Photo credit: Chris Scott
These reflections were written by Daniel Shorr, DR-NLP Business Development Manager
On 3 June 2026, more than 120 doctoral students, academic colleagues and external partners gathered at the Edinburgh Futures Institute for the annual UKRI Centres for Doctoral Training in NLP Partner Summit.
The Summit brings together the University of Edinburgh’s CDT in Designing Responsible Natural Language Processing and the preceding CDT in Natural Language Processing. It creates space for students to share emerging research, partners to bring forward real organisational challenges, and researchers and practitioners to examine how language technologies can be developed and used responsibly.
This year, we also welcomed representatives from beyond our established partner network, broadening the discussion across responsible AI, language technologies, ethics, standards and governance.
Photo credit: Chris Scott
Complementary perspectives on the future of NLP
Keynotes from Professor Verena Rieser, Research Leader at Google DeepMind and a University of Edinburgh alumna, and Professor Dirk Hovy, Dean of Digital Transformation and AI at Bocconi University, provided complementary bookends to the day.
Professor Rieser explored the ethical and technical challenge of balancing duty of care with the development of genuinely useful AI systems. At first glance, alignment can appear straightforward: build systems that are respectful, safe and helpful. In practice, difficult choices must be made about who systems are being built for, which goals they should serve and how competing values should be reconciled.
Professor Hovy examined the significant limitations—and opportunities—within current NLP and language-model development practices. Human experience is diverse, socially situated and often contradictory. Individuals themselves hold ideas and preferences that evolve across time and context, while human learning draws on far richer forms of experience and feedback than current models can readily capture.
Together, the keynotes reinforced the importance of globally inclusive and multidisciplinary collaboration, including sustained engagement with the social sciences, in addressing the next generation of NLP research questions.
Photo credit: Chris Scott
Connecting doctoral research with practice
Students shared their experiences of working on live organisational challenges through the Designing Responsible NLP case studies course. Projects developed with the London Stock Exchange Group (LSEG), Publishing Scotland and Mentor360 demonstrated the value—and complexity—of applying interdisciplinary research methods to practical questions shaped by organisational needs.
Delegates also engaged with an interactive student-led project examining whether people can distinguish authentic media from AI-generated speech and video. Developed initially for the Edinburgh Science Festival, the activity prompted discussion around deepfakes, misinformation and the growing difficulty of judging the authenticity of digital content.
Across parallel sessions, students and invited contributors explored explainability, contestability and standards for trustworthy AI; the potential consequences of modifying or withdrawing AI companion services; the risks of continually learning and highly personalised models; public understanding of large language models; and participatory approaches to AI in health and wellbeing.
Work involving the British Standards Institution (BSI) highlighted how emerging research questions can connect with the development of practical standards and governance frameworks.
A further strand examined abstraction and representation across language, cognition and artificial intelligence, reflecting the breadth of disciplines needed to understand both the capabilities and the limitations of contemporary language technologies.
Photo credit: Chris Scott
Sharing research and shaping new questions
Throughout the day, students from both CDTs presented research posters spanning the technical foundations of language models, their societal implications, real-world applications and governance.
The poster session enabled partners and academic colleagues to engage directly with researchers, offer perspectives from their own sectors and seeded possibilities for placements, collaborative projects and future knowledge exchange.
A Partner Challenge session placed questions from participating organisations at the centre of the discussion. Students, researchers and partners considered where doctoral research could contribute to emerging organisational needs, while also examining the practical constraints and uncertainties involved in deploying language technologies.
Professor Ivan Titov, Director of the CDT in Natural Language Processing, closed the formal programme before conversations continued during the reception and final poster session.
We are grateful to Harvey Nash for supporting the reception, and to all the students, speakers, moderators, partners and colleagues whose contributions made the Summit possible.
The breadth of the discussions—from alignment and human wellbeing to standards, deepfakes, personalisation and the social foundations of language—demonstrated both how rapidly the field is developing and how much remains unresolved.
We look forward to building on the relationships and research questions that emerged from the day.
Click here to see write-up from the 2025 Summit.
