My digital double and me by Politecnico di Milano
My digital double and me is an infopoem reflecting on the insanely specific and biased inferences algorithms make about us, in a continuous exchange of informations between platforms. Its genesis lies in the inferences Spotify made about me, based on data about my online behaviour coming mainly from third parties. The dataset is taken from the dossier of personal insights about me directly requested to Spotify, which includes more than 500 inferences about my lifestyle, my personality, my social class, my user and purchase habits, my interests, my work and much more. For the infopoetry I selected the most relevant and unified similar ones, mantaining the focus also on how often each assumption appears. The chosen ones were eventually translated with a text-to-image AI (Stable Diffusion), using each one as a prompt. The resulting generated images are merged proportionally, considering the frequency of occurrence of each assumption in the dataset and, therefore, how much each one is associated with me. The final result is, to all intents, the appearance of my digital double according to those algorithms and their stereotypes. The visual personification of my digital double, aims at stressing the human component of those assumptions and letting the user perceive that they aren’t just words but labels with a huge weight. We can clearly understand from it that our digital traces define us in ways that often do not reflect our person. In this way the project underlines how the awkward inferences that algorithms can make and the stereotyped results of a text-to-image model speak for themselves in dispelling the myth of unbiased and objective AI.