E-Thesis 691 views 603 downloads
Designing Novel Approaches to Personalise Behaviour Change in Intelligent Systems / DARREN SCOTT
Swansea University Author: DARREN SCOTT
DOI (Published version): 10.23889/SUthesis.65829
Abstract
AI personalisation presents a promising source of innovation for improving the quality of behaviour change technologies. Current approaches are limited in their success, and a proposed solution is the inclusion of intelligent tailoring to best align users with their interventions. This thesis presen...
| Published: |
Swansea, Wales, UK
2024
|
|---|---|
| Institution: | Swansea University |
| Degree level: | Doctoral |
| Degree name: | Ph.D |
| Supervisor: | Roach, Matt J. ; Wacharamanotham, Chat |
| URI: | https://cronfa.swan.ac.uk/Record/cronfa65829 |
| Abstract: |
AI personalisation presents a promising source of innovation for improving the quality of behaviour change technologies. Current approaches are limited in their success, and a proposed solution is the inclusion of intelligent tailoring to best align users with their interventions. This thesis presents three key contributions that explore this promise: A classification system and accompanying survey to examine the current research landscape of intelligent personalisation; The E˙ect-Led Design process which combines high-eÿcacy, limitless expert design concepts with focused user discussion and refinement to best explore how to implement high eÿcacy AI that is acceptable to users; and a conceptual framework, the principles of which are tested in real-world situations to examine whether the intelligent algorithms are able to learn human behaviour and whether proposed systems of personalisation encourage motivation in users. The survey paper identified current trends in the contemporary personalised technology space and explored where the scope for innovation sits. E˙ect-Led Design showed promise in developing significantly di˙erent design concepts to those seen in contemporary applications, and both experts and users commented positively on the process. The studies testing the principles of the experimental platform showed the approaches were positively received by users in terms of motivation and engagement. However, initial implementation issues meant that algorithms did not return any significant evidence of learning. Further explorations into the algorithm through simulated studies using real-world data uncovered alterations that enabled learning. These combined outcomes provided a means to better explore the inclusion of AI in the digital intervention space, with a dedicated design process and investigation of the feasibility of a conceptual framework in this domain showing both the current potential of such a system and where future work can push these ideas to provoke e˙ective behaviour change. |
|---|---|
| Keywords: |
Behaviour change, machine learning, personalisation, physical activity |
| College: |
Faculty of Science and Engineering |