Talk materials

Abstracts and suggested reading, organised by day. Entries marked TBA will be updated.

Monday 13 July

Zheng Wang

Prefrontal scaling of reward prediction error readout gates reinforcement-derived adaptive behavior in primates

Reinforcement learning (RL) driven by reward prediction errors (RPEs) enables adaptive behaviour across species. While the neural encoding of RPE signals in midbrain and striatal circuits is well documented, how prefrontal architectural differences translate reinforcement signals into species-specific behavioural flexibility remains unknown. Using cross-species RL modelling, neuroimaging, and network-based transcriptomics, this work shows that humans and macaques share molecular and neuroanatomical infrastructure for RPE computations, with convergent upregulation of monoaminergic and synaptic genes in prefrontal and subcortical regions. Despite this conservation, humans show superior reversal-learning performance, disproportionately recruiting dorsal anterior cingulate (dACC) and dorsolateral prefrontal cortex (dlPFC). The findings establish a prefrontal-dependent readout constraint in which adaptive decision-making scales with the areal extent of dACC/dlPFC engagement governing RPE-to-action translation.

Suggested reading

  • Miyamoto et al. (2026). Brain activity, disruption and connectivity comparisons identify origins of human metacognition in other primates. Nature Human Behaviour.
  • Fujimoto et al. (2026). Ventrolateral prefrontal cortex in macaques guides decisions in different learning conditions. Nature Communications.
  • Ding et al. (2026). Model-based and model-free valuation signals in the human brain vary markedly in relation to individual differences in behavioral control. Cell Reports.

Oliver Robinson

Using Reinforcement Learning Models to Understand Anxiety

I will present a series of projects exploring the role of reinforcement learning processes in the psychopathology of anxiety. I will start by describing a novel simulation based meta-analysis that we developed that suggests that, across all the work prior to the meta analysis, altered punishment learning rates are important in driving anxiety symptoms. I will then take you through more recent translational and psychopharmacological (benzo) work from the lab that suggests that in other contexts punishment sensitivity differences can drive anxiety symptoms. I will conclude by suggesting that this sort of work may help us understand how different treatments (e.g. psychological or pharmacological) might help improve symptoms via fundamentally different mechanisms.

Suggested reading

Bee Quinn

Computational modelling of reinforcement learning tasks with hBayesDM.

Computational models provide valuable insight into the parameters underlying individual differences in reinforcement learning, but how do we reliably extract these parameters from experimental task data? This workshop provides a practical introduction to Bayesian MCMC modelling using the hBayesDM library. Attendees will be instructed on how to use R for the hands-on practice fitting of computational models to experimental data, extracting reinforcement learning parameters, and interpreting the output. The workshop also covers model validation through evidence comparison, data simulation, and parameter refitting. Attendees will leave with a working pipeline for fitting and validating computational models, applicable to a range of tasks both within and beyond reinforcement learning.

Suggested reading

  • Daw, N. D. (2011). Trial-by-trial data analysis using computational models. Decision making, affect, and learning: Attention and performance XXIII, 23(1), 3-38.
  • Ahn, W. Y., Haines, N., & Zhang, L. (2017). Revealing neurocomputational mechanisms of reinforcement learning and decision-making with the hBayesDM package. Computational Psychiatry (Cambridge, Mass.), 1, 24.

Tuesday 14 July

Yixin Zhu · Peking University

Reverse-Engineering the Dark Matter of Intelligence: A Unified Framework for Modeling Minds and Probing Machines

Contemporary AI excels with large data on narrow tasks, whereas human intelligence does the reverse, generalising from sparse experience. This talk attributes the gap to the unobservable structure underlying behaviour — the “dark matter” of intelligence: functionality, physics, intentionality, causality, and utility (the FPICU framework; Zhu et al., 2020). At its centre is a unified framework for recovering such structure in both biological and artificial agents: a structured process model, grounded in cognitive theory, whose interpretable parameters are jointly constrained by theory, task, and data, then adjudicated through dissociation paradigms, model–human–baseline comparison, and cross-task generalisation. The same framework governs computational modelling of human cognition (intuitive physics, causal learning, exploratory play) and the systematic probing of large language and vision models in moral reasoning and creativity.

Suggested reading

  • The paradigm — common sense and the “dark matter” of intelligence: Zhu, Y., et al. (2020). Dark, Beyond Deep: A Paradigm Shift to Cognitive AI with Humanlike Common Sense. Engineering, 6(3), 310–345. · Lake, B. M., Ullman, T. D., Tenenbaum, J. B., & Gershman, S. J. (2017). Building Machines That Learn and Think Like People. Behavioral and Brain Sciences, 40, e253.
  • Probabilistic & generative models of cognition: Tenenbaum, Kemp, Griffiths & Goodman (2011). How to Grow a Mind. Science, 331(6022), 1279–1285. · Griffiths, Chater, Kemp, Perfors & Tenenbaum (2010). Probabilistic Models of Cognition. Trends in Cognitive Sciences, 14(8), 357–364.
  • Intuitive physics: Battaglia, Hamrick & Tenenbaum (2013). Simulation as an Engine of Physical Scene Understanding. PNAS, 110(45), 18327–18332. · Kubricht, Holyoak & Lu (2017). Intuitive Physics: Current Research and Controversies. Trends in Cognitive Sciences, 21(10), 749–759.
  • Causal learning & exploratory play: Gopnik et al. (2004). A Theory of Causal Learning in Children. Psychological Review, 111(1), 3–32. · Gopnik & Wellman (2012). Reconstructing Constructivism. Psychological Bulletin, 138(6), 1085–1108.
  • Intentionality & theory of mind: Baker, Jara-Ettinger, Saxe & Tenenbaum (2017). Rational Quantitative Attribution of Beliefs, Desires and Percepts in Human Mentalizing. Nature Human Behaviour, 1, 0064. · Jara-Ettinger (2019). Theory of Mind as Inverse Reinforcement Learning. Current Opinion in Behavioral Sciences, 29, 105–110.
  • Probing & evaluating large models — the “psychology of AI”: Binz & Schulz (2023). Using Cognitive Psychology to Understand GPT-3. PNAS, 120(6), e2218523120. · Shiffrin & Mitchell (2023). Probing the Psychology of AI Models. PNAS, 120(10), e2300963120. · Mitchell & Krakauer (2023). The Debate Over Understanding in AI’s Large Language Models. PNAS, 120(13), e2215907120. · Demszky et al. (2023). Using Large Language Models in Psychology. Nature Reviews Psychology, 2, 688–701.

Ayelet Landau

Rhythmic attention negotiates competition along the visual hierarchy

Navigating the environment involves engaging with multiple objects, each activating specific neuronal populations that compete when objects appear together. Classical attention theories suggest selection biases one population over another, while recent research shows perception fluctuates over time: a single object processed over time shows a ~8 Hz fluctuation, whereas attention distributed over two objects shows a 4 Hz fluctuation, possibly reflecting division of the 8 Hz rhythm between competing objects. This talk explores these rhythmic phenomena — attentional sampling — across the visual hierarchy, arguing that sampling is a selection mechanism that negotiates neuronal competition, manifesting as early as eye channels and extending to complex features higher in the hierarchy and potentially beyond the visual modality.

Suggested reading

  • Re, D., Kusnir, F. & Landau, A. N. (2025). Attentional sampling resolves competition along the visual hierarchy. Trends in Cognitive Sciences. doi:10.1016/j.tics.2025.06.004
  • Kusnir, F. & Landau, A. N. (2025). Temporality and the brain: the long and winding emergence of time in cognitive neuroscience. Human Arenas. doi:10.1007/s42087-025-00497-8
  • Re, D., Karvat, G. & Landau, A. N. (2023). Attentional sampling between eye-channels. Journal of Cognitive Neuroscience, 35(8), 1350–1360. doi:10.1162/jocn_a_02018
  • Landau, A. N. (2018). Neuroscience: a neural mechanism for rhythmic sampling. Current Biology, 28(15), 830–832. doi:10.1016/j.cub.2018.05.081

Zilong Ji

Computational modelling of the head direction system in the brain

This hands-on workshop introduces the computational principles underlying the head direction system in the brain. We will begin with a brief overview of head direction cells and the upstream angular velocity system, before focusing on computational models of the head direction circuit, particularly continuous ring attractor networks. No strong mathematical background is required, although participants should have basic Python programming experience. Throughout the session, we will build a ring attractor model from scratch, run simulations together, and visualise the tuning properties of individual head direction cells. By the end of the workshop, participants will gain an intuitive understanding of how recurrent neural circuits can generate stable representations of head direction. I have attached three papers for you to read. Please start with Taube_2007_ANRV, an overview of the head direction system. Then read Zhang_1996_JNS, focusing on the key ideas and skipping the mathematical details. If you have more time, read Taube_1990_JNS, which first reported head direction cells..

Suggested reading

  • Taube, J. S. (2007). The head direction signal: origins and sensory-motor integration. Annual Review of Neuroscience, 30, 181–207. doi:10.1146/annurev.neuro.29.051605.112854
  • Zhang, K. (1996). Representation of spatial orientation by the intrinsic dynamics of the head-direction cell ensemble: a theory. Journal of Neuroscience, 16(6), 2112–2126. doi:10.1523/JNEUROSCI.16-06-02112.1996
  • Taube, J. S., Muller, R. U. & Ranck, J. B. Jr. (1990). Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. Journal of Neuroscience, 10(2), 420–435. doi:10.1523/JNEUROSCI.10-02-00420.1990

Wednesday 15 July

Jian Li

The perception, implementation and alternation of human altruism

Altruism refers to concern for the welfare of others, often at a cost to the altruistic actor. Despite intensive research, how altruistic reputation is inferred and how altruistic preference is modulated by decision context remain fiercely debated. Across a series of experiments, this work shows a combination of factors influencing the inference and implementation of altruistic preference, with dedicated neural networks convening social and private information to compute a value signal behaviourally manifested as altruistic choice. Causal evidence using pharmacological and TMS methods demonstrates the elimination or reversal of the perception and manifestation of altruism by targeting these dedicated neural-network activities.

Suggested reading

Maarten Speekenbrink

The versatility of hidden Markov modelling for psychology

Hidden Markov models provide a flexible framework to analyse sequences of observables (e.g. behaviours, decisions, events, or answers to questionnaire items) by assuming these results from an underlying sequence of latent unobservable states. As such, these models are highly relevant to psychology, where we are often interested in uncovering the mental processes that lead to observable behaviour. Yet, application of hidden Markov models in psychology remains relatively rare. Via three applications, I will illustrate the versatility and usefulness of hidden Markov models in psychology, focusing on the analysis of eye movements in reading, identifying mental health states in a diverse psychiatric population, and modelling interactions in repeated economic games.

Suggested reading

Shuang Tian

Using Deep Neural Networks to Probe Neural Representations

Deep neural networks (DNNs) have become powerful computational models for studying neural representations in the human brain, helping us understand how the brain encodes information about the world. This workshop introduces several approaches for integrating DNNs with neural data and discusses when each approach should be used, what neuroscientific questions it can answer, and how the results should be interpreted. Participants will then be guided through a hands-on ROI-based fMRI analysis using CLIP, applying the complete workflow from research question to model selection, analysis, and interpretation. Participants will leave with a practical workflow that can be readily adapted to different types of neural data.

Suggested reading

  • Combined evidence from artificial neural networks and human brain-lesion models reveals that language modulates vision in human perception. doi:10.1038/s41562-025-02357-5

Thursday 16 July

Lusha Zhu

Representation of and decision-making on social networks

Social networks shape our beliefs and choices by constraining what information we receive and from whom. Yet the mechanism by which the human brain interacts with networked environments remains unclear. Two computational challenges stand out when we try to interact with interconnected peers. First, information flowing along network connections is typically statistically interdependent and varies in its informativeness. How, then, does the brain effectively integrate network-derived information? Second, individuals can hardly take into account the topological structure of the entire network when interacting with it. So which social connections are considered, which are ignored, and how do the streamlined mental representation shape our perception and navigation of the social world? In this talk, I will present a series of recent work that uses human behavioral experiments, computational modeling, fMRI, and graph neural network(GNN) to investigate how people learn from and mentally represent social networks. Our finding provide a unified account of a range of seemingly disparate biases in social perception and decision-making, shedding light on the cognitive roots of important societal conundrums such as biased social sensing and misinformation propagation.

  • Jiang, Y., Mi, Q. & Zhu, L. Neurocomputational mechanism of real-time distributed learning on social networks. Nature Neuroscience (2023).
  • Ho, M. K. et al. People construct simplified mental representations to plan. Nature https://doi.org/10.1038/s41586-022-04743-9 (2022) doi:10.1038/s41586-022-04743-9.
  • Tomov, M. S., Yagati, S., Kumar, A., Yang, W. & Gershman, S. J. Discovery of hierarchical representations for efficient planning. PLoS Comput Biol 16, e1007594 (2020).

Marco Wittmann

Neural mechanisms of social structure learning and their behavioural consequences

Navigating social environments is a fundamental challenge for the brain. While social information is often represented at an individual level, learning the structure in which individuals interact and relate to each other is equally important for adaptive social decision-making. This talk presents evidence across behavioural, fMRI, and ongoing MEG and transcranial-ultrasound studies suggesting that medial prefrontal cortex represents the combinatorial possibilities for social interaction in a compressed format resembling basis functions observed in other domains. These basis functions capture group-level structure and predict specific choice patterns, including counterintuitive effects where decision-irrelevant players influence behaviour. Initial MEG results suggest such structure representations emerge rapidly after stimulus presentation, and preliminary non-invasive deep-brain-stimulation evidence is consistent with a role of prefrontal cortex in embedding social information into its structural context.

Suggested reading

  • Wittmann, M. K., Lin, Y., Pan, D., Braun, M. N., Dickson, C., Spiering, L., et al. (2025). Basis functions for complex social decisions in dorsomedial frontal cortex. Nature. doi:10.1038/s41586-025-08705-9
  • Dayan, P. & Hinton, G. E. Feudal Reinforcement Learning.

Philippa Johnson

Capturing within-session dynamics with hidden Markov models

Suggested reading

  • Ashwood, Z. C., Roy, N. A., Stone, I. R., International Brain Laboratory, Urai, A. E., Churchland, A. K., Pouget, A. & Pillow, J. W. (2022). Mice alternate between discrete strategies during perceptual decision-making. Nature Neuroscience, 25(2), 201–212. doi:10.1038/s41593-021-01007-z

Yuki Ma (non-mandatory workshop)

TBA — materials to follow.

Friday 17 July

Yanchao Bi

Weaving vision and language into semantic memory

The human brain stores a tremendous amount of knowledge about the world, the foundation of object recognition, language, thought, and reasoning. What are the neural codes of semantic-knowledge representation? Is the knowledge that “roses are red” simply the memory trace of perceiving the colour of roses, stored within colour-sensitive neural systems? What about knowledge not directly perceived by the senses, such as “freedom” or “rationality”? This talk presents recent studies addressing these issues with various special populations and cross-species comparisons, discussing them in the context of classical findings and neural frameworks of semantic memory.

Suggested reading

Megan Peters

On using computational and cognitive neuroscience tools to reverse engineer conscious subjective experience

How does your brain create conscious subjective experience? Everybody has a theory. Some people have very “mature” theories. Some people argue that their theory is the one that has the most empirical support. But we definitely don’t have a “leading theory” of consciousness, and figuring out which theories are even on the right track requires us to develop new analytical, experimental, and theoretical approaches. In this talk, I’ll discuss how we’re doing just that, from the context of two branches of two ‘adversarial collaborations’ on consciousness which I co-lead, where we use and modify tools from psychophysics to machine learning to neuroimaging. Has one theory lost one of these contests yet? You’ll have to come to the talk to find out.

  • Tian, K., Maniscalco, B., Epstein, M., Shen, A., Castaneda, O. G., Arzu, G., ... & Denison, R. (2024). Attention robustly dissociates objective performance and subjective visibility reports. Journal of Vision, 24(10), 408-408.
  • Maniscalco, B., Graham Castaneda, O., Odegaard, B., Morales, J., Rajananda, S., Denison, R. N., & Peters, M. A. K. (2026). The relative psychometric function: A general analysis framework for relating psychological processes. Psychological Review. Advance online publication. https://dx.doi.org/10.1037/rev0000625
  • Peters, M. A. (2022). Towards characterizing the canonical computations generating phenomenal experience. Neuroscience & Biobehavioral Reviews, 142, 104903.