| |
Abstract:
Abstract: In everyday tasks, selecting actions in the proper
sequence requires a continuously updated representation of temporal
context. Traditional models address this problem by positing
hierarchies of processing units that mirror the hierarchical
structure of each naturalistic task. This approach, however, leads
to a number of difficulties, including a reliance on overly rigid
sequencing mechanisms, a limited ability to account for context
sensitivity in behavior, and a failure to address learning. We
consider an alternative framework in which representations of
temporal context are learned by recurrent connections within a
connectionist network that maps environmental inputs to actions..
Applying this approach to the specific but typical everyday task of
coffee-making, we examine a model's ability to account for several
central characteristics of normal and impaired human performance.
The model learns to deal flexibly with a complex set of sequencing
constraints, encoding contextual information at multiple
time-scales within a single, distributed internal representation.
Mildly degrading this context representation leads to errors
resembling everyday "slips of action". More severe degradation
leads to a pattern of behavior resembling that observed in action
disorganization syndrome. Analysis of the model's operation yields
novel, testable predictions relevant to both normal and apraxic
performance. Taken together, the results indicate that recurrent
connectionist models offer a useful framework for understanding
routine sequential action.
|
