Seminar Video: Prof Judy Gichoya on Shortcut Learning for Medical AI—Now Available

The full recording of Prof Judy Wawira Gichoya’s seminar, “Shortcut Learning for Medical AI: Opportunity or Threat,” is now available to watch online.

About the Seminar

On July 23, 2025, Prof Judy Gichoya—Associate Professor of Radiology at Emory University and Director of the Healthcare AI, Innovation and Translational Informatics (HITI) Lab—delivered an eye-opening talk at the University of Pretoria. The seminar, hosted by the Data Science for Social Impact Research Group (DSFSI) and AfriDSAI, addressed a crucial issue facing AI in healthcare: shortcut learning.

Shortcut learning occurs when machine learning models identify and exploit easy, superficial patterns in data rather than learning the real, underlying clinical features that matter. While this can lead to high test scores in controlled settings, it often results in poor performance, unfair outcomes, and real risks when these systems are deployed in the real world.

Key Highlights

• What shortcut learning looks like in medical AI and why it’s so common
• Real-world examples of shortcuts causing unexpected failures in clinical AI
• The serious implications for fairness, generalizability, and trust
• Strategies to detect, prevent, and mitigate shortcut learning
• Why addressing shortcut learning is essential for building responsible, reliable, and Africa-centric health AI

Prof Gichoya not only described the technical problem, but also challenged researchers, clinicians, and the broader AI community to take practical steps—like improving datasets, adopting causal reasoning, and ensuring deep collaboration between technical and clinical experts.

Watch the Full Seminar

You can watch the full seminar recording here:
https://youtu.be/giaDbLzMbMo

Get Involved

We thank Prof Gichoya for her visit and her leadership in building responsible, equitable medical AI for Africa and the world. For more about DSFSI, AfriDSAI, or to get involved in upcoming events and research collaborations, join the newsletter https://dsup.substack.com

Feel free to share this video with colleagues, students, and anyone passionate about the future of AI in healthcare.