AI-based recommender systems for sustainable behavior
Tailoring nudges for a sustainable life
Supporting people leading more sustainable lifestyles is becoming increasingly important, so the question becomes: how can we use the power of artificial intelligence to make sure we are influencing people exactly the way they need?
Nudging as a tool
As Therese Wernstedt explained in a previous post, nudging, the process of nudging someone towards a behaviour without restricting their freedom, can be a powerful tool in supporting people to live more sustainable lifestyles. Of course, not all nudges will affect everyone in the same way - pushing a vegan to eat more vegan food may not produce the desired results; similarly, asking someone who doesn't like walking or biking to walk or take the bike to work may not be very helpful. Different themes, nudge methods, tone and even timing can affect a person's response to a particular nudge. This is where we can use AI-based recommender systems to make sure that content is tailored to specific users.
Recommender systems in today's society
Recommender systems are ubiquitous in today's society, determining everything from what content people consume on social media platforms to what ads or special offers they are shown when shopping online. Various methods and technologies can be used to create such systems; Thompson sampling was proposed as early as the 1930s to address the so-called exploration-exploitation problem, but only found practical use in the early 00s. Today, the most powerful systems deployed by giants such as Google or Meta may instead use Reinforcement Learning models based on Deep Neural Networks that require staggering amounts of data, but simpler models can still produce powerful results, and with much more modest amounts of data.
Exploring algorithm
One way to personalise content is to segment users and nudges in a number of dimensions; for users, this could be things like age group, environmental activity, or vegetarianism. For nudges, it might be the tone, the subject matter, or the nudging technique used. Based on this, similarities and differences between different users and nudges can be calculated. The next step is for the algorithm to start exploring - if a user responds well to a nudge, then similar users are likely to respond well to similar nudges. Over time, as the algorithm learns the response of more users to more nudges, it will develop an idea of which users like which nudges.
A force for good
As digital technology advances, the potential for using these powerful technologies in combination with behavioral science techniques such as nudges will grow exponentially. While it's easy to get carried away with all the exploitative ways these technologies can be used for profit, it's important to remember that they can also be a force for good.
