# "I'm Not Confident in Debiasing AI Systems Since I Know Too Little": Teaching AI Creators About Gender Bias Through Hands-on Tutorials

**Authors:** Kyrie Zhixuan Zhou, Jiaxun Cao, Xiaowen Yuan, Daniel E. Weissglass, Zachary Kilhoffer, Madelyn Rose Sanfilippo, Xin Tong  
**arXiv:** [2309.08121](https://arxiv.org/abs/2309.08121)  
**Date:** 2023-09-15  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2309.08121v1/

## Key points

- Gender bias in AI systems leads to negative user experiences, perpetuates injustices, and causes mental harm to women.
- Current educational curricula for AI creators lack adequate preparation to recognize and mitigate gender bias in their work.
- Hands-on tutorials were designed to raise awareness among AI creators about gender bias in AI systems.
- The tutorials aimed to enhance participants' understanding of the sources of gender bias and debiasing techniques.
- Evaluation with a group of 18 AI creators showed a significant improvement in awareness and knowledge regarding gender bias after engaging with the tutorials.
- The study highlights the potential for educational interventions to complement deficiencies in CS/AI courses related to gender bias education.
- The authors provide design implications and a rubric for future research, educational initiatives, and design efforts addressing gender bias in AI systems.
- Equipping AI creators with tools and knowledge is crucial for effectively tackling gender bias within their work.

## Summary

In their paper titled "I'm Not Confident in Debiasing AI Systems Since I Know Too Little": Teaching AI Creators About Gender Bias Through Hands-on Tutorials, authors Kyrie Zhixuan Zhou, Jiaxun Cao, Xiaowen Yuan, Daniel E. Weissglass, Zachary Kilhoffer, Madelyn Rose Sanfilippo, and Xin Tong address the pervasive issue of gender bias in AI systems. They highlight how this bias leads to negative user experiences and perpetuates injustices while causing mental harm to women. The authors identify a critical gap in current educational curricula for AI creators that fails to adequately prepare them to recognize and mitigate gender bias in their work. To bridge this gap, they designed hands-on tutorials aimed at raising awareness among AI creators about gender bias in AI systems. These tutorials were specifically crafted to enhance participants' understanding of the sources of gender bias and various debiasing techniques that can be employed. The effectiveness of these tutorials was evaluated with a group of 18 AI creators comprising researchers, industrial practitioners (such as developers and product managers), and students with prior knowledge of AI. The results demonstrated a significant improvement in participants' awareness and knowledge regarding gender bias in AI systems after engaging with the tutorials. This positive outcome highlights the potential for such educational interventions to complement existing deficiencies in CS/AI courses related to gender bias education. Based on their findings, the authors provide valuable design implications and a rubric that can guide future research endeavors, educational initiatives, and design efforts aimed at addressing gender bias in AI systems. Overall, this study sheds light on the importance of equipping AI creators with the necessary tools and knowledge to effectively tackle gender bias within their work. By offering practical hands-on tutorials focused on this crucial issue , the authors advocate for a more inclusive and ethically responsible approach to developing AI technologies that prioritize fairness and equality for all users .