Deciding Not to Decide

Hillary-Clinton-lecture-jpgThis post is an executive summary of a 2013 paper about deciding not to decide. (“Deciding Not to Decide: Computational and Neural Evidence for Hidden Behavior in Sequential Choice,”  by Sebastian Gluth, Jorg Rieskamp, and Christian Buchel, that appeared in PLoS Comput Bio 9(10). Quite frankly the detail of the paper is beyond me, but the general ideas are interesting.

Many decisions are not triggered by a single event but based on multiple sources of information. When purchasing a new computer, for instance, we certainly look at the price, but not without accounting for further aspects like capabilities, quality and appearance. According to Gluth et al, usually, these multi-attribute decisions evolve sequentially, that is, as long as the collected evidence is insufficient to motivate a particular choice we search for more information to resolve our uncertainty. Importantly, such ‘‘decisions not to decide’’ are not directly observable but can promote significant changes in behavior.

Gluth explains that the timing of decisions is well captured by sequential sampling models (SSMs), a mathematical approach that allows making inferences on both, how and when people decide. The core structure of an SSM consists of an evidence accumulation process that proceeds until an internal criterion (a decision threshold) is met and a specific response is elicited.  They are used to model rapid as well as slow decisions, which may last up to several seconds. However, even though the assumption of a time consuming accumulation process implies that decisions are delayed until a threshold has been reached, an explicit decision not to decide is typically not considered by SSMs. By modeling response time distributions in a sequential choice paradigm, Gluth et al demonstrate that people decide not to decide when given the opportunity to sample more information. Importantly, this explicit decision to wait is distinguishable from an implicit delay in ongoing decisions as it actively inhibits this ongoing process.

Using functional magnetic resonance imaging (fMRI) and electroencephalography (EEG),  they support the computational results by evidence from time-frequency analyses of the EEG data showing that the decision not to decide is accompanied by an increase in oscillatory power in the beta band (14–30 Hz),  which apparently is a well established neural marker of the active inhibition of motor responses.  The obtained pattern is consistent with the authors’ hypothesis that participants repeatedly alternated between considering and postponing the decision in the
sequential task.

Their study further advocates combining computational modeling with brain imaging techniques, as both methodologies attempt to look into the ‘‘black box’’ of the human
mind. In this context, the temporal precision of EEG nicely dovetailed with the fine-grained response time modeling allowing the authors to discover a decision that would otherwise not have been observable.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.