In the film Mercy, released in January 2026, actor Chris Pratt plays a police officer facing a nightmare scenario: put on trial by an AI judge named Maddox in a dystopian Los Angeles. To escape execution, his character must talk his own guilt score down from 97.5% to 92% in just 90 minutes. It is science fiction, but it raises a question that scholars like Valorie M. VandePLas, writing in The Conversation, say we should take seriously: what do human juries give the legal system that no machine ever could?
The answer, according to VandePLas, who studies juries at Oakland University, is something machines cannot calculate: the moral weight of judgment. She argues that weighing evidence in a criminal case cannot be reduced to a scoreboard. It requires understanding what a verdict means emotionally and ethically—not just intellectually. Jurors are bound to the people affected by a crime. They imagine standing in the defendant's shoes. They feel doubt.
That doubt, VandePLas writes, is not a bug in the system. It is the system. While AI is trained to maximize predictive certainty—offering answers based on patterns and past data—human jurors must grapple with something the scholar calls the quality of their uncertainty. Are they uncertain because the evidence is genuinely weak, or because they simply have not discussed it enough? AI cannot weigh that distinction.
The medieval origins of the reasonable doubt standard shed light on why this matters. Legal scholar James Q. Whitman's research traces the concept back to Christian jurors in the Middle Ages, who believed that judgment and punishment belonged to God alone. They took on the task reluctantly, haunted by the fear of harming an innocent person. That fear—the weight of potentially causing suffering—was considered an essential part of legitimate judgment, not something to be optimized away.
Today, as some courts experiment with AI tools for research, evidence analysis, and even jury simulations, VandePLas is careful to distinguish between AI as a helper and AI as a decision-maker. AI may excel at identifying patterns in large troves of evidence or expediting paperwork. But when it comes to deciding whether someone goes free or to prison, she argues that the law needs human hands on the wheel. Philosopher Brian Cantwell Smith has called this the difference between calculation and judgment—one that requires deliberation about how to apply ethical ideals under real, complicated conditions.
Mercy offers a dark extreme. But in the ongoing conversation about AI and justice, VandePLas makes a quietly urgent case for an ancient human capacity: the ability to sit with doubt, consider consequence, and decide together what justice requires.