John McCoy
I am an Assistant Professor of Marketing at the Wharton School of the University of Pennsylvania.
At Penn, I'm also affiliated with the Wharton Neuroscience Initiative and Penn's MindCORE hub.
Trained as a computational cognitive scientist, I'm broadly interested in understanding how people think and decide, and using this understanding to solve marketing problems.
I use a combination of behavioral experiments and formal modeling, drawing on ideas and techniques from psychology, economics, marketing, Bayesian statistics, and computer science.
Much of my work relates to two main themes: measurement at the level of individuals and crowds of both beliefs and preferences, and consumer semantic cognition. For more details, see my CV.
Selected research
A few recent working papers
-
McCoy, J. "Putnam Products: Is Non-alcoholic Beer Beer or Oat Milk Milk?", 2025
-
Aka, A. and McCoy, J. "Predicting the Memorability of Brand Slogans", 2025
-
McCoy, J. and Prelec, D. "Belief Decomposition: A Mechanism for Collective Inference", 2025
Crowd wisdom (collective-level measurement)
-
My collaborators and I have proposed a new solution to extracting wisdom from the crowd: select not the "most popular" answer, but rather the "surprisingly popular" answer. That is, elicit from respondents their own answer and their prediction about the answers of others and select the answer that is more popular than the crowd predicts.
Representative press:
Wall Street Journal, Bloomberg, Quartz,
MIT news, Scientific American, NPR -
We incorporate the ideas behind the "surprisingly popular answer" into a Bayesian hierarchical model. This has advantages such as producing a posterior distribution over answers, tracking expertise over multiple questions, and allowing for noisy answers.
-
Perhaps the most natural application of crowd wisdom research is to forecasting, and we've surveyed the forecasting literature from perspective of geopolitical forecasting and of consumer behavior, particularly the complex, interacting roles of people and algorithms.
Individual-level measurement, decision making, and semantic cognition
-
People may not even represent basic details of imaginary scenes.
Bigelow, E., McCoy, J.*, and Ullman, T.*. "Non-Commitment in Mental Imagery", Cognition, 2023 (Press: Scientific American Discover Magazine )
-
Words, semantics, and memorability.
Aka, A., Bhatia, S., and McCoy, J. "Semantic determinants of memorability", Cognition, 2023
-
Measuring preferences: in a working paper, we apply conjoint analysis across product categories.
-
Measuring attitudes: we proposed a method for measuring how people perceive groups or agents different to themselves.
McCoy, J.* and Ullman, T.* “A Minimal Turing Test”, Journal of Experimental Social Psychology, 2018
OSF link with pre-print, data and analysis
Representative press:
-
While some of my work has shown that we can measure nuanced information, for example meta-beliefs that aid in crowd-sourcing, people don't always reveal everything that they know. This is, for example, sometimes true of doctors communicating with patients.
-
People's imagination can be a source of insight into their models of the world. For example, how people imagine possible (magical!) worlds relates to their intuitive theories of physics
-
People also imagine choices involving various kinds of transformative experiences, which we can study from the perspective of metaphysics and cognitive science.
-
We examine how people perceive the risk preferences of others, and how people attribute blame.
Graph theory papers
-
I have written some papers in graph theory, mostly dealing with paired domination: