In the WebSurf task, humans forage for videos paying costs in terms of wait times on a time-limited task. A variant of the task in which demands during the wait time were manipulated revealed the role of attention in susceptibility to sunk costs. Consistent with parallel tasks in rodents, previous studies have found that humans (undergraduates measured in lab) preferred shorter delays, but waited longer for more preferred videos, suggesting that they were treating the delays economically. In an Amazon Mechanical Turk (mTurk) sample, we replicated these predicted economic behaviors for a majority of participants. In the lab, participants showed susceptibility to sunk costs in this task, basing their decisions in part on time they have already waited, which we also observed in the subset of the mTurk sample that behaved economically. In another version of the task, we added an attention check to the wait phase of the delay. While that attention check further increased the proportion of subjects with predicted economic behaviors, it also removed the susceptibility to sunk costs. These findings have important implications for understanding how cognitive processes, such as the deployment of attention, are key to driving re-evaluation and susceptibility to sunk costs.

In this study, we tested the relationship between error-related signals of cognitive control and cortisol reactivity, investigating the hypothesis of common systems for cognitive and emotional self-regulation. Eighty-three participants completed a Stroop task while electroencephalography (EEG) was recorded. Three error-related indices were derived from the EEG: the error-related negativity (ERN), error positivity (Pe), and error-related alpha suppression (ERAS). Pre- and posttask salivary samples were assayed for cortisol, and cortisol change scores were correlated with the EEG variables. Better error–correct differentiation in the ERN predicted less cortisol increase during the task, whereas greater ERAS predicted greater cortisol increase during the task; the Pe was not correlated with cortisol changes. We concluded that an enhanced ERN, part of an adaptive cognitive control system, predicts successful stress regulation. In contrast, an enhanced ERAS response may reflect error-related arousal that is not adaptive. The results support the concept of overlapping systems for cognitive and emotional self-regulation.

Depression is clinically characterized by obvious changes in decision making that cause distress and impairment. though several studies suggest impairments in depressed individuals in single tasks, there has been no systematic investigation of decision making in depression across tasks. We compare participants diagnosed with Major Depressive Disorder (MDD) (n = 64) to healthy controls (n = 64) using a comprehensive battery of nine value-based decision-making tasks which yield ten distinct measures. MDD participants performed worse on punishment (d = −0.54) and reward learning tasks (d = 0.38), expressed more pessimistic predictions regarding winning money in the study (d = −0.47) and were less willing to wait in a persistence task (d = −0.39). Performance on learning, expectation, and persistence tasks each loaded on unique dimensions in a factor analysis and punishment learning and future expectations each accounted for unique variance in predicting depressed status. Decision-making performance alone could predict depressed status out-of-sample with 72% accuracy. The findings are limited to MDD patients ranging between moderate to severe depression and the effects of medication could not be accounted for due to the cross sectional nature of the study design. These results confirm hints from single task studies that depression has the strongest effects on reinforcement learning and expectations about the future. our results highlight the decision processes that are impacted in major depression, and whose further study could lead to a more detailed computational understanding of distinct facets of this heterogeneous disorder.

Sunk cost sensitivity describes escalating decision commitment with increased spent resources. On neuroeconomic foraging tasks, mice, rats, and humans show similar escalations from sunk costs while quitting an ongoing countdown to reward. In a new analysis taken across computationally parallel foraging tasks across species and laboratories, we find that these behaviors primarily occur on choices that are economically inconsistent with the subject’s other choices, and that they reflect not only the time spent, but also the time remaining, suggesting that these are change-of-mind re-evaluation processes. Using a recently proposed change-of-mind drift-diffusion model, we find that the sunk cost sensitivity in this model arises from decision-processes that directly take into account the time spent (costs sunk). Applying these new insights to experimental data, we find that sensitivity to sunk costs during re-evaluation decisions depends on the information provided to the subject about the time spent and the time remaining.

An engineer's viewpoint on psychiatry asks: What are the failure modes that underlie psychiatric dysfunction? And: How can we modify the system? Psychiatry has made great strides in understanding and treating disorders using biology; however, failure modes and modification access points can also exist extrinsically in environmental interactions. The network analysis suggested by Borsboom et al. in the target article provides a new viewpoint that should be incorporated into current theoretical constructs, not placed in opposition to them.