Ethical Dimensions in the Use of Computational Text Analysis: Topic Modeling Autism Parenting Blogs
Computational tools can facilitate interpretive work in the qualitative social sciences by rapidly summarizing large amounts of data, including publicly available textual data. Doing so delegates to such models some measure of authority to speak on behalf of those people whose stories (i.e. whose data) are being analyzed. This ongoing project draws on sociological accounts of representation to examine ethical dimensions in the application of topic modeling to publicly available blogs written by parents of children on the autism spectrum. The project investigates: (1) the kinds of statements that topic models, and other computational techniques, can render regarding persons discussed in such blog texts; (2) the risks and benefits of such topic modeling; (3) what sorts of translational research models could effectively engage the authors of these blogs, and in some cases individuals described in the blogs, in a dialogue regarding the value and risks of applying computational analysis to their stories; and (4) whether (or to what extent) current human subjects research protocols provide protections to vulnerable individuals represented in such texts.