Hybrid Authorship and the Limits of Detection

Speaker: Dr. Sheila Queralt

Abstract: This talk presents ongoing research from the CorpIdentIA project, which critically evaluates the performance and reliability of AI text-detection tools in forensic contexts. The project is structured in multiple phases combining experimental and analytical approaches, including corpus design across different authorship modalities (human, AI-generated, and hybrid texts), systematic testing of detection systems, and qualitative linguistic analysis of classification errors. Particular attention is given to hybrid and manipulated texts (e.g. paraphrased or humanised AI outputs), which pose significant challenges to current detection technologies. Preliminary findings reveal important limitations in these tools. By integrating quantitative performance metrics with linguistic analysis, the project aims to assess the evidential value and forensic applicability of AI-detection systems. The talk will outline the research design, present ongoing results, and discuss implications for forensic practice.

Event Venue & Nearby Stays

Aston University, Aston Street, Birmingham, United Kingdom

Concerts, fests, parties, meetups - all the happenings, one place.

Get App