The 2nd Workshop on Human-Centered Recommender Systems

Current recommendation metrics miss what matters most: Traditional success measures like accuracy, clicks, and engagement don't capture what actually benefits people. The workshop emphasizes that the AI industry needs to shift away from these narrow metrics toward measuring outcomes that align with human values like trust, safety, fairness, and well-being.

Recommender systems have outsized influence on society: These systems shape how people discover information, form opinions, and connect with others, yet their design often prioritizes engagement over human benefit. This concentration of influence demands a more responsible approach to how recommendations are built and deployed.

Human values must be built into recommendation design from the start: Rather than treating ethics as an afterthought, the workshop advocates for integrating human-centered values—including transparency, fairness, and safety—directly into the recommendation process itself, not as add-ons.

Solving this requires collaboration across disciplines: Addressing responsible AI in recommendations isn't just a technical problem. The workshop brings together experts in human-computer interaction, AI safety, and social computing alongside computer scientists, recognizing that human-centered design requires diverse perspectives.

Three key areas need focus: understanding users, involving them in decisions, and measuring real impact: The workshop's framework emphasizes not just understanding what people need, but actively involving them in how systems work and measuring whether recommendations actually improve their well-being rather than just driving engagement.