Design and development of a motivational conversational agent for brain injury rehabilitation

Author: Judith Hocking

Hocking, Judith, 2022 Design and development of a motivational conversational agent for brain injury rehabilitation, Flinders University, College of Nursing and Health Sciences

Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact with the details.


Adults with traumatic brain injury (TBI) can experience many symptoms including functional impairments, decreased memory, and low motivation. Recovery following TBI is possible due to the physiological mechanism of neuroplasticity – in which the brain is able to form new neuronal connections in response to rehabilitative training. Rehabilitation following TBI is supported through multi-disciplinary rehabilitation, in which client-centred goals are set and then pursued. However, rehabilitation care is time limited. Novel care approaches which leverage service provision and are motivational for clients, are needed.

Conversational Agents (CAs) provide a personal, human-computer conversation interface and can be designed to engage the user in a focused task with motivational content. Specific motivational behaviour change approaches can be applied to CA design: Motivational Interviewing (MI) has been integrated into CAs for users without cognitive impairment; and Self-Determination Theory (SDT) has been recommended for human-centred design for digital technologies including CAs. Additionally, both MI and SDT have been recommended for brain injury rehabilitation.

This thesis outlines the development of a motivational embodied CA (ECA) for brain injury rehabilitation. This type of ECA for this purpose has not previously been developed. Key considerations for developing the ECA were: addressing the clinical needs of clients with TBI: contextualizing the ECA to the clinical setting, and incorporating SDT and MI within the conversation dialogues. Living Laboratory design methodology was utilised for this project: including co-design with clinicians and clients and testing of the ECA in the real-life setting (two ambulatory care clinics, and the client’s home environment).

The three phases of development of the ECA – called RehabChat – were conducted to optimise the feasibility and acceptability of this ECA. The phases were in-house development and testing, co-design workshops, and a mixed methods feasibility pilot trial. The initial ECA prototype was developed in-house, and then tested. Testing incorporated alpha testing to check thoroughly for glitches as well as general usability; and beta testing to detect any glitches and appraise usability more closely. For the co-design workshops and the feasibility pilot trial, full ethics approval was gained, and clients and clinicians of the collaborating clinics were recruited.

The co-design workshops comprised three cohorts (current clients, discharged clients and clinicians) and four rounds of co-design workshops. Separate meetings were conducted for each cohort for the first three rounds of meetings, followed by one final fourth workshop meeting comprising all cohorts. Iterative changes were made to the ECA during these workshops. The ECA was then thoroughly checked by completing second alpha testing. The refined stable model ECA was then used in the feasibility pilot trial.

For the mixed methods feasibility pilot trial, client-clinician dyads completed a two-week intervention using RehabChat alongside usual rehabilitation care. As well, clinician-participants, for whom no clients could be recruited, participated in a mock client-clinician session to use the ECA and then provide feedback. Results which revealed that participants thought RehabChat was motivational and that it was easy to use alongside usual care, are discussed and recommendations for future research and development of RehabChat are presented.

Keywords: brain injury, rehabilitation, conversational agent, virtual human, co-design, digital health

Subject: Health Sciences thesis

Thesis type: Doctor of Philosophy
Completed: 2022
School: College of Nursing and Health Sciences
Supervisor: Associate Professor Belinda Lange