About
Hello there! I’m Kehang.
I’m a third year Ph.D. candidate at Harvard University. I’ve been jointly advised by Prof. John Horton from MIT Sloan’s IT group and Prof. David Parkes from Harvard’s EconCS group.
My Research
The central problem I aim to address is how to make mechanism design more applicable in real-world scenarios. Many theoretically optimal mechanisms are seldom applied in practice due to their complexity.
My previous work focuses on using Large Language Models (LLMs) as simulated agents — Homo Silicus — in traditional economic lab experiments. By comparing benchmarks from both economic theory and experiment, we found that LLM agents exhibit many cognitive constraints similar to those of real humans, i.e., LLMs behave in ways akin to humans. The Homo Silicus approach provides new empirical insights for mechanism designers.
My recent work concentrates on simplifying mechanisms through the involvement of LLM proxies. People can express their intentions in natural language, and an LLM agent will act as their proxy within the mechanism.
Working paper:
Automated Social Science: Language Models as Scientist and Subjects
(With John Horton and Benjamin Manning) Slide
Selected Media: Marginal revolution, AI Breakfast, LLM in Science, The Future of Being Human, One Useful Thing, AI in Education
Evidence from the Synthetic Laboratory: Language Models as Auction Participants
- (With Anand Shah, Yanchen Jiang, John Horton and David Parkes)
- Accepted by Conference of Economics and Computation 2024 as Poster
My Journey So Far
Before joining Harvard, I was very lucky, undeservedly so, to be host by Nobel Laureate in Physics Prof. Frank Wilczek in MIT. I was working on using Quantum field theory to model correlated dynamics inside a class of material (quantum spin ice) News.
My research interests turned to human and society after an enriching collaboration with Prof. Hanspeter Pfister from Harvard Viusal Computing Group. I ran large-scale human-subject studies to investigate the mental models of people making sense of visualizations CHI 24’. Reading Between the Pixels: Investigating the Barriers to Visualization Literacy(/files/reading-between-pixels.pdf)
A Pinch of Extra
When I’m not geeking out over research, you can find me capturing the world through my camera lens 📸, embracing the beauty of nature 🌲, hitting the ski slopes 🎿, or trekking through scenic trails 🥾.
Selected Honors
- Introduction to Technical AI Safety Fellowship, 2024
- Purcell Fellowship (Harvard), 2021
- Guo Moruo Scholarship (Highest honor for USTC undergrad students), 2020
- Yan Jici Scholarship (Highest honor for Physics department undergrad students), 2020