Send station · the claim, dispatched
Information Bias
The tendency to seek or value additional information even when no possible answer could change the decision or action one will take.
Supports
The bias files before requesting a test as confirmation.
Baron, Beattie & Hershey (1988), Experiment 4 (abstract diagnosis, provided probabilities)
Subjects gave substantial positive ratings to normatively worthless tests
n≈14
Baron, Beattie & Hershey (1988), Experiments 1–3 (elicited probabilities)
Across Experiments 1–3
n≈14
Refutes / corrects
Baron, Beattie & Hershey (1988), interview probes (globoma & umphitis problems)
A majority of the 33 answered correctly
n≈33
In the wild
Choosing Wisely campaign launched to curb tests/procedures that do not change management
In 2012 the ABIM Foundation and specialty societies launched Choosing Wisely
2012
Overuse of cardiac biomarker testing in an emergency department
A Johns Hopkins Bayview Choosing Wisely project documented that despite guidance to limit troponin to three draws per 2…
2013
Dispatch log · replication limited-evidence. The original demonstration is a small-sample (n≈10–33 per study) within-subject finding using abstract diagnosis vignettes, widely cited but, unlike the same authors' outcome-bias work, not subject to large pre-registered direct replications of the information-bias effect specifically. The closely related phenomenon — clinicians ordering tests that cannot change management — is robustly documented in the over-testing / low-value-care literature, but those field findings are driven by many factors (defensive medicine, fee-for-service incentives, patient demand) beyond the cognitive bias, so they corroborate the pattern rather than cleanly replicate the lab effect. Treat the lab effect as a foundational but modestly-powered demonstration.
Pressure diagram · why it sorts this way
Baron, Beattie and Hershey attribute the bias to a heuristic in which people evaluate a question by whether it could yield evidence about the hypotheses, while failing to consider the decisive argument against it: that a question is only valuable if some answer could change the hypothesis acted on. People treat information as good in itself, seek to raise confidence in an already-favored option, or try to rule out 'irrelevant' alternatives that would not be treated regardless. It is framed as a failure to ask 'how would this change what I do?' rather than a motivational distortion.
It overlaps with, but is distinguished from, confirmation/congruence bias (seeking answers expected to support the favored hypothesis) and from a desire to reduce psychological discomfort with uncertainty. Some theorists argue much real-world 'information bias' is rational once one adds plausible unmodeled utilities — anticipated regret, reassurance value, legal/defensive cover, or the option of unexpected findings — so the gap from the normative model may be partly a too-narrow utility specification rather than pure error.
Manual override
Apply a value-of-information check: for each possible result of the inquiry, write down the action you would take. If the action is identical across all outcomes, skip the inquiry. Baron et al. found that telling people about diagnosticity reduces but does not eliminate the bias, so make 'how would this change my decision?' an explicit gating question rather than relying on awareness alone.
In Baron, Beattie & Hershey (1988), the bias persisted even for subjects who clearly understood the task and even when the test was given a $5 cost; some subjects only recognized the test was 'irrelevant' after the interviewer walked them through the action-identity argument.
Adjacent stations
- Congruence biasIntroduced in the same 1988 Baron, Beattie & Hershey paper: overvaluing questions likely to yield a positive result for the most likely hypothesis. A confirmation-style cousin of information bias.
- Certainty biasAlso from the 1988 paper: overvaluing questions that could establish or rule out a hypothesis with 100% certainty, even when that certainty is decision-irrelevant.
- Confirmation biasConfirmation bias is about seeking/interpreting evidence to support a prior belief; information bias is about valuing information that cannot change any action, regardless of which side it favors. They overlap in test-selection behavior.
- Value of information (expected value of sample information)The decision-analytic normative benchmark: information has value only if it can change the optimal action. Information bias is the systematic deviation from this benchmark.
- Ambiguity / uncertainty intolerance and analysis paralysisDiscomfort with uncertainty motivates gathering decision-irrelevant data; excessive gathering can also stall decisions.