The computational cost of active information sampling before decision-making under uncertainty

Abstract

Humans often seek information to minimise the pervasive effect of uncertainty on decisions. Current theories explain how much knowledge people should gather prior to a decision, based on the cost-benefit structure of the problem at hand. Here, we demonstrate that this framework omits a crucial agent-related factor: the cognitive effort expended while collecting information. Using an active sampling paradigm, we unveil a speed-efficiency trade-off whereby more informative samples take longer to find. Crucially, under sufficient time pressure, humans can break this trade-off, sampling both faster and more efficiently. Computational modelling demonstrates the existence of a cost of cognitive effort which, when incorporated into theoretical models, provides a better account of people's behaviour, and also relates to self-reported fatigue accumulated during active sampling. Thus, the way people seek knowledge to guide their decisions is shaped not only by task-related costs and benefits, but also crucially by the quantifiable computational costs incurred.

Control over sampling boosts numerical evidence processing in human decisions from experience

“…Also in our experiment, participants in the full-control condition chose to sample less than they could have 61 . Although one explanation is that small samples can render choices objectively simpler 47,62 , our findings suggest that small samples may also defy typical accuracy trade-offs if the decision to stop sampling lies in the autonomy of the sampling agent (see also 63 ). Granting participants full control over sampling may thus not only enable but directly promote reliance on small samples through more efficient processing of the sample evidence.…”

Aversive motivation and cognitive control

“…The other key component in the EVC model is the Cost of cognitive control, which refers to the aversiveness of the mental effort required to exert cognitive control and successfully perform the task (Kool & Botvinick, 2018;Shenhav et al, 2017). This cost is assumed to be a monotonic but likely non-linear function (e.g., quadratic) of the intensity of control being allocated (Massar et al, 2020;Petitet et al, 2021;Soutschek & Tobler, 2020;Vogel et al, 2020). Because the model assumes that it is optimal to maximize drift rate, the drift rate would not be constrained without a cost function.…”

Do humans prefer cognitive effort over doing nothing?

“…Known as the explore/exploit tradeoff (Cohen et al, 2007), theorists have recently posited that boredom acts as a signal to start exploring one's environment to gain valuable information (Agrawal et al, 2021). That is, given that dull tasks typically provide little useful information, boredom might instigate the drive to explore the environment in search of more rewarding tasks (i.e., engaging in effort; Petitet et al, 2021). For instance, tasks that provide little valuable information are perceived as more boring and are quicker to be skipped in favor of an alternative task (Geana et al, 2016).…”