November 18, 2009
Presenter: Robb Ruttledge
Respondent: Sally Newman
How we learn about rewards
The activity of dopamine neurons in our brains are thought to encode
an error signal, the difference between the rewards we receive and the
rewards we expected to receive. This signal is thought to be used to
learn the value of options, values that then guide decision making. In
this session of the Forum, I discussed how scientific hypotheses are
tested and talked about two attempts I made to test this dopamine
hypothesis. First, I measured brain activity using functional MRI as
people won and lost money in a gambling task. I used an axiomatic
model derived from economic theory to try to falsify the hypothesis
for my brain data. It turned out that the signal in the striatum,
where the dopamine neurons send connections, satisfies the model and
can be an error signal. Second, Parkinson's disease affects the
dopamine system and patients take dopamine drugs to treat their
symptoms. I tested whether these dopamine drugs affect learning when
Parkinson's patients make choices in a learning task, as you would
expect if dopamine neurons participate in learning. The dopamine drugs
did in fact affect learning exactly as the theory predicts. Both of
these results support the hypothesis that dopamine neurons encode some
kind of error signal.