Design Informatics to lead new centre focused on designing responsible natural language processing applications
The Institute for Design Informatics are delighted to announce that we will be leading a new Centre for Doctoral Training that is part of UKRI’s new investment in developing the next generation of AI researchers and innovators. The new CDT will establish an advanced and interdisciplinary training programme in Designing Responsible Natural Language Processing, and will train 50+ doctoral graduates over the next decade.
The CDT focuses on Natural Language Processing (NLP), a specific area of Artificial Intelligence that underpins many of the applications that are public facing and are currently the subject of public debate around AI. This includes personal assistants (like Siri, Alexa), chatbots (like ChatGPT) and “Generative AI” applications that use text prompts to generate other media (like Midjourney). These technologies offer huge opportunities to enable new ways of working and new ways of access services and knowledge, but are also leading to significant public concerns about the harms these systems may be creating.
Understanding how complex AI systems that utilise natural language, such as large language models (LLMs), operate in the world is critical to ensuring responsible, sustainable, and socially sensitive design. This requires working beyond the technical silos that much NLP development currently occurs in:
“If we are to avoid common LLM issues such as bias, systems of the future must be built by multidisciplinary, diverse teams who understand the complexities of responsibly developing, deploying and overseeing these systems in real-world settings. Our students will be trained across the technical, social, design and legal aspects of these systems, and most critically how to work collaborate across disciplines in teams of hybrid expertise.” – Ewa Luger, CDT Co-Director
A unique aspect of the training programme will be the integration of expertise across informatics and computer science, design and human computer interaction, language and speech science, and law and regulation. Core to the training our students will receive is engagement with how the law around AI, the technical development and implementation of NLP, and responsible design, are interconnected to one-another:
“Novel NLP technologies are raising difficult legal and design challenges as they are deployed in the real world. Generative AI raises new questions for privacy, copyright, human rights, and design processes. Emerging European legislation is changing the landscape too, bringing formal requirements around how to manage these risks and build more trustworthy AI. It is critical we train the next generation of interdisciplinary researchers and innovators who can navigate the boundaries between complex socio-technical risks, design choices, and legal issues, in what is a very fast-moving domain.” – Lachlan Urquhart, CDT Law Lead
The students will be physically co-located together in the Edinburgh Futures Institute (EFI), the University of Edinburgh’s brand new hub for research, innovation and education around socially just data and AI futures:
“The interdisciplinary, critical and action-oriented environment of the new Edinburgh Futures Institute aligns perfectly with the ambitions of this CDT and the values of the Institute for Design Informatics as a whole. We recognise that NLP, and AI in general, are technologies that have thus far being deployed-on citizens and the public, rather than designed with and by them. A core part of our training is making use of EFI’s postgraduate courses that attend to how issues of inclusion and exclusion are baked into the history of AI. We will also be introducing students to state-of-the-art participatory and co-creative methodologies so they are ready to involve diverse users in the earliest stages of application design.” – John Vines, CDT Director
The CDT also brings together a community of 71 industrial, public sector, charity and international research partners – recognising both the excitement and intrigue of NLP systems, but also the rapid realisation that there is a pressing need for skilled researchers and practitioners that understand the potential risks and harms of these systems when deployed in the world. As well as being directly involved in training by setting out contexts and use cases for students to work on, our partners are committed to initiating a community of practice around responsible and trustworthy NLP. They will also be able to benefit from online courses based on the materials we train our students with, and host students for placements and team-based projects.
The CDT will be Directed by John Vines (IDI, School of Informatics), with a team of co-investigators across the School’s that form the CDT: Ewa Luger (IDI, Edinburgh College of Art), Mirella Lapata (School of Informatics), Frank Keller (School of Informatics), Nicola Osborne (IDI, School of Informatics and Edinburgh College of Art), Melissa Terras (IDI, Edinburgh College of Art), Catherine Lai (School of Psychology, Philosophy and Language Sciences) and Lachlan Urquhart (IDI, Edinburgh Law School).
The CDT will be recruiting its first cohort of students for a start in September 2024. There will be further announcements made in December 2023 / January 2024 about this, as well as a new website specific for the CDT.
In the meantime, if you have any questions get in touch with John Vines in the first instance: john.vines@ed.ac.uk