How can AI help investigators interpret chemical information while keeping the reasoning behind its outputs understandable? An ARIEN research contribution addressed this question at SHIELD-AI 2026, held in Naples on 11 September within the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD).<\/p>\n
The workshop, \u201cSecurity, Human Awareness, and Risk Mitigation in AI-driven Systems\u201d, brought together research on machine learning, cybersecurity and human\u2013AI interaction. Its focus extended beyond technical performance to the risks arising when AI operates in complex environments where reliability, trust and human judgement matter.<\/p>\n
ARIEN\u2019s contribution, \u201cA Novel AI-Assisted Knowledge Discovery Service to Support Drug-Manufacturing Hypotheses\u201d was authored by Federico Negri, Vito Morreale, Ernesto La Mattina, Marialuna De Tommaso and Valentina Mazzonello.<\/p>\n
Developed within Engineering\u2019s ATLANTIS platform, the service supports the examination of possible illicit drug-manufacturing activities. It combines molecular similarity analysis with retrosynthetic route matching to identify possible target substances, highlight matched and missing precursors, and provide an interpretable risk score.<\/p>\n
This approach is intended to help analysts assess investigative hypotheses by making relationships between chemical substances and potential production routes easier to examine. Its emphasis on interpretable outputs connected the contribution with the workshop\u2019s wider interest in trustworthy AI and human oversight.<\/p>\n
Participation placed ARIEN\u2019s research alongside work examining both the capabilities and limitations of AI in security applications. It extended scientific dissemination and connected a concrete drug-investigation use case with broader questions about how analytical systems can support informed human decisions.<\/p>\n