Likelihood Ratio Misuse

This problem happens when people treat a likelihood ratio as if it gives the chance of something being true. It leads to wrong conclusions because the ratio only compares evidence strength, not the final chance of truth.

Likelihood ratio misuse refers to interpreting a likelihood ratio as the posterior probability rather than as a comparative evidence metric between competing hypotheses. This error conflates evidence weighting with belief updating and ignores prior probabilities required for valid Bayesian inference.

A doctor tells a patient that a diagnostic test is "10 times more likely to be positive if you have the disease than if you don't." The patient hears "10 times more likely" and concludes they almost certainly have the disease—but if the disease affects only 1 in 1,000 people, the actual chance of having it after a positive test is still quite low. The patient skipped the starting probability and jumped straight to the ratio as if it were the final answer.

In a forensic DNA case, an expert testifies that the likelihood ratio for a DNA profile match is 1,000,000:1, meaning the evidence is one million times more probable under the hypothesis that the defendant is the source than under a random-match hypothesis. The jury interprets this as a 99.9999% probability of guilt—a textbook prosecutor's fallacy—because they conflate the likelihood ratio with the posterior probability of guilt. However, if the prior probability of guilt (based on independent case evidence) is, say, 1 in 10,000 (a city-wide suspect pool), Bayes' rule yields a posterior of approximately 99%, not 99.9999%, and at even lower priors the posterior drops substantially further. The failure to integrate prior information creates a systematic bias that artificially inflates confidence and undermines adjudication validity.

People take the evidence ratio and treat it like the final answer. Because they skip the starting beliefs, the decision shifts wrongly toward the stronger-looking evidence.

A likelihood_ratio_misuse mechanism operates when an adjudication component applies a likelihood ratio without incorporating prior probability constraints, creating a weighting_asymmetry across hypotheses. The scoring function privileges the evidence-conditioned density of one hypothesis, producing an asymmetric update that skews posterior estimates.

Always ask what the starting belief was before using the ratio. Combine that starting belief with the ratio to get the real chance.

Enforce explicit prior elicitation and apply Bayes' rule to convert likelihood ratios into calibrated posterior probabilities. Use priors as constraints in the adjudication pipeline to prevent evidence-only updates.

Overstated certainty; Neglected prior context; Biased hypothesis selection

An adversarial actor—such as an expert witness, forensic analyst, or litigant—can deliberately present a likelihood ratio as a standalone probability to a numerically unsophisticated decision-maker (judge, juror, or panel), inflating the apparent certainty of their favored hypothesis without technically lying. By selectively omitting prior probability discussion, the adversary ensures that unfavorable base rates (e.g., a rare event's low prior) never enter the adjudicator's reasoning, systematically biasing the posterior in their desired direction. In intelligence or clinical contexts, analysts can similarly frame competing-hypothesis likelihood ratios as conclusive probability estimates to suppress deliberation and foreclose alternative interpretations.

Require explicit prior elicitation and documentation as a mandatory step in any adjudication or inference pipeline before a likelihood ratio is admitted as evidence, forcing the prior into the record and making omissions auditable. Train decision-makers in Bayesian literacy so they can identify when a ratio is being substituted for a posterior, and establish procedural norms (e.g., court instructions, analytical tradecraft standards) mandating the full Bayes' rule computation. Use calibration audits that compare posterior estimates against base-rate benchmarks to flag cases where likelihood ratios appear to have been treated as terminal probabilities.