
Responsible AI (h)as a Learning and Design Problem
Michael Madaio, a Senior Research Scientist at Google Research, argues that responsible AI development is as much a learning problem as a technical one. Speaking in Stanford's CS547 Human-Computer Interaction Seminar, he presents findings from his research on how AI practitioners in industry actually learn about responsible AI on the job, often without the skills or training needed to use existing toolkits effectively, especially as pre-trained models lower the barrier to building AI applications. He discusses opportunities to support practitioners' in-situ learning during the design process itself, drawing on his background in human-computer interaction and prior work at Microsoft Research's FATE group on fairness, accountability, transparency, and ethics. The talk closes with open questions about how HCI, design, and the learning sciences can contribute to building AI systems responsibly, framing the problem as organizational and educational rather than purely algorithmic.