Skip to content

AI Model Consensus: Champions of Fairness

View Fullscreen

About This MicroSim

This interactive Venn diagram compares which historical figures four different AI models identified as people who worked to make the world more fair.

The Experiment

We asked four AI models the same question: "List famous people who tried to make the world more fair for everyone on Earth."

The models queried were: - Claude (Anthropic) - OpenAI GPT (ChatGPT 5.2) - Grok (xAI) - DeepSeek

Key Findings

Universal Consensus (All 4 models): Nelson Mandela, Mahatma Gandhi, Martin Luther King Jr.

Strong Consensus (3 of 4): Malala Yousafzai, Eleanor Roosevelt, Harriet Tubman

Unique Selections: Each model also included figures not mentioned by others, revealing different perspectives on fairness advocacy.

Hover over any dot to learn more about that person and which models mentioned them. Larger dots indicate more models agreed on that person.

Lesson Plan for High School Students

Overview

This lesson uses AI-generated lists of historical fairness champions to explore how different perspectives shape our understanding of justice, equity, and social change. Students will analyze patterns in AI responses, evaluate criteria for measuring contributions to fairness, and reflect on whose voices are included or excluded from historical narratives.

Grade Level: 9-12

Subject Areas: Social Studies, Ethics, Media Literacy, Computer Science

Duration: 50-60 minutes (expandable to 90 minutes with extensions)

Learning Objectives

By the end of this lesson, students will be able to:

  1. Identify historical figures who advocated for fairness and describe their contributions (Bloom: Remember/Understand)
  2. Analyze patterns of consensus and divergence across AI model responses (Bloom: Analyze)
  3. Evaluate the criteria used to define "making the world more fair" (Bloom: Evaluate)
  4. Assess how AI training data and design choices influence outputs (Bloom: Evaluate)
  5. Construct arguments for including additional figures in fairness discussions (Bloom: Create)

Materials Needed

  • Access to the interactive Venn diagram MicroSim (computer/tablet per student or group)
  • Student worksheet (provided below)
  • Whiteboard or shared document for class discussion
  • Optional: Access to research resources for extension activity

Vocabulary

Term Definition
Consensus General agreement among a group
Divergence Difference or deviation from a common point
Training data The information used to teach an AI system
Bias A tendency to favor certain perspectives over others
Advocacy Public support for a cause or policy
Social justice Fair distribution of opportunities and privileges in society

Lesson Structure

Opening Hook (5 minutes)

Think-Pair-Share: Ask students: "If you had to name three people in history who made the world more fair, who would you choose and why?"

  • Students think individually (1 minute)
  • Share with a partner (2 minutes)
  • Volunteer pairs share with the class (2 minutes)

Record responses on the board. Tell students they'll compare their choices with what four different AI systems selected.

Direct Instruction (10 minutes)

Introduce the Experiment:

  1. Explain that four major AI models were asked the same question students just answered
  2. Introduce the four AI models briefly:
  3. Claude (Anthropic) - designed with focus on being helpful and harmless
  4. ChatGPT (OpenAI) - widely used general-purpose AI
  5. Grok (xAI) - designed to be more direct and less filtered
  6. DeepSeek - developed in China with different training priorities
  7. Explain that each AI generated a list of 10 people, and the Venn diagram shows where they agreed and disagreed

Key Concepts to Introduce:

  • AI models learn from text written by humans
  • Different training data leads to different outputs
  • Consensus doesn't mean "correct"—it reveals shared cultural narratives
  • Absence from lists doesn't mean someone was unimportant

Guided Exploration (15 minutes)

Interactive MicroSim Activity:

Have students explore the Venn diagram individually or in pairs:

  1. Find the Center (3 minutes): Identify the three figures all four models agreed on. Hover to read about each one.

  2. Question: What do Mandela, Gandhi, and MLK have in common?

  3. Explore the Overlaps (5 minutes): Find figures mentioned by 2-3 models.

  4. Question: Why might some models include Harriet Tubman while others didn't?

  5. Examine the Edges (5 minutes): Find figures unique to only one model.

  6. Question: What does a unique selection tell us about that AI's perspective?

  7. Notice Patterns (2 minutes): Look at the overall distribution.

  8. Question: Are there patterns in geography, gender, time period, or type of activism?

Small Group Analysis (12 minutes)

Divide students into groups of 3-4. Assign each group one of these analysis tasks:

Group A - Geographic Analysis: Where are these figures from? Create a tally of continents/regions represented. What regions seem underrepresented? Why might that be?

Group B - Gender Analysis: Count the gender breakdown across all lists. Is there balance? What might explain any imbalance you find?

Group C - Time Period Analysis: When did these figures live? Are recent advocates included? Historical figures? What time periods dominate?

Group D - Type of Change Analysis: What methods did these figures use? (legislation, protest, education, armed resistance, writing, etc.) Which methods appear most often?

Groups report findings to the class (2 minutes each).

Class Discussion (10 minutes)

Facilitate discussion using these questions:

  1. Consensus Questions:
  2. Why do you think Mandela, Gandhi, and MLK appeared on all four lists?
  3. Does universal AI agreement mean these are objectively the "best" advocates for fairness?

  4. What narrative about social change do these three figures represent?

  5. Divergence Questions:

  6. What might explain why DeepSeek included different figures than Claude?
  7. If an AI was trained primarily on texts from South America, how might its list differ?

  8. Why might some models include controversial figures that others exclude?

  9. Critical Thinking Questions:

  10. Who is missing from ALL the lists that you think should be included?
  11. How does the phrasing of the question ("made the world more fair") influence the answers?
  12. Should AI systems be expected to agree on questions about values and history?

Closing Reflection (5 minutes)

Exit Ticket: Students respond in writing to ONE of these prompts:

  1. Name one person you would add to these lists and explain why in 2-3 sentences.
  2. What is one thing this activity taught you about how AI systems "think"?
  3. How might your own background influence which fairness advocates you consider most important?

Assessment Rubric

Criteria Emerging (1) Developing (2) Proficient (3) Advanced (4)
Pattern Recognition Identifies obvious patterns with prompting Identifies some patterns independently Identifies multiple patterns with evidence Identifies subtle patterns and explains significance
Critical Analysis Accepts AI output without questioning Questions AI output with basic reasoning Analyzes factors influencing AI output Synthesizes multiple factors affecting AI bias
Historical Knowledge Limited knowledge of figures discussed Demonstrates basic knowledge of key figures Connects figures to broader movements Explains historical context and lasting impact
Argumentation States opinion without support Provides limited evidence for claims Constructs clear arguments with evidence Builds nuanced arguments considering counterpoints

Differentiation Strategies

For Students Who Need Support:

  • Provide a pre-filled graphic organizer with figure names
  • Pair with a stronger reader for exploration
  • Focus on the three consensus figures rather than the full diagram
  • Provide sentence starters for discussion

For Advanced Students:

  • Research the training data sources for one AI model
  • Compare this experiment with a similar prompt about "people who made the world less fair"
  • Investigate how the same question might be answered in different languages
  • Design a follow-up experiment with a modified prompt

Extension Activities

Extension 1: Research Project (Homework)

Choose one figure from the diagram you hadn't heard of before. Research their life and contributions, then write a one-page profile answering:

  • What specific actions did they take to promote fairness?
  • Who opposed them and why?
  • What is their legacy today?

Extension 2: Design Your Own Prompt

If you could ask AI models a different question about fairness, what would it be? Write your question and predict how different models might respond differently. Consider:

  • "Who is fighting for fairness TODAY?"
  • "What ORGANIZATIONS have done the most for fairness?"
  • "What LAWS have made the world more fair?"

Extension 3: The Missing Voices

Research one person or group that doesn't appear on any AI list but should be recognized for advancing fairness. Create a presentation explaining:

  • Who they are and what they did
  • Why they might be underrepresented in AI training data
  • How their story changes our understanding of fairness movements

Extension 4: AI Bias Investigation

Compare outputs from multiple AI chatbots on the same fairness question. Document:

  • Exact prompts used
  • Complete responses from each AI
  • Analysis of differences
  • Hypotheses about why differences exist

Standards Alignment

Common Core State Standards:

  • CCSS.ELA-LITERACY.RH.9-10.7: Integrate quantitative or technical analysis with qualitative analysis
  • CCSS.ELA-LITERACY.RH.11-12.7: Integrate and evaluate multiple sources of information
  • CCSS.ELA-LITERACY.SL.9-12.1: Initiate and participate in collaborative discussions

C3 Framework for Social Studies:

  • D2.His.1.9-12: Evaluate how historical events and developments were shaped by unique circumstances
  • D2.His.5.9-12: Analyze how historical contexts shaped perspectives
  • D4.1.9-12: Construct arguments using claims and evidence from multiple sources

ISTE Standards for Students:

  • 1.3 Knowledge Constructor: Students critically curate digital resources
  • 1.6 Creative Communicator: Students communicate complex ideas clearly

Teacher Notes

Preparation Tips:

  • Test the MicroSim on classroom devices before the lesson
  • Review the historical figures so you can provide context as needed
  • Prepare for potentially sensitive discussions about controversial figures
  • Have backup discussion questions ready if conversation stalls

Potential Sensitive Topics:

  • Some figures (like Gandhi) have complicated legacies that students may raise
  • Discussion of "unfair" figures in the companion lesson involves dictators and genocide
  • Students from different backgrounds may have strong feelings about regional representation

Common Misconceptions to Address:

  • "AI is objective" — AI reflects the biases in its training data
  • "Consensus means correct" — Agreement among AI models reflects shared cultural narratives, not truth
  • "Missing means unimportant" — Absence often reflects whose stories get written down and digitized

Discussion Questions

  1. Why do you think Mandela, Gandhi, and MLK appeared in all four lists?
  2. What patterns do you notice in the unique selections each AI made?
  3. How might the training data of each AI model influence its choices?
  4. Are there important figures missing from all lists? Who would you add?

Source Files