Risky Code by Northeastern University

Risky Code is a digital workshop that uses information and interaction design to increase the algorithmic literacy of workshop participants. The workshop is composed of four modules that lead workshop participants through a series of interactive activities that communicate underlying algorithmic mechanisms and engage participants in critical-thinking activities about algorithmic decision-making.

Risky Code combines information and interaction design principles in the first module to create a model of algorithmic decision-making. Laundry was selected as an analogy to explain an algorithmic process because it is a common task that most people perform. Additionally, laundry involves a binary classification process (e.g., sorting for a particular type of load), similar to many other real-world algorithmic decisions which also rely on binary classification (e.g., choosing whether to parole an incarcerated person, choosing if a child should be assessed for neglect and abuse, etc.).

In the second half of the workshop, participants are asked to engage with the application by building several data visualizations. First, participants build a network diagram that models different stakeholder groups and values from an algorithmic decision-making scenario. Finally, participants build a visualization to assess algorithmic risk by stakeholder group. This participatory process aims to encourage critical thinking about the suitability of algorithmic decision-making for certain scenarios.

Combining stakeholder values with risks from algorithmically informed decision-making is believed to be a novel approach to encourage critical thinking about the impacts of algorithmically informed decision-making. Designers who intervene in algorithmic decision-making face an inherent tension between showing a system's complexity, both socially and technically, and making the system interpretable and understandable to an audience outside computer scientists. Utilizing information design, interaction design, and participatory practice, Risky Code seeks to strike a balance in this tension by contextualizing the tradeoffs to different types of users from algorithmic decision systems.