Explaining Human Choice Probabilities with Simple Vector Representations
Peter Diberardino and I have just finished and posted our preprint with the above title to arxiv.
In this work, we have mathematically formalized participant behavior in probability matching using histogram vectors. This representation covers not only the well-studied case of seeking a reward, but also the rarely studied scenario where people try to avoid consequences, and permits us to define probability antimatching in avoidance contexts. The similarities and differences between the seeking and avoiding scenarios, as well as their matching and antimatching counterparts, motivates our theoretical proposal for how decision making under uncertainty varies with the decision maker’s role. Seekers, for whom the cost of failure is usually smaller than their “prey”, choose not merely to succeed, but also to learn what to expect as revealed by probability matching. Whereas in an avoidance scenario, minimizing behavior predominates. A simple vector representation of probability that combines these two modes of responding accounts for the array of human participant choice frequencies seen in our experiments despite varying outcome distributions, whether a participant was prey or predator, and across variation in environment complexity
We have also made the data and analysis files available as a repository at codeberg. Want to double check our stats you can re-run them yourselves. Want to tweak a figure of ours? Edit the Rnw file and re-compile. The goal is to make the research verifiable, reproducible, findable, and extendible. We hope you take advantage of it.