Connectionist Models of Developmental Disorders:
A Critical Appraisal
A Critical Appraisal
Christopher M. Conway, Michelle R. Ellefson,
Rick Dale,
and Morten H. Christiansen
Abstract
This chapter critically reviews connectionist models of developmental cognitive disorders
that are implemented by alterations to architecture, training dynamics, or input/output
representations. Models are assessed in terms of four criteria: data contact, task
veridicality, input representativeness, and deficit implementation. Data contact concerns
the degree at which relevant empirical data is captured. Task veridicality assesses the
extent to which a model's task corresponds to that of humans. Input representativeness
refers to the degree to which the model's input matches the input available to a person.
Finally, deficit implementation assesses the appropriateness of the network manipulations
used to simulate a developmental deficit. We conclude that although some shortcomings
exist, connectionist models show much potential for elucidating the underlying
mechanisms of developmental cognitive disorders.
Please email suggestions/errors to mhc27@cornell.edu |