The Council of Writing Program Administrators has made available a very useful Research Bibliography on the Uses and Limitations of Automated Writing Evaluation Software (PDF). This is part of a set of WPA-ComPile Research Bibliographies. There are paragraph long summaries of the articles that are quite useful.
What seems to be missing is an ethical discussion of automated evaluation. Do we need to tell people if we use automated evaluation? Writing for someone feels like a very personal act (even in a large class). What are the expectations of writers that their writing would be read?