Empowering Teens to Defang Bias in AI with Algorithm Auditing

Yasmin B. Kafai, University of Pennsylvania and John Ottina, Long Beach School District

AI is not a neutral tool. Its powerful algorithms and large data sets have baked-in biases that can amplify stereotypes, limit what information we see, and shape our opportunities. Despite how deeply these algorithmic systems have infiltrated our day-to-day lives, they operate with little transparency, oversight, or mechanisms for evaluation or review. 

The potential consequences of harmful algorithmic bias include getting skewed news, being denied housing or a job, and even being arrested. For young people, biases can influence everything from what videos appear on their TikTok feed to the amount of their college scholarship funding. As educators, we must provide opportunities for our students to be empowered understand and identify these biases firsthand. 

At the keynote talk at the annual Computer Science Teachers Association (CSTA) conference in Cleveland this July, we will share ideas for how teens could be empowered to take control of their digital lives through algorithm auditing. Together with John Ottina, a veteran CS high school teacher, and with other teachers from around the country, we have developed and tested an approach called algorithm auditing for high school students.

Algorithm auditing provides students with hands-on experiences in evaluating AI systems from the outside-in. Widely used in professional contexts such as healthcare diagnostics and employment decision-making, algorithm auditing involves systematically testing AI systems by analyzing their outputs to uncover potential biases and societal impacts (Metaxa et al., 2021). 

Why Algorithm Auditing Belongs in High School Classrooms

During our talk in Cleveland, we will share activities and lessons learned from our work with teachers and students auditing algorithmic systems in the classroom. But for now, we want to share four reasons every student would benefit from investigating  how the algorithmic that shape their lives work. 

• Building workforce skills: Careers spanning fields in data science, AI, and policy require algorithm auditing skills (Bandy, 2021). As policymakers in the United States and abroad take steps to impose external oversight and accountability on these models, the ability to identify discriminatory practices in these algorithms skills will be increasingly meaningful and socially relevant.
• Empowering all users: As AI becomes more widespread, users —including young people—should  be empoweredto evaluate and recognize AI-related issues. This becomes part of AI literacy (Touretzky et al., 2019).
• Strengthening computational thinking and problem-solving: Teaching students to navigate and influence genAI systems equips them with skills in critical thinking, digital literacy, and other important aspects of computational thinking (Wing, 2006). Algorithm auditing allows students to investigate and challenge opaque algorithmic systems without requiring direct access to the underlying code or data. This approach empowers students to recognize potentially harmful biases and malfunctions in real-world contexts.
• Fostering ethical awareness: Auditing AI systems encourages students to think critically about fairness, bias, and accountability in technology. This practice aligns with algorithmic justice research (Costanza-Chock et al., 2022) and supports critical computing education, which integrates ethics into computational thinking.

Every high schooler should have the opportunity to learn about algorithm auditing. The good news is that it can be integrated into high school curricula relatively easily. No advanced coding is required. And teaching students to examine AI equips them with skills they can use all in all aspects of their lives. Plus, once students better understand the limitations of AI systems, algorithm auditing will be an easy sell—after all, teens (like the rest of us) can’t stand the thought of being manipulated or duped.

Works Cited

Bandy, J. (2021). Problematic machine behavior: A systematic literature review of algorithm audits. In Proceedings of the ACM on Human-Computer Interaction, 5(CSCW1), 1-34.

Costanza-Chock, S., Raji, I. D., & Buolamwini, J. (2022, June). Who audits the auditors? Recommendations from a field scan of the algorithmic auditing ecosystem. In Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency (pp. 1571-1583).

Metaxa, D., Park, J. S., Robertson, R. E., Karahalios, K., Wilson, C., Hancock, J., & Sandvig, C. (2021). Auditing algorithms: Understanding algorithmic systems from the outside in. Foundations and Trends® in Human–Computer Interaction, 14(4), 272-344.

Touretzky, D. S., Gardner-McCune, G., Martin, F., & Seehorn, D. (2019). Envisioning AI for k–12: What should every child know about AI? Proceedings of AAAI-19. https://doi.org/10.1609/aaai.v33i01.33019795.

Wing, J. (2006). Computational thinking. Communications of the ACM, 49, 33-35.

About the Authors