Introducing “Binary Calculations Are Inadequate to Assess Us,” a Mozilla Creative Media Award project created by Stephanie Dinkins
Training data sets are an essential part of AI — and, increasingly, our everyday lives. Datasets train algorithms that identify our faces, scrutinize our loan applications, and curate our news.
Despite these data sets’ importance, they’re often exclusionary. For example, data sets that train facial recognition AI mostly contain white faces. And data sets that train voice AI mostly contain American English soundbites. As a result, large groups of people — dark-skinned individuals, non English speakers — aren’t always treated fairly by AI. Often the needs, hopes, dreams, and desires of marginalized communities, along with multitudes of cultural nuance, get averaged down to uphold and serve the demands of a very narrow status quo.
Today, a new project by Stephanie Dinkins seeks to change that. Dinkins is launching “Binary Calculations Are Inadequate to Assess Us” — an art project that asks how we make the data-driven algorithms that increasingly control our daily lives more caring. The project is centered on a mobile app that gives everyday people the opportunity to intentionally donate and tag culturally significant information.
Over time, Binary Calculations will provide an openly available data commons with an explicit focus on BIPOC, queer, trans, gender non-conforming and disabled experiences.
Stephanie Dinkins is a New York City-based artist and educator, and the recipient of a 2021 Mozilla Creative Media Award. Binary Calculations is being launched to coincide with a new exhibition by Dinkins at the University of Michigan.
Says Dinkins: “This project takes a DIY approach to producing more holistic, community supportive, carefully crafted data. In the process it also raises awareness about dangerous gaps in training data sets and models alternative methodologies for creating and deploying data. Society needs to think more critically about the intentions, sources and stewardship of algorithms and training data and labeling training data sets. We need to ask: who owns them, who benefits from them, and who is harmed by them?”
Dinkins continues: “This project models alternative approaches to also provide a practical solution. We’re building open, interactive, compassionate alternatives to supplant the current exclusionary data sets that inform AI.”
This project takes a DIY approach to producing more holistic, community supportive, carefully crafted data.
Stephanie Dinkins, Mozilla Creative Media Awardee
Binary Calculations calls on volunteers — and especially volunteers of color — to contribute images and text that honor their life and culture. To do this, volunteers download the Binary Calculations app and answer questions like “What photo makes you happy?” and “Who do you pretend to be kind with?”
These answers are added to a data commons that Dinkins will later open up to the public. Other technologists and artists can then use this training data set to power their own projects, from art installations to apps and beyond.
Mozilla’s Creative Media Awards are part of our mission to realize more trustworthy AI. The awards fuel the people and projects on the front lines of the internet health movement, from activists to documentary filmmakers to researchers.
The latest cohort of Awardees are all Black artists who spotlight how AI can reinforce — or disrupt — systems of oppression. The AI systems in our everyday lives can perpetuate and amplify biases that have long existed offline: Recommendation algorithms promote racist messages. Facial recognition systems misidentify Black faces. And voice assistants like Alexa and Siri struggle to understand Black voices. As the AI in consumer technology grows more sophisticated and prevalent, problems like these will grow even more complex.