Samuel J. Gershman
YOU?
Author Swipe
View article: Transition from plasticity to dynamics-based value update for meta-learning
Transition from plasticity to dynamics-based value update for meta-learning Open
Value computation is fundamental to the survival of animals. Classical models suggest value is stored in synaptic weight through plasticity whereas more recent theories propose that recurrent network dynamics can encode and update value in…
View article: Storing long-lived memories via molecular error correction: a minimal mathematical model of Crick’s memory switch
Storing long-lived memories via molecular error correction: a minimal mathematical model of Crick’s memory switch Open
Molecular memory storage mechanisms face two serious obstacles: they must be robust to the intrinsic noise characteristic of molecular-scale processes, and they must rely on components (like proteins) whose effective lifetimes are limited …
View article: Gradient Descent as Loss Landscape Navigation: a Normative Framework for Deriving Learning Rules
Gradient Descent as Loss Landscape Navigation: a Normative Framework for Deriving Learning Rules Open
Learning rules -- prescriptions for updating model parameters to improve performance -- are typically assumed rather than derived. Why do some learning rules work better than others, and under what assumptions can a given rule be considere…
View article: Uncertainty-driven exploration during planning.
Uncertainty-driven exploration during planning. Open
View article: Distributed range adaptation in human parietal encoding of numbers
Distributed range adaptation in human parietal encoding of numbers Open
The brain represents magnitudes through the collective activity of neural populations, whose non-monotonic tuning properties determine the nature and precision of the population neural code. Whether and how this code adapts to changes in t…
View article: Uncertainty-Driven Exploration During Planning
Uncertainty-Driven Exploration During Planning Open
In complex environments, the space of possible plans is vast. Generating a good plan therefore requires judicious selection of which parts of the plan space to mentally explore. Drawing on past studies of human exploration, we propose that…
View article: Belief Neglect: A Heuristic in Social Reasoning
Belief Neglect: A Heuristic in Social Reasoning Open
Humans are capable of sophisticated social reasoning. This capability is often attributed to a folk theory of mind, which rationally links inferences about the beliefs and desires of another person to their observed behavior through a form…
View article: Action subsampling supports policy compression in large action spaces
Action subsampling supports policy compression in large action spaces Open
Real-world decision-making often involves navigating large action spaces with state-dependent action values, taxing the limited cognitive resources at our disposal. While previous studies have explored cognitive constraints on generating a…
View article: Bayesian estimation yields anti-Weber variability
Bayesian estimation yields anti-Weber variability Open
A classic result of psychophysics is that human perceptual estimates are more variable for larger magnitudes. This “Weber behavior,” however, has typically not been the focus of the prominent Bayesian paradigm. Here, we examine the variabi…
View article: The successor representation in high-risk drinking and alcohol-related contexts
The successor representation in high-risk drinking and alcohol-related contexts Open
The successor representation (SR) has been suggested to underlie nuanced forms of habitual behavior and a reduced SR variant (redSR) produces addiction-like behavior in simulations. Neither of these strategies can be detected in paradigms …
View article: Social group discovery, structure, and stereotype updating.
Social group discovery, structure, and stereotype updating. Open
Group stereotypes are difficult to change and drive discriminatory behavior across numerous consequential contexts. Across seven experiments, we test predictions made by a domain-general structure learning model to understand how people de…
View article: The successor representation in high-risk drinking and alcohol-related contexts
The successor representation in high-risk drinking and alcohol-related contexts Open
The successor representation (SR) has been suggested to underlie nuanced forms of habitual behavior and a reduced SR variant (redSR) produces addiction-like behavior in simulations. Neither of these strategies can be detected in paradigms …
View article: Entorhinal cortex signals dimensions of past experience that can be generalised in a novel environment
Entorhinal cortex signals dimensions of past experience that can be generalised in a novel environment Open
No two situations are identical. They can be similar in some aspects but different in others. This poses a key challenge when attempting to generalise our experience from one situation to another. How do we distinguish the aspects that tra…
View article: Assessing Adaptive World Models in Machines with Novel Games
Assessing Adaptive World Models in Machines with Novel Games Open
Human intelligence exhibits a remarkable capacity for rapid adaptation and effective problem-solving in novel and unfamiliar contexts. We argue that this profound adaptability is fundamentally linked to the efficient construction and refin…
View article: Adaptive Social Learning using Theory of Mind
Adaptive Social Learning using Theory of Mind Open
Social learning is a powerful mechanism through which agents learn about the world from others. However, humans don't always choose to observe others, since social learning can carry time and cognitive resource costs. How do people balance…
View article: Preemptive Solving of Future Problems: Multitask Preplay in Humans and Machines
Preemptive Solving of Future Problems: Multitask Preplay in Humans and Machines Open
Humans can pursue a near-infinite variety of tasks, but typically can only pursue a small number at the same time. We hypothesize that humans leverage experience on one task to preemptively learn solutions to other tasks that were accessib…
View article: Nucleus accumbens dopamine release reflects Bayesian inference during instrumental learning
Nucleus accumbens dopamine release reflects Bayesian inference during instrumental learning Open
Dopamine release in the nucleus accumbens has been hypothesized to signal the difference between observed and predicted reward, known as reward prediction error, suggesting a biological implementation for reinforcement learning. Rigorous t…
View article: Action chunking as conditional policy compression
Action chunking as conditional policy compression Open
Many skills in our everyday lives are learned by sequencing actions towards a desired goal. The action sequence can become a ``chunk'' when individual actions are grouped together and executed as one unit, making them more efficient to sto…
View article: A signaling theory of self-handicapping
A signaling theory of self-handicapping Open
People use various strategies to bolster the perception of their competence. One strategy is self-handicapping, by which people deliberately impede their performance in order to protect or enhance perceived competence. Despite much prior r…
View article: Neural evidence that humans reuse strategies to solve new tasks
Neural evidence that humans reuse strategies to solve new tasks Open
Generalization from past experience is an important feature of intelligent systems. When faced with a new task, one efficient computational approach is to evaluate solutions to earlier tasks as candidates for reuse. Consistent with this id…
View article: Blending Complementary Memory Systems in Hybrid Quadratic-Linear Transformers
Blending Complementary Memory Systems in Hybrid Quadratic-Linear Transformers Open
We develop hybrid memory architectures for general-purpose sequence processing neural networks, that combine key-value memory using softmax attention (KV-memory) with fast weight memory through dynamic synaptic modulation (FW-memory) -- th…
View article: Neural and behavioral signatures of policy compression in cognitive control
Neural and behavioral signatures of policy compression in cognitive control Open
Making context-dependent decisions incurs cognitive costs. Cognitive control studies have investigated the nature of such costs from both computational and neural perspectives. In this paper, we offer an information-theoretic account of th…
View article: Past suicide attempt is associated with a weaker decision-making bias to actively escape from suicide-related stimuli.
Past suicide attempt is associated with a weaker decision-making bias to actively escape from suicide-related stimuli. Open
Theory and evidence suggest that people attempt suicide to escape acute distress. However, little is known about why people select suicide instead of other ways to escape (e.g., alcohol/drug use). One possibility is that suicide-related st…
View article: Time and memory costs jointly determine a speed–accuracy trade-off and set-size effects.
Time and memory costs jointly determine a speed–accuracy trade-off and set-size effects. Open
Policies, the mappings from states to actions, require memory. The amount of memory is dictated by the mutual information between states and actions or the policy complexity. High-complexity policies preserve state information and g…
View article: Hierarchical Vector Analysis of Visual Motion Perception
Hierarchical Vector Analysis of Visual Motion Perception Open
Visual scenes are often populated by densely layered and complex patterns of motion. The problem of motion parsing is to break down these patterns into simpler components that are meaningful for perception and action. Psychophysical eviden…
View article: Synthesizing world models for bilevel planning
Synthesizing world models for bilevel planning Open
Modern reinforcement learning (RL) systems have demonstrated remarkable capabilities in complex environments, such as video games. However, they still fall short of achieving human-like sample efficiency and adaptability when learning new …
View article: Probabilistic forecasting guides dynamic decisions
Probabilistic forecasting guides dynamic decisions Open
Real-world assets---such as projects, jobs, skills, or relationships---change endogenously with continued investment, often improving at variable rates. These structural dynamics pose a challenge: How do we decide what asset to invest in a…
View article: General Intelligence Requires Reward-based Pretraining
General Intelligence Requires Reward-based Pretraining Open
Large Language Models (LLMs) have demonstrated impressive real-world utility, exemplifying artificial useful intelligence (AUI). However, their ability to reason adaptively and robustly -- the hallmarks of artificial general intelligence (…
View article: The Value of Non-Instrumental Information in Anxiety: Insights from a Resource-Rational Model of Planning
The Value of Non-Instrumental Information in Anxiety: Insights from a Resource-Rational Model of Planning Open
Anxiety is intimately related to the desire for information and, under some accounts, thought to arise from the intolerance of uncertainty. Here, we seek to test this hypothesis by studying the relationship between trait anxiety and the wi…
View article: Adaptive Social Learning using Theory of Mind
Adaptive Social Learning using Theory of Mind Open
Social learning is a powerful mechanism through which agents learn about the world from others. However, humans don’t always choose to observe others, since social learning can carry time and cognitive resource costs. How do people balance…