Human Cognition Talk: Understanding semantic change with cognitive impairment, and improving the wisdom of the crowd: Two applications of cognitive modeling in solution-oriented science.

 Solution-oriented science is the idea that research aiming to solve practical problems leads not only to useful applied outcomes, but also benefits basic theoretical progress. We present two projects using this approach, both of which rely on the development of cognitive models and the use of Bayesian methods.

The first project asks how people's semantic memory changes over the course of cognitive impairment. We use real-world clinical data from a task that requires people to identify the “odd one out” from a set of three animal names. We develop a cognitive model of the task and use it to test competing theoretical claims about why people's choices change. Contrary to the dominant claim, we find no evidence that the semantic representation of the animals changes. Instead, changes in performance can be explained in terms of worsening access to memory and the use of compensating response strategies.

The second project involves the wisdom of the crowd phenomenon, in which aggregated group decisions outperform most or even all of the individuals in the group. We show that cognitive models can further improve crowd performance in four ways. They can infer and upweight expertise among individuals. They can debias cognitive processes by inferring what people know from how they behave. They can provide a representational scaffolding for combining knowledge that is distributed across individuals. And they can maintain the diversity of a crowd by using model-based predictions as a surrogate for unavailable behavioral data. We demonstrate these ideas in a range of practical settings including probability estimation, ranking, competitive bidding, and sequential decision making.