Depictions of people in museum visual arts collections often reflect the histories of inclusion and exclusion. For example, museum collections in the 1920s likely reflect much less the lives of African Americans than those in the 1960s. U-M Museum of Art’s (UMMA) entire collection of ~24,000 pieces of artwork, spanning 150 years, has been digitized. A group of data scientists and artists will apply algorithms to recognize faces and to examine how art collections can (mis)represent humanity and how that representation changes with time.
The project team includes:
- Jing Liu, Michigan Institute for Data Science (MIDAS)
- Kerby Shedden, MIDAS, Consulting for Statistics, Computing and Analytics Research and Dept. Statistics
- David Choberka and John Turner (UMMA)
- Sophia Brueckner (Stamps School of Art & Design).
This project, funded by the U-M Arts Initiative, is the result of months of discussion about how data science and arts can give each other a stronger voice to promote social justice.
The issue of fair and unbiased representation looms large as Big Data and Artificial Intelligence (AI) are impacting many aspects of our society, from political messaging to targeted marketing, from parole granting to hiring practices. With biased data, algorithms will reach superficially compelling but ultimately false conclusions and amplify the biases and the injustices that exist in the data. This project will illustrate this point by generating a collection of composite representations of human faces from the UMMA dataset, and presenting these representations through an art installation, interactive digital displays and narratives. Through this new way to communicate research that is understandable, meaningful, and impactful to the public, this project highlights the caveats that lie beneath Big Data and AI applications and inspire the public to participate in shaping our transition into a data-driven society.