The idea of using nudges to drive behaviour change is by no means new, however, when OpenAI gets involved then it suggests we should take another look at what may be possible with state-of-the-art technology. OpenAI's startup fund is partnering with Thrive Global and they are focusing on health-related behaviours such as sleep, fitness, and stress management. That’s all very well, however, the big win for organizations will be if we can take this behaviour change technology and use it to improve compliance with safety regulations, better run meetings, better time management, and so on.
How nudges work with AI
The idea of using nudges to drive behaviour change contrasts with the idea of relying on a one-off training session. We can give people a workshop on healthy eating, but it’s more effective if we can have the smartphone send brief messages to encourage healthy choices at the moment when someone is shopping for food.
The point of AI is to personalize the nudges so that they are more effective. That’s what OpenAI is hoping to do with Thrive. Now, it’s true that even without AI nudges can be personalized. Fitness apps allow you to set personal goals and control how often you are nudged. What AI offers is the possibility of even more personalization.
What might personalization look like?
One of the annoying things in HR is how often vendors talk about “personalization” without being clear about exactly what will be different. In this case, it’s up to us to speculate what kind of personalization might be effective in driving the kind of behaviour change we want in a business. Imagine we wanted managers to do a better job of leading meetings. An AI-based app that understood the different types of meetings a manager had could give more appropriate nudges than one that gave generic advice before each meeting.
Are existing nudges effective?
There is a fair amount of academic research on nudges. For example, a study by A. Anro et al reports that “nudging can influence health behaviours like smoking, overeating, and poor adherence to guidelines”. That sounds pretty good but they go on to say “Nudge theory strategies can influence adult dietary behaviour, resulting in a 15.3% increase in healthier dietary choices”. Fifteen percent is not nothing but when we dig into the quantitative impact we recognize that we need to be realistic in what we expect from nudging. (Source: Arno, A., Thomas, S. The efficacy of nudge theory strategies in influencing adult dietary behaviour: a systematic review and meta-analysis. BMC Public Health 16, 676 (2016)
A more business-relevant example comes from H. van Roekel et al; they report nudges can decrease email use, but again we need to look at the quantitative data and the decrease in email use was 7%. Probably a good thing but not as dramatic as we might hope. If the cost of the intervention is reasonable, and this reduction in email affects many employees over many years, then it could have a high ROI. So yes, nudges work with current technology, though not always as well as we would hope. Perhaps AI-based personalization of nudges will take the impact on behaviour to a new level. (Source: van Roekel H, Giurge LM, Schott C, Tummers L. Nudges can be both autonomy-preserving and effective: evidence from a survey and quasi-field experiment. Behavioural Public Policy. Published online 2023:1-24. doi:10.1017/bpp.2023.18.)
Is this the future of behaviour change?
AIs are getting smart enough that they can begin to mimic some of what a personal coach might do to drive behaviour change. We shouldn’t expect miracles, people often find it very difficult to change their behaviour even with the help of a coach. By the same token, even relatively dumb nudging tools, such as reporting the number of steps someone has taken that day, can indeed affect behaviour. I think it’s a safe bet that these nudging technologies will continue to improve and evolve to address more types of behaviours. Let’s keep an eye on the OpenAI-Thrive experiment (as well as others AI behaviour change experiments such as www.slingshot.xyz) and see if that gives us insight into where AI-based tools might become a game-changer for HR.