Researchers at the University of Huddersfield have developed a pioneering artificial intelligence tool designed to help investigators identify patterns of psychological abuse hidden within vast amounts of digital evidence.
The innovation, created as part of PhD research by Dhruv Patel under the supervision of Dr Anju Johnson, aims to tackle a growing challenge in modern forensic investigations — detecting coercive control and other forms of non-physical abuse embedded in messages, emails and chat logs.
Unlike physical violence, psychological abuse often unfolds over time through subtle and cumulative behaviours such as manipulation, isolation and gaslighting. These patterns can be difficult to evidence using traditional forensic methods, which typically rely on keyword searches and may overlook context.
To address this, the Huddersfield team developed the Digital Conversation Analysis Pipeline (DCAP), a hybrid AI framework that combines rule-based forensic searches with advanced deep learning to analyse language and behaviour more effectively.
The system is designed to detect linguistic indicators linked to abusive dynamics, including traits associated with Narcissistic Personality Disorder, such as lack of empathy and entitlement. By identifying these markers across conversations, the tool helps build a clearer picture of coercive and controlling behaviour.
Crucially, DCAP operates as a “human-in-the-loop” system, meaning investigators remain central to decision-making. Rather than relying on opaque algorithms, the tool highlights specific pieces of evidence, ensuring transparency and accountability in how conclusions are reached.
In testing, the system demonstrated its potential to significantly reduce investigative workload. In a simulated case involving more than 8,400 messages, DCAP narrowed the dataset to just 287 key messages — cutting manual review time by over 90 per cent while surfacing the most relevant evidence.
Researchers say the tool could have far-reaching implications not only for digital forensics, but also for mental health research and the wider justice system, where proving patterns of coercive control remains a complex challenge.
Mr Patel said the increasing volume of digital communication has made it harder for human investigators to identify long-term behavioural patterns.
“The sheer scale of digital data means subtle indicators of psychological abuse can easily be missed,” he said. “By combining textual analysis with emotional and behavioural insights, we are moving towards a more comprehensive approach that captures nuance and context.”
Dr Johnson added that the research represents a step forward in bridging the gap between raw data and actionable evidence.
“This work provides a scientifically robust framework for understanding hidden abuse,” she said. “It offers investigators and the courts a transparent and objective way to analyse complex behavioural patterns, ensuring that victims’ experiences are supported by credible, data-driven evidence.”
The research has been published in Forensic Science International: Digital Investigation and IEEE Access, marking a significant milestone in the use of AI to support justice and safeguarding outcomes.
