Stanford Seminar - Human-AI Interaction Under Societal Disagreement

Explore a thought-provoking Stanford seminar on human-AI interaction in the face of societal disagreement. Delve into the challenges of developing machine learning algorithms that must navigate conflicting perspectives on ground truth across various AI applications. Learn about Jury Learning, an innovative interactive AI architecture that allows developers to explicitly consider whose voices should influence model predictions. Discover the Disagreement Deconvolution metric, which reveals how current evaluation methods may overstate the performance of user-facing tasks. Gain insights into a new pipeline for encoding human values and goals in AI systems, bridging HCI principles with machine learning realities. Presented by Mitchell Gordon, a Stanford University PhD student in Human-Computer Interaction, this 53-minute seminar offers valuable perspectives on addressing societal disagreements in AI development and evaluation.