Will Artificial Intelligence (AI) improve the effectiveness of behaviour change interventions?

By: Corneel Vandelanotte, a Professorial Research Fellow at the Central Queensland University in Australia

Ever since my PhD I have done research examining how we can make web-based interventions more effective (1). Specifically I have been looking at the use of computer-tailoring. A tailored intervention provides participants with personally relevant feedback about a health behaviour and its correlates based on one or more brief online surveys. The feedback is provided immediately and based on so called ‘If-Then’ algorithms (IF this condition applies, THEN provide that feedback). There’s plenty of evidence that computer-tailored interventions are more effective in improving health behaviour than generic information, but there’s always room for improvement, and most intervention effects fade reasonably quick. [2]

So in a quest to improve interventions that provide personalised information many researchers, including myself, have gone the extra mile. Some researchers have examined the effectives of providing personally relevant information based on culturally and ethnically appropriate variables in addition to the more traditional correlates of behaviour change [3]. Others have sought to incorporate personalised information about the local environment [4] and I’m currently conducting a study to see if providing the health information as a personalised video works better compared to just text on a web-page [5]. There’s also the option to make the interventions ever more comprehensive, with more modules of personalised feedback provided over a longer period of time, and the inclusion of progress or iterative feedback over time. While a lot of progress has been made, it’s not clear how much is enough.

At some point tailoring to even more variables, and providing information that is even more personally relevant won’t make any difference. Beyond a certain point any extra effort made to improve the quality of the intervention will not result in improved effectiveness. We just don’t know where this point is at this stage (are we overdoing it already?). Our current interventions are already pretty comprehensive and developing even more comprehensive interventions is becoming every more time and resource consuming, which hinders translation into the real world. So this begs the question: What has AI to offer?

Can the use of AI provide us with interventions that are less difficult to develop (because of ‘smart software’ with the potential to adapt over time in response to the individuals’ changing behaviour and environment), but more effective?  Or is it actually harder to develop evolving interventions without them becoming awkward or creepy over time (how do you keep the software in line so it does what it needs to do and doesn’t provide inappropriate feedback)? Or have we simply already reached the point of saturation with our current computer-tailored interventions, so that the adaptive or just-in-time element offered by AI won’t make much of a difference? What do you think?

BIO:  Corneel Vandelanotte is a Professorial Research Fellow and Heart Foundation funded Future Leader Fellow at the Central Queensland University in Australia. His research takes a population-based approach to health behaviour change, with the aim of effectively reaching large populations at an acceptable cost. Specifically, his research is focused on developing and evaluating web- and app-based behaviour change interventions.

Join the discussion about this post on our LinkedIn group.

Want to be a guest blogger? Please email Aiste for more information.

References:

1. Vandelanotte C, De Bourdeaudhuij I, Sallis J, Spittaels H, Brug J (2005). Efficacy of sequential or simultaneous interactive computer-tailored interventions for increasing physical activity and decreasing fat intake. Annals of Behavioral Medicine, 29 (2):138-146.

2. Broekhuizen K, Kroeze W, van Poppel M, Oenema A, Brug J. (2012) A systematic review of randomized controlled trials on the effectiveness of computer-tailored physical activity and dietary behavior promotion programs: an Update. Annals of Behavioural Medicine, 44:259-286.

3. Romeike K, Lechner L, De Vries H, Oenema A. (2016). Development of a computer-tailored nutrition and physical activity intervention for lower-educated women of Dutch, Turkish and Moroccan origin using content matching and ethnic identity tailoring. BMC Public Health, 16:924.

4. Springvloet L, Lechner L, De Vries H, Oenema A (2015). Long-term efficacy of a Web-based computer-tailored nutrition education intervention for adults including cognitive and environmental feedback: A randomized controlled trial. BMC Public Health, 15:372.

5. Vandelanotte C, Short C, Plotnikoff R, Hooker C, Canoy, D, Rebar A, Alley S, Schoeppe S, Mummery WK, Duncan MJ (2015). TaylorActive – Examining the effectiveness of web-based personally-tailored videos to increase physical activity: a randomised controlled trial protocol. BMC Public Health, 15:1020.