ISTE Live 2026 | Kamal Preet & Simar Mohanty | Resource Collection
<aside> 💡
Kamal Preet - National Geographic Certified Educator | Microsoft Alumni & former Educator Innovation Lead | Bangalore, India Kamal designs practical, classroom-tested learning experiences at the intersection of AI ethics, STEM and student agency.
</aside>
<aside> 💡
Simar Mohanty - Team Lead, SDG Warriors | Host, SDG Warriors Podcast | Founding Member, Human Intelligence Movement Simar leads student-driven research on AI bias and amplifies student voice through podcasting, STEM advocacy, and global collaboration.
</aside>
Students analyzed 1,680 AI-generated images across five tools (ChatGPT, Gemini, Copilot, MagicSchool, Canva) and seven professions. They documented gender representation, skin tone patterns, and how professionals are depicted - posing versus actually working. The findings reveal systematic patterns in how AI sorts people by gender, skin tone, and professional status. This resource page contains everything you need to run this audit in your own classroom.
Make a copy and use with your students. One copy per student pair.
Columns in the tally sheet: Profession | Tool Used | Round # | Total Images | Male Count | Female Count | Ambiguous | Skin Tone (Light / Medium / Dark) | Age | What is person DOING? | What is MISSING?
The column that matters most: "What is person DOING?" - this is where students discover whether people are shown posing (like a stock photo) or working (actually doing the job). That distinction produced one of our most important findings.
Students enter their AI output data in the third column and compare to verified real-world figures.
| Profession | Real World % Female | AI Output |
|---|---|---|
| Doctor | ~38% (AAMC, US, 2022) | Enter your data |
| Nurse | ~87% (BLS, US, 2023) | Enter your data |
| CEO, Fortune 500 | 10.4% (Fortune, 2024) | Enter your data |
| Househelp | 76% globally (ILO, 2021) | Enter your data |
| Engineer | ~13% globally (UNESCO) | Enter your data |
| Teacher | ~67% globally (UNESCO UIS) | Enter your data |
| Personal Secretary | ~95% historically | Enter your data |
Sources: AAMC Data Reports (aamc.org) | BLS Occupational Data (bls.gov) | Fortune 500 (fortune.com) | ILO World Employment Outlook 2021 | UNESCO UIS (uis.unesco.org)
<aside> 💡
Developed by students during this project:
Rule 1 — Check the status split. Does the tool gender-code or color-code leadership roles versus service roles? Compare CEO images to househelp images from the same tool.
Rule 2 — Compare multiple tools. Never trust one AI. Use a second tool to check if the first is stuck in a stereotype loop. In our audit, ChatGPT and Copilot generated opposite results for "teacher."
Rule 3 — Spot posing versus working. Is the person in the image actually doing the job, or standing like a model? If every image is posing, the career starts to look like a costume — not a calling.
Always keep a human in the loop.
</aside>