Abstract
Essential to adaptive intelligence is the ability to create mental spaces where knowledge from past experiences cumulates and integrates with newly acquired information. When engaged in decision-making tasks, humans are known to create such a space and therein form decision variables, which integrate task-essential information from multiple sources in a generalizable form. Much effort has focused on the cognitive and neural processes involved in forming decision variables. However, there is limited understanding of how decision variables, once formed, are utilized to adapt to the environment. Considering the abstract and generalizable nature of decision variables, we reason that decision-makers would benefit from shaping and updating probabilistic knowledge—known as belief —within the decision-variable space. As one such belief updating, we hypothesize that an act of decision commitment restricts the current belief about the decision variable to a range of states corresponding to that decision. This implies that past decisions not only attract future ones but also exert a greater pull when those decisions are made with finer granularity—dubbed ‘the granularity effect.’ Here, we present the findings of seven psychophysical experiments that consistently confirm these implications while ruling out the stimulus and action space as potential loci of the granularity effect. Further, as a principled and unified account of the granularity effect and other history effects found in various perceptual tasks, we offer a Bayesian model where beliefs are updated separately in the stimulus and decision-variable spaces. Our work demonstrates how humans leverage the abstract and generalizable nature of the decision-variable space to effectively adapt to their surroundings, expanding the gamut of human intellect.
Competing Interest Statement
The authors have declared no competing interest.
Footnotes
↵✉ e-mail: heeseung1990{at}gmail.com