Stanford OnlineStanford CS547 HCI Seminar | Autumn 2026 | Designing AI for Societal Good
Stanford Online
Why measure harm 0:10
Chen says generative AI can shape society in small, slow ways that add up. Effects on loneliness, creativity, job quality, and mental well-being may be hard to measure at first, but waiting for full proof can mean missing the chance to correct course. The aim is to act before these harms are deeply set, even when they are still only suspected.
Beyond safer products 11:02
The talk argues that making a less harmful chatbot is not enough. A better goal is to build harm measurement and harm mitigation systems that can show effects at a social scale and help shift public opinion, policy, and business models. The speaker frames this as work for system builders, not only activists or policy people, and then turns to history for guidance.
Policy and public opinion 19:00
Policy change matters, but less than the talk first suggests. In a democracy, policy cannot move far ahead of public opinion, so people need civic understanding before lawmakers can act. If citizens already see that chatbots may hurt creativity, learning, or well-being, they could choose to use them less without waiting for a rule. The same logic appears in measurement work. Researchers often begin with humble tools that make harm visible. Smartphone overuse is one example. HCI researchers measured whether people felt their phone time was time well spent, because raw minutes were not enough. Those studies gave self-improvers something to act on, built evidence before the political moment arrived, and later helped shift public opinion and policy. The same pattern showed up in civic monitoring gadgets for air pollution. Simple measurements turned slippery harms into something trackable and actionable.
Money and market fit 37:32
A social good product has to survive in a market that already has set rules. One example follows a mental health app through hospitals, employers, and insurers. Each buyer wants a different proof. Hospitals want costly randomized trials. Employers want fewer sick days and better work. The team keeps changing the product to fit those demands.
Harm and policy work 43:00
The group then widens the frame. HCI researchers can spot harm early, balance many stakeholders, and work on policy, public opinion, and business models as well as design. They turn to chatbots, measuring harm with proxy measures from cognitive psychology and studying how they affect well-being, identity talk, and social ties.
Evidence synthesis matters 56:00
Think tanks often gather scientific evidence from many scholars for policy work. There are also calls for more HCI researchers to do evidence reviews instead of only small studies. That kind of synthesized work matters because it helps shape policy and public opinion.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
