Exploring Trends and Gaps in Artificial Intelligence-Assisted Forensic Firearm Evidence Analysis: A Scoping Review

Abstract

Background: Artificial intelligence (AI) is increasingly used in forensic firearm evidence analysis to improve pattern recognition, automation, and analytical efficiency. Despite advances in machine learning (ML) and deep learning (DL), limitations in scientific validation, explainability, expert involvement, and judicial readiness constrain operational adoption. This review examines the evolution of computational and AI-assisted approaches and the factors influencing their integration into forensic practice. Methods: This PRISMA-ScR scoping review searched Wiley Library, PubMed, IEEE Xplore, ScienceDirect, Google Scholar, and Scopus for literature published between 2000 and December 2025. Of 101 records identified, 3 duplicates were removed, leaving 98 unique records for title and abstract screening. Following screening, 47 records were excluded, and 51 studies met the inclusion criteria and were retained for qualitative synthesis. The findings were interpreted alongside complementary authoritative and foundational literature to support framework development. Results: The reviewed literature demonstrates advances in ML, DL, three-dimensional analysis, and automated comparison, particularly in feature extraction, pattern recognition, and similarity assessment. Persistent limitations concern dataset representativeness, external validation, uncertainty quantification, explainability, standardization, expert involvement, governance, and judicial considerations. Integration of these findings with foundational literature informed a Human-Centered Hybrid AI Framework combining scientific validity, computational intelligence, explainability, expert collaboration, governance, and judicial accountability. Conclusions: Computational performance alone is insufficient for scientifically reliable and judicially defensible AI-assisted firearm examination. The proposed framework positions AI as an explainable, governed decision-support capability operating with qualified forensic experts rather than as an autonomous decision-maker. It provides an evidence-informed reference architecture for future validation, standardization, system development, and responsible forensic implementation.