RESEARCH INTO PRACTICE / Source, interpretation and review
From source material to an inspectable conclusion
A marked feature is more informative when another person can inspect its source, understand its interpretation and follow how it supported the conclusion. The connection should survive later review and research reuse.
In this article
What does this mark on the image mean?
Consider an examiner who marks a possible ridge ending near a distorted area of a fingermark. A second person opens the finished report and sees a numbered point on a cropped image. They can identify the label, but several questions remain. Which original capture does it belong to? Was the feature considered clear or uncertain? Was it later treated as corresponding with the reference?
The point becomes useful when those connections remain available. Another examiner should be able to return to the source detail, read the interpretation and explanation, inspect the comparison relationship and see how it was used in the evaluation. The record must preserve the distinction between the material and what a person concluded from it.
This example also explains why a complete form is only a starting point. A field may contain an answer while leaving its evidential basis difficult to inspect. A detailed record needs both the requested information and relationships that allow another reader to understand what the information means.
Documentation preserves an account of judgement
Needham and colleagues’ 2022 documentation study found that prompts encouraged consistency in recording particular categories. Examiners nevertheless differed in the feature positions, characteristics and totals they recorded. The study also discusses the difficulty of knowing whether an annotation set exhausts everything an examiner considered available. Completion and substantive agreement remain separate questions.[1]
I began building PROFAS in 2017, informed in part by Megan Needham’s developing research, which I knew before its publication in 2022. That work helped establish the importance of preserving the observations and reasoning behind a conclusion.
ENFSI’s evaluative-reporting guideline provides relevant earlier background. It calls for conditioning information and propositions to be recorded, and recognises that changed conditioning information may require an assessment to be reviewed. Where a probability assignment draws on experience, the relevant knowledge and professional activity need a documented basis available for scrutiny. These principles help explain why a conclusion should retain its grounds; they do not establish PROFAS compliance with the guideline.[2]
Follow the feature through the examination
In PROFAS, the fingerprint annotation refers to a particular source capture and a position in its original image coordinates. Its recorded interpretation and note remain attached. Display changes, such as zooming or cropping the view, do not redefine the source position. A comparison relationship can then cite the particular observations on questioned and reference material that support it.
The evaluation records the proposition, explanation and limitations alongside exact references to accepted records. Stage submission fixes a defined set of revisions. A later change should therefore be considered in relation to its stage and earlier basis, rather than allowing the current arrangement of annotations to stand for the complete history.
Verification adds another account. The verifier’s eligible independent baseline precedes release and opening of the target examination. The subsequent report cites the exact submitted target or baseline records it discusses. This makes the review’s basis inspectable while preserving the earlier contribution. Prior exposure outside the software and the correctness of either account still require separate consideration.
The source changes; the distinction remains
DNA mixture interpretation provides a useful contrast. NIST’s 2024 foundation review distinguishes generating and measuring a DNA profile from interpreting it with assumptions, protocols, tools and expert judgement. Different laboratories can obtain different results on the same mixture. A numerical result does not remove the need to retain the conditions and choices behind it. This is a current extension of the rationale, rather than an original influence.[3]
The PROFAS DNA workflow keeps native signal positions, separately imported peak analysis and examiner interpretation identifiable. Its statistical basis is recorded as not performed or externally supplied, with the associated report and provenance when supplied. PROFAS does not calculate mixture likelihood ratios. Recording an external result preserves its basis without validating the supplier’s method or conclusion.
Measured surfaces introduce a different source frame again. A supported X3P height grid has native sample locations and declared physical axes. Accepted measurements use that grid; lighting and vertical exaggeration affect its display. Research by Duez and colleagues also illustrates why an unmarked region is ambiguous: an examiner may not have recognised it, may have rejected it or may have relied on sufficient information elsewhere. The absence of markup does not recover that explanation.[4]
A reusable dataset needs an explicit unit
If these contributions are later selected for research, the meaning of a row matters. It might represent a source capture, an accepted annotation revision, an examination, an imported peak or a review. Counting all of them as independent examples would obscure the relationships that made the original record interpretable.
Gebru and colleagues’ Datasheets for Datasets asks creators to describe what their instances represent, their composition, missing information and relationships. The FAIR principles add clear usage conditions, detailed provenance and relevant community standards. Both provide useful guidance for a dataset project. They are not evidence that a particular collection has valid forensic labels or permission for every proposed use.[5][6]
PROFAS’s structured records provide a basis for investigating such collections while retaining source and encounter relationships. A dataset design still needs to define its task, contributors, reference basis and permitted reuse. Multiple examiners’ agreement, recorded commentary and a completed review do not automatically create ground truth or grant rights to train a model.
Make the basis of the conclusion available
Returning to the possible ridge ending, the most informative record is one that lets a reader inspect the original detail and the interpretation attached to it, then follow its role in comparison, evaluation and review. They may still disagree with the feature selection or the conclusion. The record gives that disagreement a specific basis.
The same discipline applies when a source is a signal or a measured surface. Source identity, recorded meaning, subsequent use and review should remain connected without turning one into proof of another. PROFAS aims to preserve those connections so that examinations can be discussed, challenged and studied with their supporting material in view.
FOLLOW THE EVIDENCE
Sources & further reading
The original research behind this article.
- Needham, M., Fieldhouse, S., Morris, W., Wheeler, J., and Nicholls, G. (2022). Collaborative practise in forensic science and academia: The development of a documentation strategy for fingerprint examinations in an English fingerprint bureau in the ISO 17025 era. Science & Justice, 62(3), 336–348.(opens in a new tab) Back to the article
- European Network of Forensic Science Institutes (2015). ENFSI Guideline for Evaluative Reporting in Forensic Science. Approved version 3.0.(opens in a new tab)
Guideline, approved version 3.0; cited for evaluative-reporting principles, not software certification.
Back to the article - National Institute of Standards and Technology. (2024). DNA Mixture Interpretation: A NIST Scientific Foundation Review. NISTIR 8351.(opens in a new tab)
Scientific foundation review, NISTIR 8351, December 2024.
Back to the article - Duez, P., Weller, T., Brubaker, M., Hockensmith, R. E., Lilien, R. (2018). Development and Validation of a Virtual Examination Tool for Firearm Forensics. Journal of Forensic Sciences, 63(4), 1069-1084. DOI: 10.1111/1556-4029.13668.(opens in a new tab) Back to the article
- Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., Daumé III, H., and Crawford, K. (2021). Datasheets for Datasets. Communications of the ACM, 64(12), 86–92.(opens in a new tab)
Retained author-manuscript composition questions.
Back to the article - Wilkinson, M. D., Dumontier, M., Aalbersberg, I. J., et al. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data, 3, 160018.(opens in a new tab) Back to the article