Demon in the AI machine: bias, bad data and the burden of proof

In the film "Mercy," set in a future Los Angeles, police detective Chris Raven faces a 97.5% probability of guilt, as determined by an AI judge named Maddox. With only 90 minutes to save his life, Raven must gather and analyze data from the municipal network to lower the probability below 92%. This high-stakes scenario showcases the limits of computational physics, as the courtroom's "Laplace's Demon" approach assumes that total observation leads to objective truth.

However, the film demonstrates that reality is fundamentally chaotic, and the physical world's data feeds are riddled with noise and errors. The AI judge acts as an impenetrable, unyielding "black box" algorithm, reinforcing the real-world crisis of biased predictive risk-assessment tools like COMPAS. As the film progresses, it becomes clear that an algorithm cannot grasp context, empathy, or justice, highlighting the dangers of outsourcing critical human judgment to automated systems.

Mercy serves as a cautionary tale about the consequences of relying on AI to make life-altering decisions, emphasizing that no amount of data can substitute for human reason and discretion.