How AI can (also) empower the behavioral sciences

The relationship between technology and human behavior is a two-way street: first, AI has a profound impact on how we understand human behavior, and second, behavioral sciences can help shape AI development, accuracy and adoption. BBVA’s Behavioral Economics unit is continuously exploring possible alliances between AI and behavioral economics, beyond the efficiency gains to be had from the use of tools such as ChatGPT.

A few examples of these synergies:

1. Standard labeling of cognitive and heuristic biases: AI allows us to label cognitive mechanisms present in different user experiences. This information is vital in analyzing which stimuli most guide our decisions and is an invaluable resource for designing contexts and decision architectures that shape our actions. This labeling can be used as another key input variable for behavioral segmentations or predictive algorithms of human behavior.
2. Designing decision-making architectures: AI, since its beginnings with Alan Turing and Herbert Simon (one of the fathers not only of AI but also of behavioral economics), has mimicked and been linked to the understanding of the human mind. Today, Nobel laureates like Richard Thaler have been busy exploring how AI can mitigate the errors produced by our cognitive biases and help design more effective and personalized ‘nudges’ (small changes in the context in which a decision is made) to avoid ‘one size fits all’ solutions. By harnessing the personalization capabilities offered by AI, we can make these nudges more impactful. By tailoring them, rather than making a single change for the entire population, we can also optimize their effects. This principle, which is becoming increasingly apparent in the studies carried out by international experts on the future of behavioral economics, happens to be one of the aspects underpinning the BELA project, which has been in production at the bank since 2020.
3. Growing exploitation of the scientific method: AI also enhances the scientific method, a cornerstone of the behavioral sciences. AI is able to mass generate different assumptions and hypotheses, allowing it to produce the immense amount of content needed for the hyper-personalization it promises. It can also support the industrialization of downstream analyses and help us review the vast body of academic literature, which is constantly growing and evolving.
4. New forms of research: the combination of AI and behavioral sciences is leading to innovations that allow us to simulate large-scale behaviors. Synthetic panels and archetypes (virtual study groups based on AI-generated profiles) let us carry out behavioral experiments without involving real participants, allowing us to optimize the user experience under different scenarios. For example, this would overcome a classic problem of online panels for hard-to-reach populations, such as key decision-makers within large companies or organizations who do not normally respond to this type of survey.
5. Sentiment analysis: deeper sentiment analysis is now possible, capturing emotional nuances that previously went unnoticed. Sentiment analysis through AI, whether in telephone conversations, video, chat, email complaints or even social media, allows us to understand our customers through pure observation, even on reactions, perceptions or sometimes unconscious decisions. In many cases we do not know why we make certain decisions, so the answers obtained about the reasons for a past action or future behavior are often unconsciously unreliable. By being able to obtain observed, non-declarative learning, we can design much more humane, empathetic and personalized interactions.
6. Behavioral forecasting: predictive models based on historical data and observed behavior are becoming increasingly accurate. AI lets us analyze large volumes of data and discover complex patterns that would otherwise go unnoticed. Interpreting these data from a behavioral perspective is an immense opportunity to accelerate the development of more effective predictive systems. By incorporating contextual variables and external stimuli, these models are able to anticipate future decisions, identify behavioral friction points and come up with strategies to optimize the user experience.

According to Álvaro Gaviño, from the global Behavioral Economics team at BBVA, “the development of predictive systems that simulate complex behavioral systems has a bright future.” These platforms synthesize the populations for which we want to predict behaviors, including the variables and logics that affect their decision-making. They allow us to discover non-obvious effects when the conditions of a system are modified, looking at it as a whole. “They are tools in which a multitude of stimuli and contextual variables must be considered, but which are already proving to be of immense value to companies and governments around the world,” he concludes.

At BBVA, AI is not just a tool for improving efficiency, it is our ally in getting to understand people better. The bank remains on the cutting edge when it comes to innovation, applying AI and behavioral science to build a smarter and more accessible financial future for all.