Well Known Strangers is a mechanism that aims to retrieve the aggregated digital memories people cast on social networks by using machine learning algorithms over data. The mechanism generates a physical representation of the person who produced that memory.
A plethora of agents transform our memories in the digital space. Some of them are intentional, by the user, others a mere product of the technical limitations of the medium and some are designed to control. The implications of these transformations are becoming more and more relevant, as a large part of the population is now engaging with tools for digitizing memories. My intention with this project was to expose the nature of those agents. Agents that are shifting the idea of memory, history and identity.
We use technology to extend the capabilities of the body, in order to hit harder, to dive deeper, to travel faster. A phone book for example is a device that lets us remember a large amount of numbers. We don't actually know the numbers but we feel like we do. This form of storage is called "prosthetic memories" and is fundamentally a technological construct. Our memories are highly mediated: they take the form of pictures, videos, text, code and music. And in this "casting" is were the true nature of prosthetic memory rises.
The technological convergence towards a single artifact have created a pervasive form of software, beyond the limits of the devices that are used to feed it. Well Known Strangers brings back to physical space the data that has been crunched by the process of digitizing our own memories. Those uncanny automatons are the product of the blurring of the once distinct separation between humans and data. This project is composed of 2 parts: The Voice and The Face.
The Voice: The “tweets generator” algorithm first makes an API call to Twitter servers, receiving the last 600 tweets of the selected profile. Those tweets are then striped from all the links and mentions to other users and is feed to the algorithm, based on Markov’s Chains. The algorithm analyzes the text and determines the probability of a successive words. Once this process is done, it generates a 140 character limited sentence.
The Face: Portraits are based on Facebook’s facial recognition software. A Python script retrieves all the public accessible images where the target profile has been tagged and saves them to a directory. The images on the directory are matched against themselves to retrieve the part of the image where the user’s face is residing. Once the faces are acquired a second analysis is performed on the images for the position of the face and creating an aggregated image of the face. The same information is passed to a third piece of software that uses the tracked features of the faces and matches them with a generic face model, based on Facebook’s own face model, finally generating a 3D representation of the person.