Authored By: Ahmad Abdulrahim Parpia
Strathmore University
ABSTRACT
Artificial intelligence (AI) is the “in-thing” in today’s digitalized world. However, when it comes to traditional areas of justice, legal professionals are quite apprehensive about the use of AI to expedite their professional tasks. It is generally feared that AI could render certain roles, such as contract law practitioners, obsolete. Nevertheless, there is one field where the use of AI could greatly enhance the easier facilitation of justice; forensic investigations. With the increase in the sophistication of crimes, it is important to have a conversation on how AI can be used to trace the perpetrators of crimes. In this essay, my central claim is that if used properly, AI is a blessing, not only in facilitating criminal proceedings, but also in modernizing our justice systems.
INTRODUCTION
In 2022, the Court of Appeals of New York dealt with an appeal on the grounds of a judge admitting into evidence an AI software’s deduction of the probability of the Defendant committing murder and robbery.[1] In that case, an AI, TrueAllele, which answers the question of “how much more the suspect matches the evidence [than] a random person would”, through the DNA present, was used by the law enforcement to determine that it was the Defendant who had committed the crimes charged.[2]
The AI model, TrueAllele, showed, to a high degree of probability, that the DNA collected at the crime scene belonged to the defendant. The law enforcement used this information to corroborate the evidence adduced to show that it was the Defendant who had committed the robbery and murder.[3]
TrueAllele is not the only AI that aids today’s forensic investigations.[4] AI models are widely being used in both crime scene reconstruction (3D mapping, blood spatter trajectory analysis, etc.), as well as for the analysis of evidence obtained by law enforcement agencies.[5] Admittedly, they have enhanced efficiency, minimized human error and bias during the forensic investigation, and reduced costs associated with manual labour and expert analysis.[6]
AI models process crime scenes faster, are able to analyse patterns, and have the ability to handle large amounts of digital data.[7] Nevertheless, AI is still a new addition to the field of forensic investigations. As such, it undoubtedly poses many challenges that the legal profession must adapt to in order to effectively enhance access to justice.[8] It requires some degree of human intervention when it comes to the training of the AI models, as well as in the interpretation and validation of the results generated by the AI models. For instance, before admitting the evidence obtained from TrueAllele, the Trial Court in the case of People vs Wakefield ordered that the evidence go through stringent procedure to ensure its accuracy.[9]
The use of AI therefore raises questions about its reliability, legal admissibility, ethical implications of using AI, and its larger role in advancing access to justice; all of which shall be discussed in this piece.
AI TECHNOLOGIES IN FORENSIC INVESTIGATIONS
The field of forensic investigations has been described as a highly fragmented system due to the different individuals who make use of this field.[10] First, you have crime scene investigators who collect evidence from the scene; second, you have scientists who examine the clues to make a deduction; and finally, you have lawyers who take the evidence to court together with the expert evidence of the scientists, to either push for a conviction or an acquittal. Additionally, the traditional method of conducting an autopsy has many limitations, such that some minute observations may be missed out by a tired or negligent eye.[11]
Introducing AI models in forensic investigations has therefore been a pivotal step in facilitating evidence examinations. One such task carried out by AI models is crime scene reconstruction to find the answers to the “who, what, when and how” of events.[12] Forensic investigators look into blood spatter analysis, gunshot trajectories, footprints, all of which can be analysed using AI models, to provide a largely accurate depiction of what really occurred, including the relevant timeframe of events.[13]
AI models are also used to determine the Postmortem Interval (PMI), which is the time between the death of the person and the discovering of the body.[14] Data input in the AI model regarding the various stages of decomposition gives an accurate prediction of the time of death of the person. [15] Similarly, AI colour detection techniques have also been frequently used by forensic investigators[16] to note the various colour changes in bruises.[17]
Another function that AI models play in forensic investigations is posture monitoring.[18] This is where a CCTV captures footage of a person of interest, which is then run through an AI model in order to determine to a high degree of probability that the person suspected is indeed the person caught in the CCTV footage. Digital evidence analysis of emails, computer files, and chats is widely conducted by AI models.[19]
Instances in which the bodies of victims have been mutilated beyond recognition, certain parameters, such as fingerprints, iris pattern, DNA patterns, and other forms of biometrics are fed into the AI machine, which is then able to generate the identity of the victim. However, for this to work, such a person’s information would have to pre-exist in the coding of the AI model.
Another breakthrough in the field of forensic analysis is Probalistic Genotyping Software (PGS). An AI that allows forensic investigators to link a genetic sample, usually crime scene evidence, to a person of interest in their case.[20] It computes the probability that a person of interest was the contributor of a DNA sample, through the comparison of specific genetic profiles of all persons of interest created by the AI software.
These advancements in the use of AI in forensic investigations are certainly not the only new technologies used in the field. Instead, they purpose to demonstrate that the AI now plays an integral role in the forensic investigation process. As such, it is important to dive into the accuracy and reliability of the current AI systems.
ACCURACY, RELIABILITY, AND CHALLENGES OF AI IN FORENSIC INVESTIGATIONS
AI tools in forensic investigations, have demonstrated high accuracy in complex cases involving mixed or degraded DNA samples.[21] Their ability to detect subtle patterns and deliver consistent, reliable results enhances objectivity and reduces human error, making them a reliable asset in modern forensic science.[22]
However, forensic investigation is a complex field of science, which attempts to bring together the physical and the human worlds together in a distinctive manner.[23] Adding AI to this mix, while facilitating the process, certainly has its drawbacks. This is especially over AI’s susceptibility to algorithmic bias and over-reliance by legal actors unfamiliar with its limitations. Errors such as false positives or negatives can compromise trial fairness. AI tools require independent validation[24], as noted by the President’s Council of Advisors on Science and Technology (PCAST), and human oversight remains crucial to contextualizing findings and safeguarding against injustice.[25]
ACCESS TO JUSTICE: IS AI A TOOL OR A BARRIER
Admissibility of forensic evidence requires a scientific foundation.[26] AI offers significant promise in enhancing access to justice by expediting forensic analysis, reducing costs, and minimizing human error, especially in overburdened legal systems. However, its benefits are often curtailed by challenges in admissibility. Courts require expert testimony to validate forensic evidence[27], yet AI-generated reports raise questions: who qualifies as the “expert”? The developer, the operator, or the machine itself? This ambiguity can delay proceedings or lead to exclusion of potentially vital evidence. Unless procedural rules evolve to accommodate the unique nature of AI tools, their utility may be undermined. Thus, AI is both a tool for progress and a barrier without legal adaptation.
CONCLUSION
AI in the field of forensic investigations has certainly been a useful and impactful introduction. Severa AI models have presented a significant promise in the field due to their expeditious analysis, reduced human error, and their ability to deal with large amounts of data. However, they must be used carefully, with a proper legal framework developed for their admissibility and the extent to which they can be used. Without such guidance, forensic investigators would fail to consider other possibilities if there exists only one high probability.
REFERENCE(S):
Alaa El-Din EA, ‘Artificial Intelligence in Forensic Science: Invasion or Revolution?’ 10 Egyptian Science and Clinical Toxicology Journal 2, 2022, 20-32.
Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993)
Edmond G and Cunliffe E, ‘Cinderella Story? The Social Production of a Forensic “Science”’ 106 The Journal of Criminal Law and Criminology (1973-) 2, 2016, pp. 219-273.
Galante N, Cotroneo R, Furci D, Lodetti G and Casali MB, ‘Applications of artificial intelligence in forensic sciences: Current potential benefits, limitations and perspectives’ 137 International Journal of Legal Medicine 445-458, 2023.
GAO, Probabilistic Genotyping Software, 1.
Georgieva L, Dimitrova T, Stoyanov I ‘Computer-aided System for the Brouse Color’s Recognition’ International Conference on Computer Systems and Technologies CompSysTech’ 2005, p. 1-5.
Government Accountability Office (GAO), Science & Tech Spotlight: Probabilistic Genotyping Software GAO-19-707SP, September 2019, 1.
https://www.linkedin.com/pulse/ai-crime-prevention-forensic-science-how-saovf/.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9506671/
Khakare R, Fatangare S, ‘Survey on Prediction of Post Mortem Interval using Artificial Intelligence in Forensic Examination’ 7 International Research Journal of Engineering and Technology (IRJET) 5, May 2020, 5760-5763.
Morgan R, ‘Forensic science, revealing the unseen and the unknown’ in Coldwell P and Morgan R (eds), Picturing the Invisible: Exploring Interdisciplinary Synergies from the Arts and the Sciences, UCL Press, 2022, 15.
People vs Wakefield (2019) 175 A.D.3d 158.
People vs Wakefield (2022) 38 N.Y.3d 367.
Golomingi, C. Haas, A. Dobay, S. Kottner, L. Ebert, ‘Sperm hunting on optical microscope slides for forensic analysis with deep convolutional networks – a feasibility study’, Forensic Science International: Genetics 56 (2022).
Shenoy S, Nagar V and Akhith, ‘Artificial Intelligence-Based Techniques for Crime Scene Reconstruction and Investigation: An Overview’ 14 Journal of Forensic Research 4, 2023.
[1] (2022) 38 N.Y.3d 367.
[2] (2019) 175 A.D.3d 158.
[3] (2019) 175 A.D.3d 158. The test calculates the probability that a random person’s DNA would be found on the piece of evidence by determining how often each allele occurs in the general population. In that case, “Specifically, TrueAllele concluded that it was 5.88 billion times more probable that defendant was a contributor to the mixture on the amplifier cord than an unrelated black person, that it was 170 quintillion times more probable that defendant was a contributor to the mixture on the outside rear shirt collar than an unrelated black person, that it was 303 billion times more probable that defendant was a contributor to the mixture on the outside front shirt collar than an unrelated black person, and that it was 56.1 million times more probable that defendant was a contributor to the mixture on the victim’s dorsal forearm than an unrelated black person”.
[4] Shenoy S, Nagar V and Akhith, ‘Artificial Intelligence-Based Techniques for Crime Scene Reconstruction and Investigation: An Overview’ 14 Journal of Forensic Research 4, 2023, page 3.
[5] Galante N, Cotroneo R, Furci D, Lodetti G and Casali MB, ‘Applications of artificial intelligence in forensic sciences: Current potential benefits, limitations and perspectives’ 137 International Journal of Legal Medicine 445-458, 2023.
[6] https://www.linkedin.com/pulse/ai-crime-prevention-forensic-science-how-saovf/ on 27/6/2025.
[7] Shenoy et al, ‘Artificial Intelligence-Based Techniques for Crime Scene Reconstruction and Investigation’, page 1.
[8] Alaa El-Din EA, ‘Artificial Intelligence in Forensic Science: Invasion or Revolution?’ 10 Egyptian Science and Clinical Toxicology Journal 2, 2022, 20-32.
[9] (2019) 175 A.D.3d 158.
[10] Morgan R, ‘Forensic science, revealing the unseen and the unknown’ in Coldwell P and Morgan R (eds), Picturing the Invisible: Exploring Interdisciplinary Synergies from the Arts and the Sciences, UCL Press, 2022, 15.
[11] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9506671/ on 26/6/2025.
[12] Morgan R, ‘Forensic Science, Revealing the Unseen and the Unknown’, 11.
[13] Shenoy et al, ‘Artificial Intelligence-Based Techniques for Crime Scene Reconstruction and Investigation’, page 2.
[14] Khakare R, Fatangare S, ‘Survey on Prediction of Post Mortem Interval using Artificial Intelligence in Forensic Examination’ 7 International Research Journal of Engineering and Technology (IRJET) 5, May 2020, 5760-5763.
[15] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9506671/ on 26/6/2025.
[16] Khakare R, Fatangare S, ‘Survey on Prediction of Post Mortem Interval using Artificial Intelligence in Forensic Examination’ 7 International Research Journal of Engineering and Technology (IRJET) 5, May 2020, 5760-5763.
[17] Georgieva L, Dimitrova T, Stoyanov I ‘Computer-aided System for the Brouse Color’s Recognition’ International Conference on Computer Systems and Technologies CompSysTech’ 2005, p. 1-5.
[18] Edmond G and Cunliffe E, ‘Cinderella Story? The Social Production of a Forensic “Science”’ 106 The Journal of Criminal Law and Criminology (1973-) 2, 2016, pp. 219-273.
[19] https://www.linkedin.com/pulse/ai-crime-prevention-forensic-science-how-saovf/ on 28/6/2025.
[20] Government Accountability Office (GAO), Science & Tech Spotlight: Probabilistic Genotyping Software GAO-19-707SP, September 2019, 1.
[21] Shenoy et al, ‘Artificial Intelligence-Based Techniques for Crime Scene Reconstruction and Investigation’.
[22] Khakare R, Fatangare S, ‘Survey on Prediction of Post Mortem Interval using Artificial Intelligence in Forensic Examination’ 7 International Research Journal of Engineering and Technology (IRJET) 5, May 2020, 5760-5763.
[23] Morgan, ‘Forensic Science, Revealing the Unseen and the Unknown’, 11.
[24] (2019) 175 A.D.3d 158.
[25] GAO, Probabilistic Genotyping Software, 1.
[26] Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993)
