Behaviour change and AI: panacea or plain creepy?

AI holds both great potential and risk for the marketing industry. But little has been discussed about using it to strengthen behavioural change interventions. The Human Behaviour Change Project run by UCL is creating an AI to scan and analyse the world’s literature on behaviour change to answer the big question: which interventions work, how well and why? But beyond this academic use of AI, there is potential for using AI in marketing communications to either scale or hyper-personalise nudges.

There are a few ways this could manifest itself:

1. Combining the predictive modelling of AI with the proven ability of behavioural science to prompt behaviour and overcome the ‘last mile problem’.

One of the powers of AI is in predictive analytics. It can rapidly identify prospects, be those new customer targets, those at risk, or those most likely to change their behaviour. But this is where data science hits the ‘last mile problem’ – it can identify who to target, but it cannot alone suggest how to prompt the desired behaviour change. This is where behavioural tactics could come into play and compliment the AI modelling. Proven behavioural change techniques could then be employed to encourage people towards the desired behaviour.

2. AI-driven long-tail of communications to keep nudging people towards their chosen goal.

If someone is trying to stop smoking by taking part in Stoptober, we can ensure that they feel supported and part of a social movement through public and personal communications and working with the right partners to ensure they are supported face-to-face. But if, despite all this they share on social media how hard they are finding it or that they are going to find tonight’s social gathering hard – how do we keep them on track?

This is where AI could help us to identify the signs and signals that shows someone is about to fall off the wagon and serve further behaviour change interventions. The AI model could learn what the best signs and signals are and learn what interventions are best applied for that type of person in that type of situation. This could enable us to both hyper-personalise nudges, while doing it at scale.

3. Tech solutions with behaviour change at the heart powered by AI.

Predictive modelling works within the realm of communications – a ‘push’ approach. But as we continue to change our relationship with technology, does a ‘pull’ approach become more appropriate?

We are moving from an age where technology is used to connect us (to our money, to our friends, to our mail) to an age where it is more focused on helping us achieve our goals. Chris Risdon, former head of behavioural design at Capital One, talks about this as moving from the ‘utility age to the augmentation age’. The closer we get to technology the more it is about us.

So instead of just giving a person access to view your money (utility), it could be about helping them achieve stated behavioural goals (augmentation). People are inviting positive influence which means we need to deliver this, not just through marketing communications but through products that help them achieve their goals. This creates a clear space for tech solutions, powered by AI, that are built around helping people towards a desired behaviour.

Examples of this include Chip, the fintech app which uses an AI algorithm to monitor your regular spending and identifies when you can save money, which it does on your behalf. The recent introduction of Open Banking in the UK will enable this type of ‘augmentation’ on a wider scale in financial services, but it isn’t just limited to this sector. People don’t have the time, money or mental effort to achieve everything they want – they are looking for brands or services to support them to achieve their goals.

But there is a darker side to combining AI and behavioural science. The process relies on being able to identify and monitor people in some way. We are seeing that people are increasingly concerned over privacy and control of their own data. Using AI to deliver direct interventions could raise awareness of how much of your personal data is out there and how, even with a little data, a lot can be inferred. This is where AI could be seen as plain creepy.

While there aren’t many AI and behaviour change examples out there now, it is an area ripe for development. If you are an organisation who wants to support people to change their behaviour and achieve their goals, then considering combining AI and behaviour change techniques should be in your development pipeline. If you can do it while overcoming data and privacy concerns, then it could well be the panacea.

Jane Asscher is chief executive and founding partner at 23red