Likelihood Ratio Misuse

A likelihood ratio can get mistaken for the actual chance that something is true. That mistake produces wrong conclusions, because the ratio only compares how strong the evidence is, not the final probability.

This misuse interprets a likelihood ratio as the posterior probability itself, rather than as a comparative evidence metric between competing hypotheses. It conflates evidence weighting with belief updating and skips the prior probability that valid Bayesian inference actually requires.

A doctor tells a patient 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 profile match is a million to one — the evidence is a million times more probable if the defendant is the source than if it's a random match. The jury interprets this as a 99.9999% probability of guilt, conflating the ratio with the actual chance of guilt. But if the prior probability of guilt, based on independent case evidence, is only 1 in 10,000 given a city-wide suspect pool, the properly calculated posterior comes out closer to 99% — and at a lower prior, it drops substantially further still. Skipping that prior inflates confidence well beyond what the evidence actually supports.

The evidence ratio gets treated as though it were the final answer on its own. Skipping the starting belief is exactly what pulls the decision too far toward the stronger-looking evidence.

A likelihood ratio gets applied without incorporating the prior probability, creating an asymmetry across the competing hypotheses. That imbalance privileges whichever hypothesis the evidence favors, skewing the resulting estimate away from what the true prior would support.

Always asking what the starting belief was, before using the ratio, is the direct fix. Combining that starting belief with the ratio gives the actual probability.

Explicit prior elicitation, combined with proper Bayesian updating, converts the likelihood ratio into a calibrated posterior directly. Treating the prior as a mandatory input in the adjudication process prevents an evidence-only update from slipping through.

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.