lansky@SRI-VENICE.ARPA (Amy Lansky) (03/19/87)
Anyone interested in giving a talk, please contact Amy Lansky --
LANSKY@SRI-AI.
EVALUATING "CONNECTIONIST" NETWORKS AS MODELS OF HUMAN LEARNING
Mark A. Gluck (GLUCK@SU-PSYCH)
Stanford University
11:00 AM, MONDAY, March 23
SRI International, Building E, Room EJ228
We used adaptive network (or "connectionist") theory to extend the
Rescorla-Wagner/LMS rule for associative learning to phenomena of
human learning and judgment. In three experiments, subjects learned
to categorize hypothetical patients with particular symptom patterns
as having certain diseases. When one disease is far more likely than
another, the model predicts that subjects will substantially
overestimate the diagnosticity of the more valid symptom for the Rare
disease. This illusory diagnosticity is a learned form of "base-rate
neglect" which has frequently been observed in studies of probability
judgments. The results of Experiments 1 and 2 provided support for
this prediction in contradistinction to predictions from probability
matching, exemplar retrieval, or simple prototype learning models.
Experiment 3 addressed representational issues in the design of the
network models. When patients always have four symptoms (chosen from
four opponent pairs) rather than the statistically equivalent
presence/absence of each of four symptoms, as in Experiment 1, the
network model predicts a pattern of results quite different from
Experiment 1. The results of Experiment 3 were again consistent with
the Rescorla-Wagner/LMS learning rule as embedded within an
connectionist network.
VISITORS: Please arrive 5 minutes early so that you can be escorted up
from the E-building receptionist's desk. Thanks!