MONDAY, August 29
Nora Newcombe, Temple University
Charting our way in space and time
Episodic memory and navigation are two adaptive functions that share some neural substrates and that develop in humans from fragile beginnings in the first two years to mature competencies (with individual differences) by early adolescence. But how are they related, both in development and maturity? In this talk, I will discuss the two developmental trajectories, and some possible answers to that question.
Nicholas Turk-Browne, Yale University
Learning and memory in early development
Cognitive neuroscience provides a rich account of how different brain systems give rise to diverse forms of learning and memory. However, these theories are based on adult data and largely neglect early development, perhaps the greatest period of learning in life. A key challenge for studying this age range is the limited set of behavioural measures available, especially in infants. Neural measures from EEG and fNIRS provide a window into the infant mind, but have coarse spatial resolution and lack access to deep-brain structures such as the hippocampus that are critical for adult learning and memory. I will present our recent efforts to adapt fMRI, which addresses some of these limitations, for studying human infants during cognitive tasks. I will focus on one line of studies, concerning a mystery about how the brain supports statistical learning. We have shown in adults that the hippocampus is important for statistical learning, and statistical learning is a core building block of the infant mind, yet the hippocampal system is often assumed to be immature in infants (e.g., to explain infantile amnesia). This and our other fMRI studies in awake infants aim to advance understanding of the functions and plasticity of the youngest minds and brains.
Aaron Peikert, Max Planck Institute for Human Development
Reproducible Research: A talk on how to do the same thing more than once
Computational reproducibility is the ability to obtain identical results from the same data with the same computer code.
The high rate of irreproducible research limits the reach of results and decreases the efficiency of researchers.
Reproducible research is a building block for transparent and cumulative science because it enables the originator and other researchers, on other computers and later in time, to reproduce and thus understand how results came about.
In this talk, I present a conceptual analysis of what it takes to work reproducible, provide hints on how to work reproducible in practice, and what problems researchers are likely to encounter.
TUESDAY, August 30
Rani Moran, University College London
Cognitive maps guide credit assignment
An extensive reinforcement learning literature has paid considerable attention to how cognitive maps serve choice by allowing agents to prospectively compute values of task-states during planning. But whether and how knowledge of a cognitive map contributes to the way agents assign credit during reward-feedback (i.e., how they revaluate actions and task-states following the receipt of an outcome) remains underexplored. In many real-life situations attributing reward-outcomes to underlying task-states and actions is difficult and hence credit assignment is challenging. Importantly, knowledge stored in a cognitive map can aid reward-attribution and hence, guide credit assignment. I will present evidence that in situations entailing state-uncertainty (i.e., important aspects of the decision environment are latent) or exuberant reward-feedback (which only partially relates to one’s actions), humans exploit knowledge stored in a cognitive map to guide credit assignment. I will discuss how these findings broaden our understanding regarding the scope of functions cognitive maps serve in adaptation and I will argue that they call for a more nuanced conceptualization of model-based and model-free control processes and their interactions.
Michael J. Frank, Brown University
Clustering and generalisation of abstract structures in reinforcement learning
Humans, and even infants, are remarkably adept at generalizing knowledge between experiences in a way that can be difficult for computers. Previous computational models and data suggest that rather than learning about each individual context, humans build latent abstract structures and learn to link these structures to arbitrary contexts, facilitating generalisation. In these models, task structures that are more popular across contexts are more likely to be revisited in new contexts. However, these models predict that structures are either re-used as a whole or created from scratch, prohibiting the ability to generalise constituent parts of learned structures. This contrasts with ecological settings, where some aspects of task structure, such as the transition function, will be shared between contexts separately from other aspects, such as the reward function. Moreover, in many situations people can transfer structures that they have learned to entirely new situations, by analogy, even when surface aspects of the transition and reward functions change. I will present novel computational models that address how agents and humans can learn and generalise such abstract and compositionl structure, supported by empirical data.
WEDNESDAY, August 31
Tomás Ryan, Trinity College Dublin
Forgetting as a form of adaptive plasticity
Forgetting' generally refers to the loss of previously formed memories. Although multiple forms of forgetting have been characterised, ranging from natural "every day" forgetting to unnatural pathological forgetting, a formal scientific framework with which to explain and investigate the neuroscience of forgetting is lacking. This may be because forgetting has been regarded as a defect of the brain, and it has been assumed to have many diverse and incidental causes. However, contemporary research is challenging this paradigm and an alternative perspective has emerged where forgetting may be viewed as an adaptive feature of the brain that allows an organism to respond optimally to its environment.
In my lecture, I will summarize behavioural studies that imply that forgetting serves adaptive functions to allow organisms to generalise and abstract from initial experiences. I will discuss a growing body of findings that demonstrate that forgetting is based on active neurobiological mechanisms that respond to environmental experience. I will introduce "engram cell labelling" methodologies, which allows us to genetically label, observe, and manipulate the specific ensembles of neurons that encode particular memories in the rodent brain. I will then describe our recent research on innate and acquired forms or long-term forgetting in the mouse, by focussing on infantile amnesia during development on one hand, and natural forgetting in adults on the other. I will show how many forms of forgetting are in fact reversible, and that the core information endures within the brain's engrams. I will present a formal model of natural forgetting, based on our empirical data, that will inform future experimental investigations. Finally, I will outline a novel framework that considers both natural and unnatural forgetting to be predictive processes that involve the interaction of a subject's priors with perceptual experience.
Nico Dosenbach, University of Washington at St. Louis
Development and plasticity of human functional brain networks
In contrast to Brain-Wide Association Studies (BWAS), Precision Functional Mapping (PFM), a deep phenotyping approach to human neuroimaging, can leverage extremely small samples (down to n = 1). This lecture reviews the original PFM studies and current ones that have generated new functional neuroanatomical insights. The focus is on investigations of the impact of sensory and/or motor deprivation, or disuse, and brain lesions early in development on brain function. It covers PFM of brain plasticity, focusing on experiments that tracked casting of the dominant upper extremity with daily resting-state functional MRI scans. Mechanisms that shape brain circuits during early development may persist into adulthood, helping to maintain the organisation of disused circuits. Focal cortical injuries sustained in early childhood can be compensated for more quickly and more completely than those sustained later in life. However, the mechanisms underlying cortical plasticity are only beginning to be understood. We have done an exhaustive investigation of one of our patients (PS1; an adolescent male), who sustained large, bilateral perinatal strokes in 1999, but nevertheless had typical neurodevelopment and his injuries went unnoticed until he was 13 years old.
THURSDAY, September 1
Catherine Hartley, New York University
Developmental tuning of action selection
Throughout our lives, we rapidly acquire knowledge through experience. This knowledge is structured — it reflects regularities in our environments such as sequential relations between events, contingencies between actions and outcomes, and similarities across contexts. Across development, we exploit this structure to support the flexible pursuit of valued outcomes. In this talk, I will present studies examining at the cognitive, neural, and computational levels how the learning, memory, and decision-making processes that support or constrain adaptive behavioural flexibility change over the course of development from childhood to adulthood. I will show that development confers marked changes in the cognitive representations and computations engaged to evaluate and select actions. I will discuss how these changes may optimize behaviour for an individual’s developmental stage and how neurocognitive development may influence individual vulnerability and resilience to different forms of psychopathology.
Tobias Hauser, University College London
Building the tools for Developmental Computational Psychiatry
Most psychiatric illnesses emerge during childhood and adolescence. This is also the period in life, during which the brain and cognition change most dramatically. However, we still know little about whether and how aberrant brain development may lead to mental health problems. Developmental computational psychiatry has the goal to identify and understand the neural and computational mechanisms that go awry during development and lead to mental illness. In my talk, I will outline why I believe that development is critical if we want to understand and treat psychiatric disorders in general. I will then discuss the prerequisites for successfully studying these developmental processes, and I will show how far we have come in building these tools for studying these computational mechanisms during development. Lastly, I will provide an outlook about where I envisage the field to move in the next years.