Experts from the university’s Centre for Argument Technology are to develop software capable of detecting and disguising trademark linguistic patterns used by individuals online.

The team, located within the university’s School of Science and Engineering, have been awarded $2.5 million of funding as part of a larger project consortium led by SRI International in California. The project is funded by the Intelligence Advanced Research Projects Activity (IARPA), the research and development arm of the United States Government’s Office of the Director of National Intelligence.

Professor Chris Reed, who is leading the research team, said: “Whether we realise it or not, every person has patterns within the ways they engage in dialogue that, over a body of work, can reveal their identity. This is what we refer to as a dialogical fingerprint.

“The way a person interacts using language can also provide clues as to their identity. For example, a person may tend to answer questions or deliberately avoid doing so. If they do answer, do they expand on a simple yes or no with evidence and explanation, or do they tend to quickly move on. All these attributes can feed into a wider linguistic profile.

“People around the world take significant risks in highlighting injustice. By being able to identify an individual’s dialogical fingerprint it becomes possible to develop the means to remove tell-tale traits, protecting their privacy and allowing them to share their experiences in safety.”

The Dundee research forms part of IARPA’s Human Interpretable Attribution of Text Using Underlying Structure (HIATUS) program, which aims to advance human language technology. The goals of the initiative are to help protect the identities of authors who could be endangered for speaking out, as well as developing means of identifying counterintelligence risks.

The Dundee team will utilise cutting-edge artificial intelligence technology throughout the project, processing dialogue models dating back centuries to develop a complete understanding of linguistic patterns.

“It’s fun to be working with deep-learning models that are at the vanguard of AI technology and linguistic processing and combining these with theories and models from the era of Aristotle,” added Professor Reed.

“We already have practical experience of working at the interface of philosophy, linguistics and AI and are excited about harnessing this understanding to resolve modern, real-world problems.”