Topic group: AI DATA BIAS
AI Arcade taps into the nostalgic allure of 80s arcade culture to prompt contemplation on the evolution of technology and the existing biases that permeate AI systems. Visitors are transported into an arcade setting where they engage with vintage-style games that teach the principles of pattern recognition and reinforcement learning. As they carry the pre-assumption/historical bias of what a “pear” should look like and how “tic-tac-toe” is played, we challenge these preconceptions by revealing how AI systems “learn” from patterns and adapt through feedback. Yet an intriguing twist emerges: the feedback loop itself, supplied by developers and AI users, might inadvertently introduce biases, potentially cementing these preconceptions within AI systems. This leads us to a profound question: Do these models truly learn as humans do, or are they molded to execute specific functions?
Our artistic intent is to invite reflection on the dualities of nostalgia and progress, intuition and algorithm, and the biases that linger in both humans and AI. Visitors are encouraged to write down their interpretations of the rules or patterns they perceive within the games and deposit them in a black box at the booth. This interactive element is meant to foster a sense of agency in challenging an AI model’s assumptions and promote an ongoing dialogue about the potential biases that are embedded within the algorithms. By engaging in these immersive experiences, the audience becomes participants in a larger narrative—one that challenges them to rethink their preconceived notions, embrace the evolving landscape of technology, and ponder the role of biases in shaping the AI-driven world we inhabit today. As the arcade lights dance, we envision visitors departing the exhibition with a heightened awareness of the possibility of pervasive bias ingrained in AI systems.
Max Haarich (DE), Sarah Cistonm (US)