[CTN] title and abstract for Dr. Maia's CTN seminar, July 26

[Matthijs van der Meer]

Dr. Tiago Maia
Assistant Professor of Clinical Neurobiology
Department of Psychiatry, Columbia University
and New York State Psychiatric Institute

http://childpsych.columbia.edu/brainimaging/CV_maia.html

Title: Dopamine, norepinephrine, and behavior: computational and
empirical investigations

The last decade and a half has brought remarkable advances to our
understanding of the computational functions of the dopaminergic and
noradrenergic systems. I have focused on leveraging this understanding
to explain complex behavioral findings that have eluded classical
psychological and neurobiological explanations. Along the way, I found,
almost by serendipity, that this focus on developing detailed
neurocomputational theories of behavior often leads to novel neural
predictions that can be tested (and confirmed) empirically. I will
illustrate this general approach using two examples. In the first, I
used a standard actor-critic model to explain a wide range of behavioral
findings in avoidance learning (the process of learning to avoid
aversive outcomes). The decision to use the actor-critic was based on
the substantial evidence for a role of prediction errors (signaled by
dopamine) in reinforcement learning in the brain and the mounting
evidence for an actor-critic-like organization of the striatum, but the
model was originally aimed only at explaining behavior, not neural
findings. Nonetheless, the model made novel predictions about dopamine
release during avoidance learning, and we have recently confirmed those
predictions experimentally. In the second example, I used a model of the
role of norepinephrine in modulating the gain of target neurons to show
that low norepinephrine explains a broad range of behavioral deficits in
attention-deficit/hyperactivity disorder (ADHD). This model was also
originally aimed only at explaining behavior. Nonetheless, it made novel
predictions about abnormalities in brain connectivity in ADHD and the
amelioration of such abnormalities with the medications commonly used to
treat ADHD. We have recently confirmed those predictions using fMRI with
patients with ADHD on and off medications and a healthy control group.
Taken together, these findings illustrate the ability of computational
accounts to integrate complex neural and behavioral findings, showing
that computational theories originally developed at one level often lead
to novel explanations and novel, accurate predictions at the other level.