Ethical Algorithm Lesson Plan The Ethical Algorithm
Lesson Facilitation Guide
UNIT: Ethics & AI
DURATION: 90-120 MIN
Learning Objectives
Identify how bias can be introduced into machine learning models through training data.
Analyze real-world case studies of algorithmic discrimination .
Develop a "Developer's Code of Ethics" for software development.
Propose technical and policy-based solutions to mitigate machine learning bias.
Essential Question
"If code is math, can it still be unfair?"
Preparation & Materials
Teacher Materials
• Code Ethics Slide Deck
• Case Study: The "Auto-Hire 3000" Files
• Ethics Auditor Rubric
Student Materials
• Bias Detective Activity Sheet
• Ethics Journal (or digital doc)
• Developer Manifesto Template
Instructional Sequence
15m
The "Unfair" Sorting Hat
Engage students with a simulation. Sort students into groups based on a hidden, arbitrary "algorithm" (e.g., shoe color). Discuss how it feels to be sorted by opaque criteria.
Key Question: How do we know if a system is "objective"?
25m
Mechanics of Bias
Use Code Ethics Slides to explain the feedback loop: Training Data → Learning Model → Predictions. Focus on historical and sample bias .
Vocabulary: Training Data, Transparency, Proxy Variables.
40m
Activity: Bias Detective
Distribute Bias Detective Activity . Students act as "Ethics Auditors" for a company whose AI is excluding qualified candidates from specific zip codes.
Task: Identify the biased data point and suggest a technical fix.
20m
The Developer's Manifesto
Students draft an "Ethical Code" for software dev. Use the Ethics Auditor Rubric to evaluate critical thinking and proposed safeguards.
Teacher Pro-Tips & Differentiation
Support
Provide a "Bias Word Bank" and pre-highlighted data sets for students. Use visual diagrams of neural networks to show where data enters the system.
Extension
Challenge students to research "Generative AI" bias (e.g., image generators) and how those models inherit biases from the entire internet.
Code Ethics Slides SYSTEM_AUDIT // UNIT_FINAL
The Ethical Algorithm
Power, Bias, and Responsibility in the Age of Machine Learning
The Core Conflict
Code is Logic
If algorithms are just math and data, they should be objective, fair, and neutral.
Data is Human
Data reflects human history, which includes prejudice, inequality, and error.
"Algorithms are opinions embedded in code." — Cathy O'Neil, Weapons of Math Destruction
The Feedback Loop
Training Data
Historical records used to "teach" the AI.
The Model
Mathematical patterns are recognized.
Biased Outcome
The AI reproduces past patterns of inequality.
Garbage In, Garbage Out: If the data is biased, the math will be too.
Case Study #1
The Visibility Gap
Researchers found that facial recognition AI had error rates up to 35% for darker-skinned women, while only 0.8% for lighter-skinned men.
Cause: Training sets were 75% white and 80% male.Consequence: False arrests and surveillance errors.
🕵️♀️
Analyzing Data Points
The Developer's Burden
As creators of code, you are no longer just "engineers." You are the architects of social opportunity.
Vigilance
Question every dataset source.
Voice
Speak up when models look unfair.
Value
Prioritize safety over speed.
Bias Detective Activity Sheet Case File: Bias Detective
CODE_AUDIT // SECTION_04
NAME: ____________________________________
DATE: __________________
The Mission
A tech giant, Apex Systems , has just released "Auto-Hire 3000," an AI that ranks job applicants for software engineering roles. However, reports show that the AI is ranking candidates from Tech-Valley High School higher than all others. You have been hired as an Ethics Auditor to find the bug in the data.
Evidence Log: Sample Training Data
Applicant ID Years Exp. Coding Score (1-100) School Zip Code Hired? (Last 10 yrs) #9012 2 92 94025 (Elite) YES #4451 5 95 90012 (Urban) NO #3321 1 88 94025 (Elite) YES #1109 8 99 91104 (Suburban) NO #6722 3 91 94025 (Elite) YES
1. THE SMOKING GUN: Look at the "Hired?" column. What pattern do you notice about candidates from Zip Code 94025 compared to the others?
2. THE PROXY VARIABLE: The AI wasn't told to "discriminate" based on income. Explain how using Zip Code as a feature might introduce unintentional bias if 94025 is a high-income area.
3. SYSTEM REPAIR: As the Lead Developer, which piece of data would you remove from the training set to make the model fairer? Why?
4. THE DEVELOPER'S OATH: Write one sentence that every software engineer should say before launching an AI that makes decisions about people's lives.
AUDIT COMPLETE
Ethics Auditor Rubric Ethics Auditor Rubric
Evaluation for The Ethical Algorithm Unit
STUDENT: ___________________________
SCORE: ________ / 20
Category Expert (4-5) Proficient (2-3) Novice (0-1) Identifying Bias Clearly identifies specific proxy variables and explains how they lead to skewed results. Identifies that the system is "unfair" but struggles to name the technical cause. Unable to identify why the algorithm is producing biased outcomes. Societal Impact Articulates long-term consequences for communities and systemic inequality. Recognizes people are hurt, but limits analysis to individual cases. Shows little understanding of how software affects human lives. Technical Solutions Proposes valid technical fixes (e.g., feature removal) and explains trade-offs. Suggests general fixes like "make it better" without concrete technical steps. Proposals are non-technical or do not address root causes. Ethical Logic Manifesto is comprehensive and reflects a strong internal developer compass. Manifesto contains basic rules but lacks depth or professional tone. Ethical rules are missing or incoherent.
Auditor Notes & Feedback
Apex Systems Compliance Dept // Internal Use Only
CERTIFIED ETHICS AUDIT
