Design considerations for systems meant to support humanities and social sciences research
Eetu Mäkelä (University of Helsinki)
From the viewpoint of the humanities and social sciences, collaborations with computer scientists often fail to deliver. At the Human Sciences – Computing Interaction research group, we’ve tried to understand why this is, and what to do about it. In this talk, I will discuss three key elements that we’ve discovered:
- Often, datasets in the humanities and social sciences are not neatly representative of the object of interest. Systems need to provide ways in which to evaluate and counter the biases, confounders and noise in the data.
- Often, there is also a large gap between what is in the data, and what would be of interest. This gap needs to be bridged using algorithms, but care must be given that a) what the algorithm produces actually matches the interest and b) that its application does not introduce bias of its own (also interestingly, algorithm performance metrics of interest here often differ from those generally used in computer science, such as aggregate precision/recall)
- On a process level, collaboration between researchers from different disciplines is hard due to discrepancies in expectations relating to all facets of research, from research questions through methodology to the publication of results. Projects and systems need to acknowledge this, and be designed to facilitate iterative movement in the right direction.
Aalto HELDIG DH pizza seminar on Friday 12 March 2021 at 12.00 (Zoom)