When rewards backfire: Collapsing under pressure in motor cortex.

People often underperform their abilities in high-value situations, a mysterious phenomenon known as "choking under pressure." In this issue of Neuron, Smoulder et al.1 report that target-selective signals in the motor cortex of non-human primates collapse in the face of high-value opportunities.

Learning signatures of decision making from many individuals playing the same game.

Human behavior is incredibly complex and the factors that drive decision making-from instinct, to strategy, to biases between individuals-often vary over multiple timescales. In this paper, we design a predictive framework that learns representations to encode an individual's 'behavioral style', i.e. long-term behavioral trends, while simultaneously predicting future actions and choices. The model explicitly separates representations into three latent spaces: the recent past space, the short-term space, and the long-term space where we hope to capture individual differences. To simultaneously extract both global and local variables from complex human behavior, our method combines a multi-scale temporal convolutional network with latent prediction tasks, where we encourage embeddings across the entire sequence, as well as subsets of the sequence, to be mapped to similar points in the latent space. We develop and apply our method to a large-scale behavioral dataset from 1,000 humans playing a 3-armed bandit task, and analyze what our model's resulting embeddings reveal about the human decision making process. In addition to predicting future choices, we show that our model can learn rich representations of human behavior over multiple timescales and provide signatures of differences in individuals.

Prolonged Physiological Stress Is Associated With a Lower Rate of Exploratory Learning That Is Compounded by Depression.

BACKGROUND: Stress is a major risk factor for depression, and both are associated with important changes in decision-making patterns. However, decades of research have only weakly connected physiological measurements of stress to the subjective experience of depression. Here, we examined the relationship between prolonged physiological stress, mood, and explore-exploit decision making in a population navigating a dynamic environment under stress: health care workers during the COVID-19 pandemic. METHODS: We measured hair cortisol levels in health care workers who completed symptom surveys and performed an explore-exploit restless-bandit decision-making task; 32 participants were included in the final analysis. Hidden Markov and reinforcement learning models assessed task behavior. RESULTS: Participants with higher hair cortisol exhibited less exploration (r = -0.36, p = .046). Higher cortisol levels predicted less learning during exploration (ß = -0.42, false discovery rate [FDR]-corrected p [pFDR] = .022). Importantly, mood did not independently correlate with cortisol concentration, but rather explained additional variance (ß = 0.46, pFDR = .022) and strengthened the relationship between higher cortisol and lower levels of exploratory learning (ß = -0.47, pFDR = .022) in a joint model. These results were corroborated by a reinforcement learning model, which revealed less learning with higher hair cortisol and low mood (ß = -0.67, pFDR = .002). CONCLUSIONS: These results imply that prolonged physiological stress may limit learning from new information and lead to cognitive rigidity, potentially contributing to burnout. Decision-making measures link subjective mood states to measured physiological stress, suggesting that they should be incorporated into future biomarker studies of mood and stress conditions.

Learning signatures of decision making from many individuals playing the same game.

Human behavior is incredibly complex and the factors that drive decision making--from instinct, to strategy, to biases between individuals--often vary over multiple timescales. In this paper, we design a predictive framework that learns representations to encode an individual's 'behavioral style', i.e. long-term behavioral trends, while simultaneously predicting future actions and choices. The model explicitly separates representations into three latent spaces: the recent past space, the short-term space, and the long-term space where we hope to capture individual differences. To simultaneously extract both global and local variables from complex human behavior, our method combines a multi-scale temporal convolutional network with latent prediction tasks, where we encourage embeddings across the entire sequence, as well as subsets of the sequence, to be mapped to similar points in the latent space. We develop and apply our method to a large-scale behavioral dataset from 1,000 humans playing a 3-armed bandit task, and analyze what our model's resulting embeddings reveal about the human decision making process. In addition to predicting future choices, we show that our model can learn rich representations of human behavior over multiple timescales and provide signatures of differences in individuals.