Humans can navigate complex graph structures acquired during latent learning.

Humans appear to represent many forms of knowledge in associative networks whose nodes are multiply connected, including sensory, spatial, and semantic. Recent work has shown that explicitly augmenting artificial agents with such graph-structured representations endows them with more human-like capabilities of compositionality and transfer learning. An open question is how humans acquire these representations. Previously, it has been shown that humans can learn to navigate graph-structured conceptual spaces on the basis of direct experience with trajectories that intentionally draw the network contours (Schapiro, Kustner, & Turk-Browne, 2012; Schapiro, Turk-Browne, Botvinick, & Norman, 2016), or through direct experience with rewards that covary with the underlying associative distance (Wu, Schulz, Speekenbrink, Nelson, & Meder, 2018). Here, we provide initial evidence that this capability is more general, extending to learning to reason about shortest-path distances across a graph structure acquired across disjoint experiences with randomized edges of the graph - a form of latent learning. In other words, we show that humans can infer graph structures, assembling them from disordered experiences. We further show that the degree to which individuals learn to reason correctly and with reference to the structure of the graph corresponds to their propensity, in a separate task, to use model-based reinforcement learning to achieve rewards. This connection suggests that the correct acquisition of graph-structured relationships is a central ability underlying forward planning and reasoning, and may be a core computation across the many domains in which graph-based reasoning is advantageous.

Increased and biased deliberation in social anxiety.

A goal of computational psychiatry is to ground symptoms in basic mechanisms. Theory suggests that avoidance in anxiety disorders may reflect dysregulated mental simulation, a process for evaluating candidate actions. If so, these covert processes should have observable consequences: choices reflecting increased and biased deliberation. In two online general population samples, we examined how self-report symptoms of social anxiety disorder predict choices in a socially framed reinforcement learning task, the patent race, in which the pattern of choices reflects the content of deliberation. Using a computational model to assess learning strategy, we found that self-report social anxiety was indeed associated with increased deliberative evaluation. This effect was stronger for a particular subset of feedback ('upward counterfactual') in one of the experiments, broadly matching the biased content of rumination in social anxiety disorder, and robust to controlling for other psychiatric symptoms. These results suggest a grounding of symptoms of social anxiety disorder in more basic neuro-computational mechanisms.

Context-sensitive valuation and learning.

A variety of behavioral and neural phenomena suggest that organisms evaluate outcomes not on an absolute utility scale, but relative to some dynamic and context-sensitive reference or scale. Sometimes, as in foraging tasks, this results in sensible choices; in other situations, like choosing between options learned in different contexts, irrational choices can result. We argue that what unites and demystifies these various phenomena is that the brain's goal is not assessing utility as an end in itself, but rather comparing different options to choose the better one. In the presence of uncertainty, noise, or costly computation, adjusting options to the context can produce more accurate choices.

Stimulus generalization as a mechanism for learning to trust.

How do humans learn to trust unfamiliar others? Decisions in the absence of direct knowledge rely on our ability to generalize from past experiences and are often shaped by the degree of similarity between prior experience and novel situations. Here, we leverage a stimulus generalization framework to examine how perceptual similarity between known individuals and unfamiliar strangers shapes social learning. In a behavioral study, subjects play an iterative trust game with three partners who exhibit highly trustworthy, somewhat trustworthy, or highly untrustworthy behavior. After learning who can be trusted, subjects select new partners for a second game. Unbeknownst to subjects, each potential new partner was parametrically morphed with one of the three original players. Results reveal that subjects prefer to play with strangers who implicitly resemble the original player they previously learned was trustworthy and avoid playing with strangers resembling the untrustworthy player. These decisions to trust or distrust strangers formed a generalization gradient that converged toward baseline as perceptual similarity to the original player diminished. In a second imaging experiment we replicate these behavioral gradients and leverage multivariate pattern similarity analyses to reveal that a tuning profile of activation patterns in the amygdala selectively captures increasing perceptions of untrustworthiness. We additionally observe that within the caudate adaptive choices to trust rely on neural activation patterns similar to those elicited when learning about unrelated, but perceptually familiar, individuals. Together, these findings suggest an associative learning mechanism efficiently deploys moral information encoded from past experiences to guide future choice.

Episodic memories predict adaptive value-based decision-making.

Prior research illustrates that memory can guide value-based decision-making. For example, previous work has implicated both working memory and procedural memory (i.e., reinforcement learning) in guiding choice. However, other types of memories, such as episodic memory, may also influence decision-making. Here we test the role for episodic memory-specifically item versus associative memory-in supporting value-based choice. Participants completed a task where they first learned the value associated with trial unique lotteries. After a short delay, they completed a decision-making task where they could choose to reengage with previously encountered lotteries, or new never before seen lotteries. Finally, participants completed a surprise memory test for the lotteries and their associated values. Results indicate that participants chose to reengage more often with lotteries that resulted in high versus low rewards. Critically, participants not only formed detailed, associative memories for the reward values coupled with individual lotteries, but also exhibited adaptive decision-making only when they had intact associative memory. We further found that the relationship between adaptive choice and associative memory generalized to more complex, ecologically valid choice behavior, such as social decision-making. However, individuals more strongly encode experiences of social violations-such as being treated unfairly, suggesting a bias for how individuals form associative memories within social contexts. Together, these findings provide an important integration of episodic memory and decision-making literatures to better understand key mechanisms supporting adaptive behavior.