Dr Rafael Mestre

Research groups

Research interests

• Responsible Research and Innovation of emergent technologies
• Multimodal machine learning and Natural Language Processing
• Computational social science
• Bio-hybrid robotics
• Digital Humanities

Current research

Dr Mestre's current research very interdisciplinary in nature and lies at the interface of fields like computer science, science and technology studies, biohybrid robotics and political science. He is interested in the ethical and social implications of emergent technologies, as well as their public co-governance. As PI of the ESRC New Investigator award "Biohybrid Futures: A framework for research governance and application of bio-hybrid robotics", he studies responsible research and innovation of this emerging technology, which was the focus of his PhD. He is also currently studying the biases and ethical aspects of novel AI technologies like large language models (LLMs), as well as their responsible application in domains like mental health, through the RAI UK International Partnership grant RAI4MH: Exploring Fairness and Bias of Multimodal Natural Language Processing for Mental Health in collaboration with the Institute for Experiential AI @ Northeastern University (US). From January 2025, he will lead the £1.2M UKRI Cross-Research Council grant SOUNDSCALE project, investigating the responsible development of distributed fibre optic sensing for smart cities through inter- and transdisciplinaryr esearch with the physical, computer and social sciences, arts and humanities, and engagement with citizens, policymakers, nonprofits and industry.

Some of his research efforts have also focused on natural language processing (NLP) and multimodal machine learning, particularly applied, although not exclusively, to argumentation mining (the field of computer science that aims at extracting and analysing arguments from natural dialogue). In general, he has focused on the study and application of multimodal methodologies within the text-as-data, image-as-data and audio-as-data paradigms in political science research. His previous works have tackled the classification of arguments from presidential debates using audio-textual information, and he has developed datasets for this purpose.

He is also interested in the application of these multimodal machine learning techniques in computational social (political) science, from social media analysis to the study of public deliberation and parliamentary debates. He collaborates in the ‘Rebooting Democracy’ project, where he works in the development of digital tools for public deliberation based on argumentation maps and in the development of democratic innovations based on citizens’ assemblies or mini publics. He's worked in projects with impact for citizens and policy makers, such as the Southampton Climate Citizens' Assembly or public dialouges in collaboration with a UK Select Committee, where he led the digital/computational aspects, from collaborative argument mapping to topic modelling to inform citizens' perceptions.

Finally, his interdisciplinary research bridges to other fields, like biomedical engineering, environmental science or digital humanities. He works in projects detecting and classifying maritime events from distributed acoustic sensing (DAS) systems using multimodal machine learning in collaboration with the National Oceanography Institute (NOC). He is also PI of smaller projects in digital humanities around multimodal machine learning and interactivity in an immersive space and multimodal argumentation mining.

Research projects