Cognitive Bias, Error and Quality in Odontology

Itiel Dror, a cognitive neuroscientist at University College London, has published extensively on how contextual information distorts forensic comparison judgments. His 2006 experiment with latent fingerprint examiners showed that the same pairs of prints, presented under different contextual frames, elicited contradictory conclusions from the same examiners. Variants of that design have since been replicated in fingerprints, footwear marks, and bite-mark analysis.

Several bias types recur across the odontology literature, and the mitigation strategy for each differs.

• Confirmation bias: the examiner looks for features that confirm a proposed match and spends less effort searching for features that would exclude it. Especially pronounced when the examiner knows who the suspect is before doing the comparison.
• Anchoring: when a first examiner has concluded 'identification', a second examiner doing a review is pulled toward that conclusion even if the evidence would support 'inconclusive' independently.
• Contextual bias proper: irrelevant case facts (a criminal record, a confession, a prior conviction) shift the threshold at which the examiner decides the features are sufficient for a conclusion.
• Investigator effect: when the forensic examiner is embedded within the investigative team, social pressure to support the working hypothesis can operate below the level of conscious awareness.

The errors that attracted the most scrutiny came in two categories: false identifications in disaster victim identification (DVI) settings, and false bite-mark attributions leading to wrongful convictions. Both categories produced formal investigations and drove reforms.

In DVI work, the most studied failure involved misidentifications in the aftermath of the 2004 Indian Ocean tsunami. The sheer scale (an estimated 220,000 dead across 14 countries) and the pressure to return remains to families quickly created conditions in which confirmation bias could operate powerfully. Post-event reviews found cases where dental comparison was completed too quickly with poor-quality antemortem records and without proper reconciliation. Some families received the wrong remains.

In criminal contexts, the most consequential errors involved bite-mark testimony in homicide cases. Beyond Ray Krone (Arizona, 1992, exonerated 2002 by DNA), several cases from Mississippi in the 1990s linked to a single forensic odontologist, Michael West, who used an idiosyncratic ultraviolet fluorescence method involving special filtered glasses to detect bite marks invisible to other examiners, a technique never subjected to peer review or proficiency testing.

Blind verification has the clearest evidence base of any quality control procedure in forensic comparison. A second examiner reviews the same evidence independently, without being told the first examiner's conclusion. If both reach the same conclusion, confidence is higher. If they disagree, the disagreement itself is informative data about the evidence, not an error to be suppressed.

In a non-blind review, the second examiner is informed of the first conclusion before starting. Studies consistently show that this produces high agreement not because the conclusions are reliable but because the second examiner anchors on the first opinion. The apparent inter-examiner agreement is artefactual. Blind review breaks this anchoring by design.

Implementing blind review requires a case management procedure that withholds the first conclusion until the second examiner has submitted an independent finding. Evidence from fingerprint and DNA laboratories that have introduced the procedure indicates the operational disruption is manageable and the quality improvement is measurable.

Proficiency testing presents examiners with test problems that have known correct answers, concealed from the examiner. Results are scored and reported, usually with aggregate statistics across the group. This is standard practice in medical laboratory accreditation (under ISO 15189) and in forensic DNA typing (under ISO 17025 and FBI QAS). It has been applied more unevenly in comparison disciplines.

The ABFO ran formal proficiency exercises in the late 1980s and 1990s. The results were not reassuring. In one widely cited 1999 exercise, 32 ABFO diplomates assessed four bite-mark cases. Agreement on whether the injury was a human bite mark at all was below 85%. Agreement on specific comparisons was lower. These results were used as evidence in subsequent Daubert hearings to argue that the discipline did not have a known or acceptable error rate.

Post-2016, both the ABFO and INTERPOL DVI have moved toward requiring documented proficiency and competency testing as a condition of participation in casework. ISO 17025-accredited dental comparison laboratories now include proficiency exercises in their quality management systems. The shift is from self-assessed expertise to externally validated performance.

Instructing examiners to be more objective does not reduce contextual bias. The cognitive mechanisms that produce it are automatic and largely unconscious. Effective interventions are structural: changing who receives what information, and when, rather than relying on individual willpower.

• Sequential unmasking: release case information to the examiner in stages, starting with only what is needed for the comparison, and adding investigative context only after the technical conclusion is recorded.
• Task-relevant versus task-irrelevant information: the examiner needs the antemortem records and the postmortem findings. They do not need to know the suspect's criminal history, whether a confession exists, or what other forensic evidence has been found.
• Documentation of the comparison before consultation: recording the features observed and the conclusion reached before discussing the case with investigators creates a contemporaneous record that is harder to retrospectively revise.
• Separation of the verification examiner: the examiner performing blind verification should not have informal access to the case file, the investigating officers, or the first examiner's notes before completing their own assessment.

Quality management systems in forensic laboratories, required under ISO 17025 accreditation, create an institutional framework for managing bias risk. These systems define procedures, require documentation, mandate proficiency testing, and produce an audit trail for external review.

For forensic odontology specifically, the Texas Forensic Science Commission in 2016 issued a report recommending that bite-mark evidence not be admitted in Texas courts until its reliability had been better established. This was significant because it came from a state oversight body, not a defence attorney, and it applied the same framework the commission uses to assess laboratory error in DNA cases. Several other states have since restricted or excluded bite-mark evidence on similar reasoning.

In DVI contexts, INTERPOL's DVI guidelines require an independent second confirmation before a positive identification is recorded on the DVI Reconciliation Form. The confirmation must be performed by a different expert, blinded to the first finding. This is operationally embedded quality control, not an add-on.