FitChat: Conversational AI for Encouraging Physical Activity in Older Adults

Recommendations are that older adults should do at least 150 minutes/week of moderate intensity physical activity and muscle strengthening activity at least twice per week (World Health Organisation). However, only 44% of adults aged over 65 meet these recommendations (Health Survey for England, 2016). As physical activity has many positive health benefits, including reducing the risk of disease, poor mental health and falls, it is important that older adults are encouraged and facilitated to remain physically active or to increase their activity levels if they are currently inactive.Presently, health applications used to monitor physical activity deliver digital behaviour change interventions as text notifications on mobile phones. Despite the popularity of this approach, there is little evidence to indicate that text notifications are effective at promoting positive behaviour change particularly in the long-term. The main problem is that text notifications offer one-way communication (from the device to the user) and hence, provide no opportunity for interaction. In addition, text notifications are easily ignored; fewer than 30% of received notifications are typically viewed by users with average delays of close to 3 hours, highlighting the need for an alternative approach.

Conversation is expected to prove particularly appropriate for delivering behaviour change interventions to older adults who can have difficulties with new technologies and may be more likely to appreciate the natural interaction offered through conversational dialogue. Hence, delivering behaviour change interventions using digital conversation provides an opportunity for achieving higher levels of adoption of and adherence to physical activity and exercises, compared to the traditional text notification approach. Outcomes of this project were published as FitChat: Conversational AI for active aging at the IntelLanG Workshop – Intelligent Information Processing and Natural Language Generation Co-located with ECAI 2020.




Prof Nirmalie Wiratunga
Nirmalie is a professor of Artificial Intelligence and has a track record leading digital health research projects (H2020, InnovateUK Cancer Challenge & SBRI IBD), most relevant here is the 5M Euro SelfBACK (H2020) project, in which she leads RGU’s human activity recognition (HAR) work package. She is an executive member of the BCS’s Specialist Group on AI, ICCBR’s Conference Chair in 2012, its Sponsor Chair in 2018 and serves on international AI programme committees (IJCAI, AAAI) and is also co-organiser of the Joint Workshop on AI in Health 2018 co-located with the premier International Joint conference in AI (IJCAI 2018)

Prof Kay Cooper
Kay’s research focuses on chronic musculoskeletal pain, human movement analysis, self-management, and physical activity. She is Director of the Scottish Centre for Evidence-based, Multi-professional Practice: A Joanna Briggs Institute Centre of Excellence, and leader of the North of Scotland hub of the Council for Allied Health Professions Research. She is experienced in the development and testing of peer-support interventions for self-management of LBP in older adults (PI: Dunhill Medical Trust R300/0513) and digital mapping of behaviour change theory and techniques. Kay also has experience of co-production and user-involvement in research, particularly with older people.

Dr Sadiq Sani
Sadiq is an experienced researcher in AI and Machine Learning. His recent work on the SelfBACK H2020 project explored algorithms and context representation for effective recognition of human activities from wearable sensor data. Sadiq has also worked on digital intervention delivery (underpinned by the behaviour change theory). Sadiq is co-chair of the annual CBR and Deep Learning workshop, which is co-located with the International Conference on Case-based Reasoning and serves as PC member and reviewer of a number of leading machine learning conferences and journals including the International Joint Conference on Neural Networks (IJCNN) and Neural Computing And Application (NCAA).

Dr Stewart Massie

Stewart has more than 15 years research experience in AI developing improved machine learning, information retrieval and data mining technologies. He also has a track record of research in smart sensors and their application in healthcare and other areas. Recent relevant funded projects include: FITsense, a health application that analyses data from static in-home sensors to underpin evidence based alerts allowing preventative intervention before incidents occur, and SmartBeacons, tourism applications that improve visitor interaction with objects by providing personalised recommendations. He has over 50 peer-reviewed publications in leading journals and conferences and serves as PC member for several international conferences