Plinko and Statistical Learning

How can you measure someone's prior? Alex Filipowicz, James Danckert, and I thought long and hard about this. We wanted something to reflect an entire distribution and not just a point estimate like the maximum. We also did not want to make assumptions about the shape of the distribution. Since we all have unique life histories it seemed that we should expect individual variation, and not assume consistency for an entire experimental sample. With the help of many others over a few years the paradigm we developed to address these questions took the form of Plinko, another name for Galton's Bean Machine. With writing and analysis help from Peter Diberardino a collection of experiments has just been published in the British Journal of Psychology. We report that people give consistent reports of their priors, they show learning and context effects, and that in adversarial circumstances their distributions converge towards an optimal one. There is of course much more that can be done and the main purpose of this blog post is to advertise the tool. Take a look if you have time and let me know what you think.