Humans can navigate complex graph structures acquired during latent learning

Network structure influences the strength of learned neural representations

From sequences of discrete events, humans build mental models of their world. Referred to as graph learning, the process produces a model encoding the graph of event-to-event transition probabilities. Recent evidence suggests that some networks are easier to learn than others, but the neural underpinnings of this effect remain unknown. Here we use fMRI to show that even over short timescales the network structure of a temporal sequence of stimuli determines the fidelity of event representations as well as the dimensionality of the space in which those representations are encoded: when the graph was modular as opposed to lattice-like, BOLD representations in visual areas better predicted trial identity and displayed higher intrinsic dimensionality. Broadly, our study shows that network context influences the strength of learned neural representations, motivating future work in the design, optimization, and adaptation of network contexts for distinct types of learning.

Decomposing dynamical subprocesses for compositional generalization

A striking feature of human cognition is an exceptional ability to rapidly adapt to novel situations. It is proposed this relies on abstracting and generalizing past experiences. While previous research has explored how humans detect and generalize single sequential processes, we have a limited understanding of how humans adapt to more naturalistic scenarios, for example, complex, multisubprocess environments. Here, we propose a candidate computational mechanism that posits compositional generalization of knowledge about subprocess dynamics. In two samples (N = 238 and N = 137), we combined a novel sequence learning task and computational modeling to ask whether humans extract and generalize subprocesses compositionally to solve new problems. In prior learning, participants experienced sequences of compound images formed from two graphs' product spaces (group 1: G1 and G2, group 2: G3 and G4). In transfer learning, both groups encountered compound images from the product of G1 and G3, composed entirely of new images. We show that subprocess knowledge transferred between task phases, such that in a new task environment each group had enhanced accuracy in predicting subprocess dynamics they had experienced during prior learning. Computational models utilizing predictive representations, based solely on the temporal contiguity of experienced task states, without an ability to transfer knowledge, failed to explain these data. Instead, behavior was consistent with a predictive representation model that maps task states between prior and transfer learning. These results help advance a mechanistic understanding of how humans discover and abstract subprocesses composing their experiences and compositionally reuse prior knowledge as a scaffolding for new experiences.

Personalized Cognitive Health in Psychiatry: Current State and the Promise of Computational Methods

BACKGROUND

Decades of research have firmly established that cognitive health and cognitive treatment services are a key need for people living with psychosis. However, many current clinical programs do not address this need, despite the essential role that an individual's cognitive and social cognitive capacities play in determining their real-world functioning. Preliminary practice-based research in the Early Psychosis Intervention Network early psychosis intervention network shows that it is possible to develop and implement tools that delineate an individuals' cognitive health profile and that help engage the client and the clinician in shared decision-making and treatment planning that includes cognitive treatments. These findings signify a promising shift toward personalized cognitive health.

STUDY DESIGN

Extending upon this early progress, we review the concept of interindividual variability in cognitive domains/processes in psychosis as the basis for offering personalized treatment plans. We present evidence from studies that have used traditional neuropsychological measures as well as findings from emerging computational studies that leverage trial-by-trial behavior data to illuminate the different latent strategies that individuals employ.

STUDY RESULT

We posit that these computational techniques, when combined with traditional cognitive assessments, can enrich our understanding of individual differences in treatment needs, which in turn can guide evermore personalized interventions.

CONCLUSION

As we find clinically relevant ways to decompose maladaptive behaviors into separate latent cognitive elements captured by model parameters, the ultimate goal is to develop and implement approaches that empower clients and their clinical providers to leverage individual's existing learning capacities to improve their cognitive health and well-being.

Expanding horizons in reinforcement learning for curious exploration and creative planning

Curiosity and creativity are expressions of the trade-off between leveraging that with which we are familiar or seeking out novelty. Through the computational lens of reinforcement learning, we describe how formulating the value of information seeking and generation via their complementary effects on planning horizons formally captures a range of solutions to striking this balance.

Dorsolateral prefrontal activity supports a cognitive space organization of cognitive control

Cognitive control resolves conflicts between task-relevant and -irrelevant information to enable goal-directed behavior. As conflicts can arise from different sources (e.g., sensory input, internal representations), how a limited set of cognitive control processes can effectively address diverse conflicts remains a major challenge. Based on the cognitive space theory, different conflicts can be parameterized and represented as distinct points in a (low-dimensional) cognitive space, which can then be resolved by a limited set of cognitive control processes working along the dimensions. It leads to a hypothesis that conflicts similar in their sources are also represented similarly in the cognitive space. We designed a task with five types of conflicts that could be conceptually parameterized. Both human performance and fMRI activity patterns in the right dorsolateral prefrontal cortex support that different types of conflicts are organized based on their similarity, thus suggesting cognitive space as a principle for representing conflicts.

The successor representation subserves hierarchical abstraction for goal-directed behavior

Humans have the ability to craft abstract, temporally extended and hierarchically organized plans. For instance, when considering how to make spaghetti for dinner, we typically concern ourselves with useful "subgoals" in the task, such as cutting onions, boiling pasta, and cooking a sauce, rather than particulars such as how many cuts to make to the onion, or exactly which muscles to contract. A core question is how such decomposition of a more abstract task into logical subtasks happens in the first place. Previous research has shown that humans are sensitive to a form of higher-order statistical learning named "community structure". Community structure is a common feature of abstract tasks characterized by a logical ordering of subtasks. This structure can be captured by a model where humans learn predictions of upcoming events multiple steps into the future, discounting predictions of events further away in time. One such model is the "successor representation", which has been argued to be useful for hierarchical abstraction. As of yet, no study has convincingly shown that this hierarchical abstraction can be put to use for goal-directed behavior. Here, we investigate whether participants utilize learned community structure to craft hierarchically informed action plans for goal-directed behavior. Participants were asked to search for paintings in a virtual museum, where the paintings were grouped together in "wings" representing community structure in the museum. We find that participants' choices accord with the hierarchical structure of the museum and that their response times are best predicted by a successor representation. The degree to which the response times reflect the community structure of the museum correlates with several measures of performance, including the ability to craft temporally abstract action plans. These results suggest that successor representation learning subserves hierarchical abstractions relevant for goal-directed behavior.

Network structure influences the strength of learned neural representations

Human experience is built upon sequences of discrete events. From those sequences, humans build impressively accurate models of their world. This process has been referred to as graph learning, a form of structure learning in which the mental model encodes the graph of event-to-event transition probabilities [1], [2], typically in medial temporal cortex [3]-[6]. Recent evidence suggests that some network structures are easier to learn than others [7]-[9], but the neural properties of this effect remain unknown. Here we use fMRI to show that the network structure of a temporal sequence of stimuli influences the fidelity with which those stimuli are represented in the brain. Healthy adult human participants learned a set of stimulus-motor associations following one of two graph structures. The design of our experiment allowed us to separate regional sensitivity to the structural, stimulus, and motor response components of the task. As expected, whereas the motor response could be decoded from neural representations in postcentral gyrus, the shape of the stimulus could be decoded from lateral occipital cortex. The structure of the graph impacted the nature of neural representations: when the graph was modular as opposed to lattice-like, BOLD representations in visual areas better predicted trial identity in a held-out run and displayed higher intrinsic dimensionality. Our results demonstrate that even over relatively short timescales, graph structure determines the fidelity of event representations as well as the dimensionality of the space in which those representations are encoded. More broadly, our study shows that network context influences the strength of learned neural representations, motivating future work in the design, optimization, and adaptation of network contexts for distinct types of learning over different timescales.

Sleep targets highly connected global and local nodes to aid consolidation of learned graph networks

Much of our long-term knowledge is organised in complex networks. Sleep is thought to be critical for abstracting knowledge and enhancing important item memory for long-term retention. Thus, sleep should aid the development of memory for networks and the abstraction of their structure for efficient storage. However, this remains unknown because past sleep studies have focused on discrete items. Here we explored the impact of sleep (night-sleep/day-wake within-subject paradigm with 25 male participants) on memory for graph-networks where some items were important due to dense local connections (degree centrality) or, independently, important due to greater global connections (closeness/betweenness centrality). A network of 27 planets (nodes) sparsely interconnected by 36 teleporters (edges) was learned via discrete associations without explicit indication of any network structure. Despite equivalent exposure to all connections in the network, we found that memory for the links between items with high local connectivity or high global connectivity were better retained after sleep. These results highlight that sleep has the capacity for strengthening both global and local structure from the world and abstracting over multiple experiences to efficiently form internal networks of knowledge.