Research
Research
Wisdom of crowds, information aggregation, group decision making, forecasting
Decision analysis, social network analysis, text analysis
Bayesian statistics, applied statistics, statistical machine learning
Huang, S., Broomell, S. B., & Golman, R. A hypothesis test algorithm for determining when weighting individual judgments reliably improves collective accuracy or just adds noise. Decision (2022).
Huang, S., Golman, R. The collective wisdom of behavioral game theory. Economy Theory (2024).
Huang, S., Golman, R., & Broomell, S. B. "Combining the Aggregated Forecasts: An Efficient Method for Improving Accuracy by Stacking Multiple Weighting Models", Decision (Accepted in May 2024).
Huang, S., Golman, R., & Broomell, S. B. "A Regularized Weighting Model of Judgment Aggregation", Decision Analysis (R&R) [SSRN].
Wang, C., Huang, S., Jia, H., Wang, X. "Dynamic Expert Elicitation in Multivariate Risk Assessment", Operations Research (R&R).
Getting More Wisdom from the Crowd: Using Topic Modeling of Explanations to Improve the Accuracy of Collective Forecasts (with Russell Golman and Stephen Broomell)
We use text analysis techniques incorporating topic modeling and Natural Language Processing methodologies (i.e., BERT) to extract text features and create a profile for each person. Then we establish multiple prediction models to estimate the statistical characteristics of individuals' judgments (i.e., judgment validity, variance and the correlation matrix) based on text features. Combining with the theoretically optimal weighting method from the wisdom-of-crowds theory, we develop a weighting scheme embedded with parameter estimation using text analysis inputs. This novel weighting method can be applied to online prediction platforms where users provide both their forecasts and textual comments.
Wisdom of Crowds in Social Networks (with Russell Golman and Stephen Broomell)
We proposed a framework to extract forecasters’ characteristics by analyzing the topological structure of their social network, and to use these estimated properties to find a better way to combine their forecasts. This research can reveal the effect of social influence on one's judgment, and can help managers identify influential and informative opinions in social networks.