(click here to get back to the root)

Personal Perspectives on AI Behavioral Science

On top of my research interests, this page reflects some of my additional thoughts on AI Behavioral Science. Further details and community discussions about this emerging field can be found in our perspective paper and recent workshops.

Why studying AI?

It’s about the two ultimate goals of doing science and research:

• To understand the world: As AI enters the society and interacts with us, it becomes an inseparable part of the world around us. So to fully understand the world, one needs to understand AI.
• To understand ourselves: For the very first time in the history, humans have “cyber siblings”. Posessing human-like (or even superhuman) intelligence, AI is a mirror that reflects on ourselves.

Why behavioral science (or why studying AI at the behavioral level)?

It provides a perfect scientific leverage:

• Behavior is observable and measurable.
• Behavior is ubiquitous, a generalizable representation:
  • For both humans and machines;
  • Across domains (e.g., economics, healthcare, education).
• Behavior comes from interactions. It can be intervened and generate impact, closing the causal loop.
• Behavior is at the meso level, bridging the micro level (e.g., neural networks, cognitive processes) and the macro level (e.g., social systems, ecological systems).