Theories of reward learning in neuroscience have focused on two families of algorithms, thought to capture deliberative vs. habitual choice. "Model-based" algorithms compute the value of candidate actions from scratch, whereas "model-free" algorithms make choice more efficient but less flexible by storing pre-computed action values. We examine an intermediate algorithmic family, the successor representation (SR), which balances flexibility and efficiency by storing partially computed action values: predictions about future events. These pre-computation strategies differ in how they update their choices following changes in a task. SR's reliance on stored predictions about future states predicts a unique signature of insensitivity to changes in the task's sequence of events, but flexible adjustment following changes to rewards. We provide evidence for such differential sensitivity in two behavioral studies with humans. These results suggest that the SR is a computational substrate for semi-flexible choice in humans, introducing a subtler, more cognitive notion of habit.

Humans and animals are capable of evaluating actions by considering their long-run future rewards through a process described using model-based reinforcement learning (RL) algorithms. The mechanisms by which neural circuits perform the computations prescribed by model-based RL remain largely unknown; however, multiple lines of evidence suggest that neural circuits supporting model-based behavior are structurally homologous to and overlapping with those thought to carry out model-free temporal difference (TD) learning. Here, we lay out a family of approaches by which model-based computation may be built upon a core of TD learning. The foundation of this framework is the successor representation, a predictive state representation that, when combined with TD learning of value predictions, can produce a subset of the behaviors associated with model-based learning, while requiring less decision-time computation than dynamic programming. Using simulations, we delineate the precise behavioral capabilities enabled by evaluating actions using this approach, and compare them to those demonstrated by biological organisms. We then introduce two new algorithms that build upon the successor representation while progressively mitigating its limitations. Because this framework can account for the full range of observed putatively model-based behaviors while still utilizing a core TD framework, we suggest that it represents a neurally plausible family of mechanisms for model-based evaluation.

Making decisions in sequentially structured tasks requires integrating distally acquired information. The extensive computational cost of such integration challenges planning methods that integrate online, at decision time. Furthermore, it remains unclear whether ‘offline’ integration during replay supports planning, and if so which memories should be replayed. Inspired by machine learning, we propose that (a) offline replay of trajectories facilitates integrating representations that guide decisions, and (b) unsigned prediction errors (uncertainty) trigger such integrative replay. We designed a 2-step revaluation task for fMRI, whereby participants needed to integrate changes in rewards with past knowledge to optimally replan decisions. As predicted, we found that (a) multi-voxel pattern evidence for off-task replay predicts subsequent replanning; (b) neural sensitivity to uncertainty predicts subsequent replay and replanning; (c) off-task hippocampus and anterior cingulate activity increase when revaluation is required. These findings elucidate how the brain leverages offline mechanisms in planning and goal-directed behavior under uncertainty.

Memory and planning rely on learning the structure of relationships among experiences. Compact representations of these structures guide flexible behavior in humans and animals. A century after 'latent learning' experiments summarized by Tolman, the larger puzzle of cognitive maps remains elusive: how does the brain learn, generalize, and transfer relational structures? This review focuses on a reinforcement learning (RL) approach to learning compact representations of the structure of states. We review evidence showing that capturing structures as predictive representations updated via replay offers a neurally plausible account of human behavior and the neural representations of predictive cognitive maps. We highlight multi-scale successor representations, prioritized replay, and policy-dependence. These advances call for new directions in studying the entanglement of learning and memory with prediction and planning.

The development of shared memories, beliefs, and norms is a fundamental characteristic of human communities. These emergent outcomes are thought to occur owing to a dynamic system of information sharing and memory updating, which fundamentally depends on communication. Here we report results on the formation of collective memories in laboratory-created communities. We manipulated conversational network structure in a series of real-time, computer-mediated interactions in fourteen 10-member communities. The results show that mnemonic convergence, measured as the degree of overlap among community members' memories, is influenced by both individual-level information-processing phenomena and by the conversational social network structure created during conversational recall. By studying laboratory-created social networks, we show how large-scale social phenomena (i.e., collective memory) can emerge out of microlevel local dynamics (i.e., mnemonic reinforcement and suppression effects). The social-interactionist approach proposed herein points to optimal strategies for spreading information in social networks and provides a framework for measuring and forging collective memories in communities of individuals.mnemonic reinforcement effect | socially-shared retrieval-induced forgetting | social networks | collective memories | emergent phenomena