Algorithmically fair? by Northeastern University

Algorithmically fair? is a data physicalization project that visualizations disparate rates of a commonly used recidivism algorithm in the United States. Through simple pie charts, the project compares how the error rates of the algorithm differ for Black and White defendants. The physical forms obstruct part of the visualizations; the audience must physically engage with the artifact to reveal the rates. The third dimension, the form's height, indicates the number of people affected by these differing rates. Light draws the audience’s attention to the error rate, highlighting the discrepancy between Black and White defendants, and asks the audience to think critically about what it means for an algorithm to be fair.

Algorithmically fair? uses uncommon materials to create data visualization and three-dimensionality to encode additional data attributes. The materiality of this piece invites interaction from the audience, creating an interesting pedagogical model of algorithmic literacy through data visualization. Algorithmically fair? is believed to be a novel way of using three-dimensional structures to visually communicate the outcomes of algorithmic processes that is attentive to art and design.