The Inversion of Doubt: Public Skepticism Is Concentrated Where Epistemic Warrant Is Highest and Absent Where It Is Lowest

Executive Summary

Public skepticism is not randomly distributed across scientific domains. The claims attracting the most sustained public doubt — vaccine safety, anthropogenic climate change, biological evolution — are among the most empirically robust in the history of inquiry. Meanwhile, macroeconomic projections, demographic forecasts, and clinical practice guidelines with documented failure rates of 40 to 97 percent circulate through prestige journalism and policy discourse with near-zero friction.

This inversion is not the product of individual irrationality. It is structurally produced by three forces: the asymmetry between falsifiability timescales in hard versus soft science, identity-protective cognition that indexes skepticism to perceived threat rather than accuracy, and a selection environment that rewards transmissibility over calibration. Where ground truth is unavailable, the correction machinery never activates, and claims circulate without measurement or challenge.

A provisional threshold operationalization is offered: domains meeting two of three measurable criteria — registered report divergence exceeding 20 percentage points, expert calibration gap exceeding 20 percentage points, and falsifiability timescale exceeding a professional generation — warrant presumptive skepticism of public acceptance rates.

Key Contributions and Methodology

The paper's primary contribution is synthesizing five independent lines of evidence into a single cross-domain argument: that high public acceptance rate can serve as evidence of low accuracy above a measurable threshold. The component mechanisms have been documented in isolation — Ioannidis on false discovery rates, Fanelli on positive result gradients, Tetlock on forecasting failure, Kahan on identity-protective cognition — but have not previously been integrated into a unified framework with threshold operationalization.

The methodology is synthetic rather than experimental. Each section draws on the strongest available evidence from its domain: NISTEP Delphi technology foresight retrospectives for hard science reliability, Fanelli's positive-result gradient and Scheel et al.'s registered reports comparison for metascientific measurement, Tetlock's Expert Political Judgment project and the Survey of Professional Forecasters for calibration data, Herrera-Perez et al.'s meta-epidemiological review for clinical reversal rates, and Pew Research Center survey data for public opinion.

Key Findings

• Epistemic warrant hierarchy: Hard sciences show prediction realization rates of about 70% or better and replication rates of 70-100% for well-established findings, while the soft sciences replicate at 36-50% with positive result rates of 91-97%
• Registered report divergence: Scheel et al. found 96% positive results in standard publications vs. 44% in registered reports — a 52-point gap revealing the selection filter's magnitude
• Expert calibration gap: Survey of Professional Forecasters respondents report 53% average confidence with 23% accuracy; in Tetlock's data, media contact frequency correlates negatively with calibration (r = -0.12)
• Clinical reversal: Of 3,017 RCTs testing an established practice, Herrera-Perez et al. found 396 (about 13%) reversed it; only about 6-8% of NCCN therapeutic recommendations rest on Category 1 evidence
• Identity-protective cognition: In Kahan's motivated numeracy experiment, higher numeracy improved accuracy on a neutral (skin cream) problem but increased partisan bias on an identical gun-control problem
• Composite threshold: Domains meeting 2 of 3 criteria (RR divergence >20 pts, calibration gap >20 pts, falsifiability timescale >30 yr) are structurally positioned for the inversion to operate

Key References