Why law schools must teach play, failure, and human-centered AI

# The problem: certainty, not exploration For decades legal education has taught students to find a single right answer: locate the controlling precedent, avoid mistakes, and demonstrate certainty under pressure. High-stakes exams reinforce that model by treating errors as failures. That training matches an older model of legal work but clashes with how practice is changing as AI handles routine tasks.

# What needs to change The authors argue for a shift toward exploration and iteration. Law students should learn methods for rapid experimentation: try ideas, measure what happens, fail, and repeat. That approach cultivates practical judgment about when to use tools and how to respond when they fail. It also develops the human skills AI cannot provide—empathy, interviewing, ethical calibration, and courtroom judgment.

# Clinic example: the University of Michigan AI Law and Policy Clinic Clinic puts these principles into practice. Students build AI tools aimed at justice-system problems, giving them responsibility for design choices, evaluation, and stakeholder interaction. Working on real projects forces students to confront messy facts, imperfect data, and competing values. That practical exposure trains future lawyers to lead teams that combine technological capability with client needs.

• Technical literacy about how AI systems work and where they break down.
• Practical skills in scoping projects, collecting and validating data, and setting realistic performance expectations.
• Ethical judgment in how tools affect access to justice and fairness.
• Communication skills for explaining trade-offs to clients, courts, and colleagues.

# Why play and failure matter

# How this prepares graduates for practice Graduates trained this way can evaluate when automation helps and when human oversight is essential. They can run pilot projects inside firms or public-interest settings, manage cross-disciplinary teams, and translate technical performance into legal risk assessments. The clinic model also exposes students to client-facing work earlier, so they practice interviewing, empathy, and real-world problem framing.

# Practical implications for law schools and firms Law schools can adopt clinic-style courses that require iteration, public-facing deliverables, and multidisciplinary collaboration. Firms and public agencies should hire graduates with experience in experimental workflows and create early-career roles that combine supervision with progressive responsibility over AI-enabled tasks.

# Bottom line Teaching certainty as the default is insufficient for an AI-enhanced legal system. Training that normalizes play and failure, centers human judgment, and gives students hands-on experience building tools for justice issues will produce lawyers who can lead and responsibly govern technology in practice.

By Vivek Sankaran, Bridgette Carr, and Sam Flynn — co-directors and collaborators associated with the University of Michigan AI Law and Policy Clinic.