Connectionist modeling has proven very successful in describing aspects of both typical and atypical development. Quite a few models have also been created to explain features of autism spectrum conditions. Yet, to date there has been no systematic review of connectionist work in autism. I surveyed all connectionist modeling work in autism spectrum conditions published in English-language, peer-reviewed venues. I located 37 publications, which represented 26 distinct studies, and also provided formal study quality assessment. The number of studies per year increased with time, demonstrating the growing popularity of the topic. Study quality also increased with time, though some of this effect was due to models becoming larger with more computing power available. The most frequently investigated areas of atypical performance in autism were less related to core cognitive features and more related to phenomena easy to model with connectionist approaches: atypical neural growth patterns, aspects of attention and perception, and to an extent, functional connectivity. Most, but not all publications passed a minimum requirement of neurological plausibility, but very few studies compared their results to human experiments. I provide guidelines to researchers interested in producing their own models.
I am in the process of writing this up as an article with Georg Dorffner.
This study was funded by the University of Iowa Presidential Graduate Research Fellowship to BP.