Kavanaugh: A Probability Poll

An original reader poll (2,350 responses) eliciting probabilities that Kavanaugh assaulted Dr. Ford, to test whether probabilistic framing tempers a charged judgment. Findings: an overall mean of 52.6% that is a 'fake balance' from canceling demographic skews; breakdowns by party and gender; respondents with more background knowledge gave more extreme and slightly more pro-innocence answers (he flags a question-design flaw — he didn't ask about Dr. Ford); 'neutral' respondents averaged ~50% (uncertain); no consensus on the probability threshold for rejecting a nominee (confirm below 25%, reject above 80%, split between); and an admit-and-apologize hypothetical that gained ~10% (poorly worded, he notes). Conclusion: people CAN give well-formed, self-consistent probabilities, but this removed neither partisan bias nor disagreement — 'not a good way to get beyond confusion' on an issue like this.

Quality 58 · Solid. Solid, high (58). Competent original survey work with honest self-critique and a modest but real finding (probabilistic framing doesn't dissolve partisan disagreement). Held to Solid by its topical, chart-dependent, of-the-moment nature.

Claude’s paradigm shift 42 · Moderate. Moderate, low (42). The 'do probabilities reduce overconfidence/bias?' question was established; this applies it to one charged case, modest novelty.

Real-world impact 2 · Minor. Minor (2). A topical reader poll with a small generalizable point; no durable real-world effect.