When machines think straighter than justices
Justice Kagan's surprise at Claude's constitutional analysis reveals an irony: while we fixate on AI hallucinations, we miss when machines reason more systematically than humans, modeling dispassionate legal analysis.
I keep thinking about Justice Kagan's reaction to Claude's constitutional analysis. Not because she praised an AI—that's noteworthy but not shocking. Because of what made her praise feel so genuine: surprise at encountering actual analytical rigor on a Confrontation Clause issue, "the court has divided on twice."
Ironic. We spend so much energy worrying about AI hallucinations that we miss the flip side—what happens when machines actually reason more systematically than the humans around them? Against that backdrop, Claude's methodical analysis of precedent probably felt like intellectual oxygen.
Adam Unikowsky's experiment showed Claude holding steady against bad-faith questioning while charting reasonable paths through complex doctrine. Meanwhile, human justices routinely let political priors, personal grievances, and sheer intellectual laziness drive their reasoning. The algorithm stayed focused on legal principles. The humans wandered into ideology.
This doesn't mean AI should replace judicial reasoning—those guardrails against manipulation remain paper-thin, and we're still learning that simple number sequences can turn language models into "homicidal owl-lovers." But it suggests something more hopeful than our current hallucination panic allows. Maybe AI's real value isn't matching human intelligence, but modeling what dispassionate legal analysis actually looks like. In a system where cognitive bias masquerades as jurisprudence, that might be revolutionary.
