[Probcogsci] Mike Mozer's third and final talk, Yu Lab Meeting, Friday 12-1, Natcomp Lab

[Angela Yu]

In the third and final installment of Mike's talk, Mike will finish
discussing the experiment that Matt Jones, Tim Curran, and he ran, which
obtains response times and ERP measures from subjects performing a
sequence of simple two choice tasks (i.e., stimulus X -> response 1,
stimulus Y -> response 2).  The experiment has three punch lines.
First, we have the strongest evidence yet that predictions about trial
N+1 depend on joint predictions about the response identity and whether
the stimulus will repeat from the previous trial, consistent with the
extension of Angela's model which we called DBM2.  Second, counter to
everyone's theories, predictions seem to be updated via error
correction, not via recency-weighted averaging.  And third, counter to
everyones' theories, recency effects decay according to a power
function, not an exponential.  These last two findings suggest that
sequential effects are manifestations of the mechanisms of human
learning and memory, and are not mere priming effects.

The talk is in the Natcomp Lab seminar room, on the second floor of  
SSRB.  If you don't have a personal code to get in, call the natcomp  
lab phone number, posted on the code-locked door next to the ladies'  
room on the 2nd floor.

See you,
Angela

---------------------------------------------------------
Angela J. Yu
Assistant Professor
Department of Cognitive Science
UCSD, Mail Code 0515
9500 Gilman Drive
La Jolla, CA 92093-0515

Email: ajyu at cogsci.ucsd.edu
Phone: 858-822-3317
Fax: 858-534-1128
Website: www.cogsci.ucsd.edu/~ajyu
---------------------------------------------------------