Projects

Inner crowd

The HAL lab investigates how a single person can act as their own “crowd.” Averaging several independent estimates from different people usually beats any one person’s guess — the classic wisdom of crowds. The same trick works within a single mind: when someone estimates a quantity twice and averages both guesses, the average is often more accurate than either guess alone. Our research shows this “wisdom of the inner crowd” can be boosted further by deliberately making the second guess from the perspective of someone you’d typically disagree with. Recently, we’ve also shown this benefit isn’t universal: taking a disagreeing perspective helps mainly for difficult questions, where first guesses tend to be far from the truth — for easy questions, where first guesses are already close, deliberately disagreeing with yourself can do more harm than good.

Human-AI interaction

The HAL lab investigates how people perceive, trust, and collaborate with algorithms and AI systems and how this can affect their accuracy. We study when and why people trust — or distrust — algorithmic advice and predictions relative to human input, including how cues such as response speed shape that trust. Recently, we are investigating how interacting with LLMs can impact people’s prediction accuracy.