Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session A: Tuesday, August 12, 1:30 – 4:30 pm, de Brug & E‑Hall

Suboptimal human decision-making reflects an efficient information bottleneck on inference

Jacob A Parker¹, Kristen Li², Alexandre Filipowicz³, Vijay Balasubramanian¹, Joseph Kable, Joshua I Gold; ¹University of Pennsylvania, University of Pennsylvania, ²University of Pennsylvania, ³Toyota Research Institute

Presenter: Jacob A Parker

Human decision-making behavior varies widely across individuals and task conditions. This variability is often interpreted as a variety of suboptimal inference strategies, but the principles that govern these suboptimalities are not well understood. We propose that one major source of variability in suboptimal decision-making reflects a specific form of bounded rationality that involves capacity-limited inference. We developed and used new theoretical and empirical approaches to study capacity-limited inference based on the information-bottleneck framework. These approaches allowed us to relate the amount of information used (capacity), to the effectiveness with which it was used (accuracy), by individual human subjects performing a variety of inference tasks. We found that substantial variability both within and across subjects reflected optimal capacity-accuracy trade-offs. Strikingly, the same capacity-accuracy tradeoffs were evident among those using heuristic (biased) inference strategies, which inherently failed to maximize performance for a given level of information use but nonetheless appeared to be implemented in a similarly capacity-limited manner. The results imply that human inference reflects consequential, and flexible, capacity limitations that impose structure on suboptimal choice behavior.

Topic Area: Predictive Processing & Cognitive Control

Extended Abstract: Full Text PDF