I am a cognitive neuroscientist, with over three decades of research experience in computational and mathematical modeling as well as empirical studies of human brain function and behavior, focusing on the neural mechanisms responsible for cognitive control and human intelligence, and how our growing understanding of these can be brought to bear in the design of computational architectures with more human-like cognitive capabilities. My work lies at the points of contact between neuroscience, psychology, computer science, mathematics, behavioral economics and psychiatry, and involves collaborations with investigators in each of these fields. I also have considerable experience in the coordination and administration of scientific research, as one of the two founding Co-Directors of the Princeton Neuroscience Institute, leading multi-institutional projects (NIMH Conte Center; Templeton Center Grant; NSF Convergence Accelerator grant; PNI-Intel Labs collaboration), and a number of open source software development projects.
[Note: links below are to reprints and numbers correspond to entries under Publications in CV]
Some of the contributions that have emerged from the theoretical work of my colleagues and I are: the first computationally-explicit models of how cognitive control may be implemented in the brain4 and the role of prefrontal cortex in control;76 the role of dopaminergic function in the gating and updating of information in prefrontal cortex,5,56 noradrenergic regulation of the explore/exploit tradeoff,55,103,127 and the interaction of these modulatory systems in adaptive regulation of exploration in reinforcement learning;111 how these mechanisms may be disturbed in psychiatric disorders; 8,53,110 the role of anterior cingulate cortex in performance monitoring61,102,182 and the optimal allocation of control;175,212 mathematical analysis of optimal control of simple decision making processes;118,152 normative approaches to understanding capacity constraints associated with working memory68,143 and cognitive control;181,211,249,X and how the brain regulates the balance between flexible control-dependent and efficient automatic processing.218,X Increasingly, our work has come to focus on how these mechanisms contribute to higher cognitive functions and human intelligence, such as the control of memory, planning, and abstract reasoning221,238,239,253 including ways in which the human brain achieves the flexibility of symbolic forms of computation115,173,252,258 while preserving the efficiency of computation in neural networks, and how this can be used to inform research in machine learning and artificial intelligence.256,264, 265, X
Empirical and methodological contributions
The theoretical work summarized above has served as the foundation for a number of empirical and methodological contributions. Empirical contributions include: the first demonstrations in humans of sustained activity in PFC associated with working memory performance;15,34 the distinction between the roles of dorsolateral prefrontal cortex (in the regulatory functions of control) and anterior cingulate cortex (monitoring and evaluative functions of control;40,51,58,92 and the role of the locus coeruleus / norepinephrine system in regulating the explore-exploit tradeoff.156,206 We have also made influential contributions to advances in quantitative methods in cognitive neuroscience, including: the introduction of cluster size correction into the analysis of fMRI data;42 the use of fMRI to directly study midbrain neuromodulatory nuclei;132 the design of systems for realtime fMRI analysis22,260 and closed-loop feedback designs;185,247 and whole brain, full correlation analysis of fMRI data178,192 and its use in realtime analysis.X Finally, I have lead or co-lead several large software development projects, including: PsyScope, the first graphical environment for the design and execution of cognitive behavioral experiments; BrainIAK (in collaboration with Intel Labs), an open-source, python-based toolbox for the implementation and optimization of advanced methods of brain image analysis; PsyNeuLink, an open-source, python-based environment for the design and exchange of computational models of brain and cognitive function; SweetPea, a framework for specifying empirical experimental designs and machine learning training environments using factorial structure, and generating maximally unbiased sampling of trials; and a standardized model description format (MDF) for expressing models of brain and cognitive function as computational graphs in machine readable form for exchange across modeling environments.