Of all the companies claiming to achieve hyper-personalization through behavioral science and artificial intelligence, Lirio seems to be the one actually doing it. Marten den Haring, now CEO, joined Lirio as Chief Product Officer in 2019, and has spent the past 5 years building and launching the platform and its personalization engine to drive meaningful behavior change in healthcare. And with Amy Bucher at the helm as Lirio’s Chief Behavioral Officer, you can be confident that the rigor of the science is as solid as it gets.
Lirio’s platform is built on the metaphor of the Large Language Model (LLM), but rather than ingesting semantic data and predicting the next word in a sequence like generative AI models, its Large Behavior Model (LBM) draws on health behavior data to predict the next behavior in a sequence of behaviors for an individual. The result is a hyper-personalized behavioral model that provides recommendations optimized for each user.
In this interview, Marten den Haring shares how he combines behavioral science and AI to design for health behavior change, and specifically how Lirio uses behavior change techniques such as “pros and cons awareness” to achieve an outsized impact in healthcare.
So, as you know, I’m Dutch. You probably associate The Netherlands with things like flowers, cheese, football (soccer), cycling, speed skating, etc. You may even know that the Dutch are considered the tallest people in the world. What you may not know is that some pretty useful inventions were made by the Dutch, including Wi-Fi, Bluetooth, Python (the programming language), the microscope, the telescope, the submarine, the first privately owned company that issued shares, the first stock exchange, and … orange carrots.
Above all, product leaders must have insatiable curiosity. Pick up a copy of Douglas Adams’ The Hitchhiker’s Guide To The Galaxy, because it tries to inspire us to always find the right questions to ask on our journey. Pair it with Max Tegmark’s “Life 3.0”, because it forces us to ask questions about what kind of future we want to create.
At Lirio, we have placed behavioral design at the heart of our research and development effort. We leverage behavioral design in our behavioral interventions, in the training of our Large Behavior Model, in the design of randomized control trials, and in the real-world deployment of our Precision Nudging solution. The result is a company on a mission to encourage health behavior change through personalized interventions, proving that behavioral design can significantly boost engagement rates and improve health outcomes across the continuum of care.
For example, one health system asked us to engage their diabetic patients who hadn’t seen a primary care provider in several years, putting them well overdue on diabetes well visits. The health system had tried every tactic in their arsenal but was unable to engage these patients – to the extent that they believed a 2% conversion would be a worthy target.
Our behavioral design process centers on identifying barriers to key behaviors and then incorporating appropriate behavior change techniques into our intervention, following the COM-B framework. We used this behavioral design process to identify 10 broad categories of barriers to primary care for patients with diabetes and ensured that each one was addressed from multiple angles.
When we looked at which types of behavior change techniques (BCTs) were most effective at nudging patients to attend a primary care visit, many of them were designed to address barriers around awareness of the need for diabetes-specific primary care. One of the most successful nudges used the BCT of “pros and cons,” which we leveraged to emphasize the specific benefits of seeing a doctor for a diabetes well visit. Ultimately, over 60% of the disengaged patients who we reached with Precision Nudging attended at least one primary care visit (well above the health system’s 2% target!).
Behavior change is hard. We are talking about influencing humans that are not always motivated to do what is right, or experience barriers to adopting healthy behaviors. More importantly, humans themselves change over time. Building product that leverages data to digitally communicate with people to influence their decisions and actions is complex, requiring multi-disciplinary teams to collaborate, experiment, iterate and improve the product. In this context, the biggest barrier is creating an environment where we can quickly access and analyze feedback data to adapt to human behavior in near real-time.
Most of the time, behavioral insights are the gap! In our case, behavioral insights enable us to go beyond targeting static population segments with a generalized approach, to tackling behavior change at the individual level with a personalized approach. In other words, behavioral insights allow us to scale and target the n of 1 across many behaviors over time.
My hope is that the practice will continue to attract people who are passionately curious about creating great products, and represent the (diverse) populations they build product for. The skills and tools we use in our profession will continue to evolve. The scope and speed of our decision-making will continue to increase. The inherent security and privacy challenges that come with a new generation of AI and data-driven products will increase the risks associated with those decisions. It is therefore critical for product organizations to become more intentional about incorporating ethics and responsibility into their research and development processes and controls. As product leaders we must never lose sight of the fact that our products impact humans, and technology remains a means to an end to better serve humanity.
Invest in developing your human superpowers: critical thinking and empathy. After all, AI (and tech more broadly) lacks the emotional intelligence that humans possess. Embrace behavioral design thinking and use it to complement your technical skills and other subject matter expertise. Know what problem you are solving, who you are solving it for, and understand the limitations of the data you have access to at any given point in time. Build products that learn and adapt to change, fast. Enjoy the experience!