Case Study
Research project: A framework for establishing situated and generalizable models of users in intelligent virtual agents
Casper Hyllested, Dr Christoph Scheepers & Koa Health:
Background: Koa Health’s team of researchers, data scientists, clinicians & engineers collaborate with leading institutions, like the Social AI CDT, to build evidence-based prediction, prevention and treatment solutions that meet users’ mental health needs end-to-end.
Digital health and personalisation of medicine are becoming ever more important aspects of health treatment regimes. It is Koa Health’s objective to transform newly obtained knowledge as well as insights from decades of scientific research into cutting-edge, evidence-based solutions for health.
This research project investigates whether intelligent virtual agents are most effective and accepted when they adapt themselves to individual users, e.g., via development of an effective model of a user’s traits which will anticipate dynamically varying states of these traits as situations change.
Generally, this work will increase our ability to construct and understand user models that virtual agents can employ, thereby (i) advancing the development of intelligent social agents, (ii) contributing to a deeper understanding of traditional assessment instruments, and (iii) enable more adaptive situated measurement approaches across multiple scientific domains. In an applied context, the framework, methods, and app prototypes will be of potential use to clinicians and individuals interested in understanding both functional and dysfunctional health behaviours.
Aleksandar Matic (Koa Health, R&D Director)
“PhD co-supervision with the Social AI CDT has allowed our company to extend our own research capabilities as their unique mix of research expertise is so complementary to our own, specifically, the application of social AI to health-related areas with the potential to create more user-centred and personalised experiences. We also anticipate that this collaboration will provide us with direct access to young researchers with a highly relevant, multi-disciplinary skill set allowing us to grow and innovate in the future, which as an ambitious start-up, offers us a recruitment edge and is a real benefit of collaborating with the Social AI CDT.’’
Dr Christoph Scheepers (School of Psychology & Neuroscience, University of Glasgow)
“Our collaboration with Koa Health is an excellent example of the benefits of industry and academia working together, both in terms of developing more effective evidence-based solutions for mental health problems and in terms of advancing our fundamental understanding of the dynamic interplay between person-specific traits on the one hand and situation-specific constraints on the other.”
Casper Hyllested is in the 3rd year of his PhD with the Social AI CDT.
‘‘My research focuses on understanding, deconstructing & measuring the individual differences in people and data. The initial aim is to implement this understanding to create both generalizable and individualized user models, allowing virtual agents to adapt to the dynamic state and trait profiles that can influence a person’s behaviours and responses.
Having both an academic and industrial supervisor gives this research a very strong foundation. It can draw upon the vast stores of knowledge and research available at the University of Glasgow whilst also benefitting from the streamlined data collection infrastructure and niche expertise at Koa Health. This allows me the privilege of contributing to the theoretical knowledge within my field, whilst remaining focussed on how to best apply that knowledge to help people. Achieving this balance of applied function and research understanding has been made possible through the collaborative nature of the project.’’
Learn more about Casper’s research: ‘A framework for establishing situated and generalizable models of users in intelligent virtual agents’
Interested in exploring how social AI research can help innovate your organisation?
Email us at social-cdt@glasgow.ac.uk or complete our contact form.