Twin Mirror

T Sensing David Benjamin

Computation and algorithms are increasingly influencing all aspects of contemporary life, including architectural design. In the past decade, advances in artificial intelligence and machine learning have endowed computers with new abilities to solve problems and reason about our world. Like all technologies, these algorithms involve many assumptions and biases. However, the biases of machine learning are particularly troubling because they are often hidden, sometimes even from their own inventors. As applications of these algorithms have grown, so too have documented cases of their bias, including racial profiling in policing, sexism in job listings, and uneven distribution of resources in urban neighborhoods. Twin Mirror makes visible the hidden assumptions of machine learning algorithms by processing a live video feed of the exhibit’s visitors through two different face-recognition models trained on two different sets of data. It then uses the data to generate a dynamic facade and bring these dual interpretations of reality to the public. The building envelope is a key site for the city’s commons. Our project uses the latest in machine learning to create a building envelope that reacts directly to city as a collective of citizens while simultaneously exposing the underlying assumptions and biases hidden within this technology.