[CTN] CTN seminar: Dr. Paul Schrater, Nov 6, PAS 2464, 3:30

[Matthijs van der Meer]

Dear all,

Please join us for next Tuesday's CTN seminar (Nov 6) by Dr. Paul 
Schrater, from the University of Minnesota. Title and abstract follow below.

Time and place are the usual, 3.30pm on Tuesday in PAS 2464.

If you would like to meet with Dr. Schrater, and/or come to dinner
afterwards, please let me know.

Hope to see you all there!

- Matt



Paul Schrater
Psychology & Computer Science
University of Minnesota

http://gandalf.psych.umn.edu/users/schrater/PaulSchrater.htm

Title: Rational Preference Learning

In models of decision-making, human preferences for choice options are
often represented by utilities - numbers assigned to each option that
express its desirability. Due to a dearth of useful alternatives,
utility models remain in widespread use despite considerable empirical
evidence revealing their deficiencies. In particular, preferences are
contingent on context and can even exhibit reversals between contexts.

We have developed a new theory of preference formation based on rational 
optimal learning of values. Our critical assumptions are
that feedback about an option's worth is relative to the comparison
items, and that the information we receive about the quality of
options is limited. In our theory, preferences are equivalent to a 
probabilistic inference about which item is the best choice of a set.

Surprisingly, we show that this representation is equivalent to standard 
utility under conditions where quality feedback is consistent. However, 
given inconsistent quality feedback, it provides explanations of many 
previously intractable violations of rational choice behavior including 
preference reversals and context effects. In short, people may not be 
consistent, optimal decision makers relative to an agent that knows the 
value of everything in the world. But they might be optimal learners of 
value in a world where value information is often contextual, 
inconsistent and coarse-grained.