In this article

Uncertain about what?

Imagine a fingermark with a clear central area and a blurred edge. An examiner can follow the main ridge flow but is unsure whether one short ridge ends or divides. After comparison, they see a plausible corresponding feature in the reference. Their uncertainty concerns several different questions: the quality of the source, the feature’s existence and type, and its relationship to the reference.

This hypothetical example shows why a single confidence label is difficult to interpret. Confidence in a feature is not the same assessment as confidence in a correspondence. Neither, by itself, expresses the strength or limitations of the final opinion.

A useful examination record identifies what the uncertainty is about. It also preserves the earlier assessment when later information makes an interpretation feel more convincing. Otherwise, the finished account can make a tentative observation appear more definite than it originally was.

What the GYRO colours communicate

Langenburg and Champod developed GYRO to make the meaning of feature annotations more transparent. Green, yellow and red describe differing confidence in a feature’s existence during Analysis. Their account also connects those assessments to the weight given to a feature, the expectation of finding it in a corresponding exemplar and the tolerance allowed for differences in appearance.[1]

Green carries high confidence and relatively little tolerance; red carries substantial uncertainty, minimal weight and greater tolerance. These colours therefore communicate an examination stance with consequences for Comparison. They do more than separate attractive areas of an image from unattractive ones.[1]

The authors emphasise that uncertainty attached to a red feature should survive comparison. Finding something apparently corresponding in the reference does not erase the initial uncertainty about the mark. Orange serves another purpose: it identifies a feature observed during Comparison using the exemplar. It marks a difference in sequence rather than an additional step on the same confidence scale.[1]

Categories are not probabilities

The GYRO paper includes two training exercises involving 24 and 25 working examiners. The authors assessed feature annotations against the known source skin. Confidence categories were associated with annotation accuracy in those exercises, with the more confident categories showing fewer false features.[1]

The reported overall false-feature proportions were 23% for the complex mark and 5% for the less complex mark. These denominators were annotations: 298 and 506 markings respectively. They were not rates of false source identification. The exercises concerned two fingermarks and do not turn each colour into a universally calibrated probability.[1]

Ulery and colleagues’ sufficiency research reinforces the distinction between a record and an objective inventory. A marked feature count reflects the examiner’s selection and interpretation. It cannot be read as the known number of corresponding features present in the impressions. Adding uncertain features to a count does not remove the need to examine their identity and meaning.[2]

Agreement is a further question. The later markup study cautioned that a majority cluster of annotations does not establish the true minutia in an ambiguous region. Shared confidence can be informative without supplying an independent reference for correctness.[3]

Keeping the assessments distinct in PROFAS

GYRO, the white-box research and the NIST human-factors work were among the foundations of PROFAS. They helped frame a design problem: how can the record preserve distinctions that a total or final opinion compresses? The response is to give observations, assessments, comparison relationships and Evaluation their own roles.

In the fingerprint workspace, a feature has a source position, a type, a classification and an explanatory note. Quality regions can identify their map family, distinguishing mark quality, ridge quality and ridge contrast. A region and a point annotation remain separate records even when they occupy the same part of the image.

The current classification choices are Unassigned, Green, Yellow and Red. These choices do not implement the whole GYRO method. Their use needs an agreed definition and explanation; the presence of a familiar colour alone does not establish its weighting, tolerance or probability. The recorded stage and reference-opening sequence provide separate timing information.

Comparison findings then express correspondence, difference, unexplained discrepancy or noncomparability using exact cited observations. Evaluation records the proposition, supporting account, discrepancies, limitations, narrative and opinion. This makes it possible to inspect how a conclusion uses the observations, while keeping their different questions visible.

Feature position, type and classification can be inspected on the source image. A comparison relationship is recorded separately from the feature itself.

Limits include the decision to stop

A quality assessment also has limits of interpretation. Ordinal assessment labels describe categories under the examination’s agreed definitions. They should not be read as calibrated numerical confidence. A coloured map family should likewise be interpreted under its stated definition rather than assumed to represent a level of fingerprint detail.

Uncertainty can concern whether the material supports further examination at all. The NIST human-factors report discusses both prematurely rejecting suitable material and continuing with unsuitable material. Its documentation chapter also calls for retaining the existence of prints considered unsuitable for comparison. A record consisting only of completed comparisons would omit those decisions.[4]

PROFAS distinguishes an assessment that has not been made from inability to assess, an undetermined interpretation and none observed. It also permits a reasoned stopping decision that preserves the current records without inventing later-stage findings. These states give another reader more information than an empty field.

Learning from the distinctions

For an educator, the useful discussion can begin with the uncertain ridge rather than the final label. What supported the feature interpretation? What tolerance was allowed? What changed after seeing the reference? Did the explanation retain the original limitation? Those questions make the recorded assessment available for discussion.

For researchers, the same distinctions offer measures to investigate under defined instructions and material. They can compare quality assessments, selected features, correspondence and opinions without treating them as interchangeable confidence scores. The platform preserves those records for inspection. Whether particular definitions improve understanding, agreement or performance remains a question for study.

FOLLOW THE EVIDENCE

Sources & further reading

The original research behind this article.

  1. Langenburg, G., and Champod, C. (2011). The GYRO system—A recommended approach to more transparent documentation. Journal of Forensic Identification, 61(4), 373–384.(opens in a new tab)

    The source available for this article is the authors’ retained 2010 draft, including its two training exercises.

    Back to the article
  2. Ulery, B. T., Hicklin, R. A., Roberts, M. A., and Buscaglia, J. (2014). Measuring what latent fingerprint examiners consider sufficient information for individualization determinations. PLOS ONE, 9(11), e110179.(opens in a new tab) Back to the article
  3. Ulery, B. T., Hicklin, R. A., Roberts, M. A., and Buscaglia, J. (2016). Interexaminer variation of minutia markup on latent fingerprints. Forensic Science International, 264, 89–99.(opens in a new tab) Back to the article
  4. Expert Working Group on Human Factors in Latent Print Analysis (2012). Latent Print Examination and Human Factors: Improving the Practice through a Systems Approach. NISTIR 7842. NIST and NIJ.(opens in a new tab) Back to the article