01/24 Hod Lipson, Cornell University - Abstract
01/31 Peter Dayan, Gatsby Computational Neuroscience Unit, University College London - Abstract
02/14 Ken McRae, University of Western Ontario - Abstract
03/21 Lawrence Barsalou, Emory University - Abstract
04/04 Susan Goldin-Meadow, University of Chicago - Abstract
04/11 Jordan Green, University of Nebraska, Lincoln - Abstract
04/18 John O'Doherty, CalTech - Abstract
04/25 Russell Epstein, University of Pennsylvania - Abstract

Abstract

1/24: Hod Lipson, Cornell University
Title: Self-Reflective Machines
Abstract: One of the most unique and challenging aspects of intelligent living systems is their ability to self-reflect: To reconstruct models of their own morphology and of their own behavior, then use those models to adapt to new circumstances. Processes such as self-reflection play a key role in accelerating adaptation by reducing costs of physical experimentation. Similarly, the ability of a machine to observe and reconstruct models of the morphology and behavior of other machines is key to effective cooperation and competition. This talk will demonstrate a number of experiments in self reflecting robotic system, and argue that reflective processes are essential in achieving meta-cognitive capacities.

1/31: Peter Dayan, Gatsby Computational Neuroscience Unit, University College London
Title: Computational Psychiatry: When Good Decisions Go Bad
Abstract: Substantial efforts across the fields of statistics, operations research, economics, computer science and control theory have provided us with a psychologically- and neurobiologically-grounded account of how humans and other animals learn to predict rewards and punishments, and choose actions to maximize the former and minimize the latter. It becomes an obvious idea to try and relate disruptions of these models to the discontents of decision-making, as seen in neurological and psychiatric disease. I will describe the reinforcement learning model of neural decision making, together with our early attempts to look at aspects of depression through the lenses of: (a) an infelicitous prior distribution over decision-making environments which indicates their lack of controllability; and (b) the failure of a serotonergically-mediated crutch which normally inhibits potentially unfortunate choices. This is joint work with Quentin Huys.

2/14: Ken McRae, University of Western Ontario
Title: CANCELED
Abstract:

3/21: Lawrence Barsalou, Emory University
Title: Grounding Knowledge in the Brain’s Modal Systems
Abstract: The human conceptual system contains categorical knowledge that supports online processing (perception, categorization, inference, action) and offline processing (memory, language, thought). Semantic memory, the dominant theory of the conceptual system, typically portrays it as modular and amodal. According to this approach, amodal symbols represent category knowledge in a modular system, separate from the brain’s modal systems for perception, action, and internal states (e.g., interoception, introspection, emotion). Alternatively, the conceptual system can be viewed as non-modular and modal, sharing representational mechanisms with the brain’s modal systems. On a given occasion, multimodal information about a category's members is reenacted (simulated) across relevant modalities to represent it conceptually. Additionally, the conceptual system can be viewed as emergent, situated, and dynamical. Misperceptions of this approach include viewing it as non-nativist, non-symbolic, and completely dependent on sensory-motor experience. These theoretical issues and related empirical evidence, both behavioral and neural, will be reviewed, while also addressing the representation of abstract concepts and the implementation of symbolic operations.

4/4: Susan Goldin-Meadow, University of Chicago
Title: How our hands help us think
Abstract: When people talk, they gesture. We now know that these gestures are associated with learning. They can index moments of cognitive instability and reflect thoughts not yet found in speech. What I hope to do in this talk is raise the possibility that gesture might do more than just reflect learning -- it might be involved in the learning process itself. I consider two non-mutually exclusive possibilities. First, gesture could play a role in the learning process by displaying, for all to see, the learner's newest, and perhaps undigested, thoughts. Parents, teachers, and peers would then have the opportunity to react to those unspoken thoughts and provide the learner with the input necessary for future steps. Second, gesture could play a role in the learning process more directly by providing another representational format, one that would allow the learner to explore, perhaps with less effort, ideas that may be difficult to think through in a verbal format. Thus gesture has the potential to contribute to cognitive change, directly by influencing the learner and indirectly by influencing the learning environment.

4/11: Jordan Green, University of Nebraska, Lincoln
Title: Developing Speech: A motor-centric interactionist approach.
Abstract: TBA

4/18: John O'Doherty, CalTech
Title: Delineating the neural circuits underlying goal-directed and habitual behavioral control in the human brain.
Abstract: In this talk I will present evidence from a combination of behavioral and fMRI measures supporting the existence of two distinct mechanisms underpinning the selection of actions in the human brain: a flexible goal-directed system in which actions are selected with reference to the incentive value of associated goal-states and the causal relationship between actions and their outcomes, and a more reflexive habitual system in which responses are selected only by antecedent stimuli without any consideration of the outcome. In common with findings from the rodent brain, these systems seem to depend on distinct cortical and striatal circuits involving the ventromedial prefrontal cortex and dorsomedial striatum in the goal-directed case, and the posterior lateral striatum in the habitual case. A better understanding of how these systems interact in order to control behavior could yield new approaches to the study of psychiatric disorders involving compulsive habit-like behaviors such as addiction or OCD.

4/25: Russell Epstein, University of Pennsylvania
Title: From scenes to cognitive maps: place recognition in the human brain
Abstract: Place recognition is the ability to use visual information to determine where one is in the world. It involves at least two stages: (1) perception and identification of the local scene, (2) recovery of the coordinates of the local scene within the extended spatial environment. Neuroimaging and neuropsychological findings implicate three brain regions in place recognition: the parahippocampal place area (PPA), retrosplenial complex (RSC), and hippocampus. The PPA and RSC respond strongly in fMRI studies when subjects view "places" (e.g. landscapes, street scenes, rooms) but weakly when they view faces or single decontextualized objects. The PPA appears to encode whole-scene visual quantities such as layout and may work in concert with other brain regions that analyze object-based information during scene recognition. RSC, on the other hand, appears to have a distinct but complementary function: situation of the local scene relative to elements of the world that are not currently visible. Finally, recent data from our laboratory implicate the hippocampus in the coding of spatial relationships between familiar landmarks, suggesting that it may support a map-like representation in which locations are coded in terms of their allocentric spatial coordinates.