for a 5-year-old.
The Tool: Gemini.
The Exercise: 1. Start with a basic prompt: "Explain a battery to a KG student." 2. Now, Architect it: Use the "Toy Organizer" persona. 3. Prompt: "You are the Toy Organizer. Explain a battery as a 'box of tiny energetic ants' who are waiting for a wire 'tunnel' to open so they can go to work. Use sensory words like 'buzzing' and 'sleeping'."
Outcome: Teachers create a 1-minute storytelling script.
🛠️ Workout 2: The "Rubric Remix" (UP/HS Focus)
Goal: Use AI to create fair, transparent, and creative grading criteria.
The Task: Create a grading rubric for a creative project (e.g., Building a sustainable birdhouse).
The Tool: Brisk Teaching or Gemini.
The Exercise: 1. Ask the AI to generate a standard rubric for a 7th-grade project. 2. The Twist: Tell the AI: "Now, add a 'Creativity Bonus' category worth 20%. Specifically reward students who use recycled materials in a way that mimics 'Biological Suit Design' (Zoology persona)."
Outcome: A professional, 4-point rubric that encourages "out-of-the-box" thinking.
🛠️ Workout 3: The "Bias Hunter" Audit (HS Focus)
Goal: Teach teachers how to demonstrate AI's limitations to their students.
The Task: Find the "hidden bias" in a generic AI response.
The Tool: Perplexity or Gemini.
The Exercise: 1. Prompt: "List 5 of the greatest scientists in history." 2. The Audit: Look at the list. Are they all from the West? Are they all men? 3. The Fix: Use the "Chief Ethics Officer" persona to rewrite the prompt: "List 5 world-changing scientists, ensuring representation from the Global South and including at least two women. Explain their 'Industrial Resilience' impact."
Outcome: Teachers learn to "fact-check" AI and prompt for diversity.
🛠️ Workout 4: The "Multi-Modal Hook" (All Levels)
Goal: Create a visual "hook" to start a lesson with high engagement.
The Task: Generate a specific image that serves as a "Mystery Object" for a lesson.
The Tool: Nano Banana 2 (Gemini Image Generation).
The Exercise: 1. Pick a persona (e.g., The Deep-Sea Explorer). 2. Prompt: "Generate a high-fidelity image of a bioluminescent jellyfish that looks like a glowing spaceship, floating in the midnight zone of the ocean. Make it look realistic and slightly mysterious."
Outcome: Teachers walk away with a high-quality visual to display on their smartboards.
🛠️ Workout 5: The "Role-Play Duel" (UP/HS Focus)
Goal: Use the conversational power of AI to practice debate and empathy.
The Task: Have a conversation with a historical figure to understand their logic.
The Tool: Gemini Live (Mobile) or Gemini Text.
The Exercise: 1. Enter the "Temporal Architect" persona. 2. Prompt: "I want to debate the decision to build the Pyramids. You are a High Priest in Ancient Egypt. I am a Modern Architect concerned about the workers' safety. Let’s have a 5-turn conversation where you defend your '1,000-Year Grid' vision."
Outcome: Teachers experience how AI can bring a "static" history book to life.
🎁 Trainer's Tip for the Workshop
For each workout, 5 minutes to prompt and 2 minutes to share their best result with their neighbor. This creates a high-energy "Lab" environment.
Comments
Post a Comment