Exploring the “Magic” of Algorithmic Predictions
We draw parallels in the predictions made by algorithms to those of divination, which likewise render the abstract into actionable recommendations. Because of their inherent opacity and grounding in quantitative data, algorithmic predictions appear to possess a sense of technical "magic" that non-experts do not often perceive as a constructed reality; by contrast, fortune telling is frequently viewed with skepticism, received not as truth, but instead as a means for entertainment or self-reflection. Our objective is thus to evaluate whether reframing algorithmic predictions with a deliberately mystical lens may lead people to reconsider the “magic” of technical objects.
We design a system which enables self-guided tarot card readings, involving a conversational interface, screen-based visualization, and natural language processing to make predictions of the user's past, present, and future. In 10 research sessions with our prototype, we found that interactions reflected broader, generalizable insights about interactions with computational prediction systems, including lack of user agency when faced with potential manipulation, privacy concerns, as well as poor performance. Within the traditions of reflective and speculative design, we discuss implications of computing's tendency toward magical presentation and imagine alternate paradigms.