What am I reading in ScholarWorks? June 2024

For this month’s What am I Reading in ScholarWorks? I was reminded once again that a perk of my job is getting to read so much of the scholarship of the Smith community. That said, my area of expertise does not always overlap with a given discipline, so when I come across a paper beyond the depth of my knowledge that I can understand and learn from—I get very excited.

Computer Science Professor, Jordan Crouser’s recent paper “Metric Ensembles Aid in Explainability: A Case Study with Wikidata” co-authored with doctoral candidate Grant Forbes from North Carolina State University is just that sort of paper. While being about explainability it is itself eminently explainable. Very refreshing. If you have an interest in machine learning models and what some of the problems are that researchers are dealing with when building models that give good results (More relevance! Less noise! Explaining the internal process!) then this paper will interest you. And as a bonus if you don’t know what Goodhart’s law is you will after reading this paper. Making a case for metric ensembles as a tool for explainability with demonstrations using Wikipedia Data is an excellent gateway into the process and makes the paper accessible to anyone with an enthusiasm for the subject or for learning something new. Enjoy!

To view this work in Smith ScholarWorks, visit:
https://scholarworks.smith.edu/csc_facpubs/398/

Smith ScholarWorks is Smith’s institutional repository of digital scholarship and research materials created and curated by faculty, students, and staff at Smith College. Browse by collections, disciplines, and authors.