Does Behavioural Design Really Work?

Imagine you’ve decided to change your life. Maybe you want to quit smoking, stick to a healthier diet, or finally make regular exercise a habit. Or perhaps you’re a manager, trying to foster collaboration in your team or get your employees to follow safety protocols. The promise of behaviour change—whether for individuals or entire groups—is enticing. After all, if you can change habits, you can change outcomes. And what could be more powerful than the idea that human behaviour can be managed scientifically?

This is the vision that drives much contemporary work in behaviour design.  From popular books such as Atomic Habits to the behavioural research of BJ Fogg at Stanford and Susan Michie in London the potential to change behaviour in a planned, systematic way has never seemed so close. Michie’s COM-B model, which posits that behaviour arises from an interaction of capability, opportunity, and motivation, provides a structured and research-based framework for understanding and influencing behaviour. It’s the backbone of countless public health campaigns and organizational strategies. The idea is simple yet appealing: if you can adjust the factors that drive behaviour, you can change the behaviour itself.

It’s no wonder that Michie’s approach has gained traction. It feels like science is finally taking charge of a notoriously unpredictable domain—human behaviour—and making it more manageable, more predictable, and more controllable. With enough research, data, and the right interventions, we should be able to achieve the outcomes we want. Right?

Well, not always. It turns out it's not as simple as that. As appealing as this rationalist, research-based approach is, it has its limits. In fact, when applied to unpredictable real-world environments it often falls short in ways that leave us scratching our heads. It may adopt the methods of science, but the reliability and replicability of interventions is a long way off what one can observe in the natural sciences.

When Variables Multiply and Control Slips Away

Consider a public health campaign designed to tackle obesity. On paper, Michie’s COM-B model suggests a clear path: improve people’s capability by providing nutritional education, create opportunities by making healthy food more available, and increase motivation through incentives or social campaigns. But in practice, this straightforward strategy can initially work but then unravel when faced with the complex web of factors that shape human behaviour in real life.

For one, people's food choices are rarely just about capability, opportunity, or motivation. They are influenced by deep-rooted cultural practices, social identities, economic realities, stress levels, emotional states, and even unconscious habits formed over years. Many of these variables are interconnected in ways that cannot be easily separated or controlled. For example, someone might know what’s healthy (capability) and have access to good food (opportunity), but still feel unmotivated because their family traditions prioritize high-calorie meals or they live in a community where healthy eating isn’t socially reinforced. 

In these complex real-world situations, where variables multiply beyond our ability to manage them, behaviour change interventions often stumble. This is arguably the case in many of public intervention campaigns that have adopted Michie's COM-B model. Take smoking cessation programs, like the UK’s “Stoptober.” Success might be defined by how many people quit smoking during the campaign month, or how many remain smoke-free after six months or a year. These campaigns often report encouraging results in terms of increased quit attempts and short-term cessation rates. But even here, complexity creeps in—social factors, stress levels, personal habits, and peer influences all play a role in whether a person sustains their commitment. In terms of incontrovertible success, it’s challenging to claim that a campaign like “Stoptober” produces long-term, permanent behaviour change for everyone. More often, it works for some but not others, and it may simply perform better than previous campaigns because it’s more targeted and thoughtful. So while Michie’s structured approach has improved outcomes compared to less formal initiatives, it hardly guarantees long-term success for every individual.

This is where Dave Snowden, an advocate for complexity science, offers a sobering but useful counterpoint. Snowden argues that in systems as complex as human behaviour, no amount of upfront planning can give you full control over outcomes. There’s always something unexpected that crops up—whether it’s the sudden stress of a personal crisis, a shift in social norms, or an unintended backlash from the intervention itself.  Context is king - and if your model can't cope with context, it won't cope very well in the long run with human behaviour.

Snowden’s view, which draws from complexity science, resonates with something most of us have experienced: no matter how rational and scientific we try to be, life doesn’t follow predictable patterns. Think of a time you meticulously planned for a work project or personal goal, only to have something entirely unforeseen throw your plans off course. Maybe it was an unexpected illness, a shift in office politics, or just bad timing. The point is, complex environments don’t follow the tidy logic of cause and effect that characterises simpler systems.

The Paradox of Control

There’s an even deeper issue at play here—one that takes us beyond science and into the realm of philosophy and Eastern thought. Both Zen Buddhism and Taoism suggest a paradox that flies in the face of rational planning: the harder you try to control an outcome, the more likely things are to go wrong. It’s a counterintuitive idea, but one that aligns closely with Snowden’s complexity-based approach. When we treat human behaviour like a mechanical system to be controlled and engineered, we can inadvertently introduce friction and resistance.

Take a workplace intervention aimed at boosting employee engagement. A manager might use the COM-B framework to identify barriers to engagement—perhaps employees lack the capability to collaborate effectively, or there aren’t enough opportunities for meaningful teamwork. So, the company introduces new training programs and schedules more team-building events. But instead of creating the desired boost in collaboration, employees might feel micromanaged or resistant to what they perceive as forced camaraderie. The more the manager pushes, the more disengaged the employees become. This phenomenon—where the very effort to control behaviour backfires—is something that ancient wisdom traditions have long recognized.

In Zen Buddhism, for example, it is often said that “trying to control life is like trying to grasp water.” The tighter your grip, the more it slips through your fingers. Similarly, in Taoism, the idea is to flow with the natural dynamics of a situation rather than impose control. These perspectives highlight a truth that modern complexity science echoes: human systems are not machines, and attempts to force them into predictable patterns often lead to unintended consequences.

Planning vs. Experimenting: The Snowden Approach

Does this mean that trying to change behaviour is futile? Snowden says no—but he offers a different approach. Instead of relying on detailed upfront planning, as Michie’s rationalist model might suggest, Snowden advocates for small-scale experiments in real-world contexts. The idea is not to try to predict and control every variable, but to introduce small interventions, observe how the system responds, and adapt based on what emerges. It’s an approach rooted in humility, recognizing that in complex systems, we can never have full control.

For example, rather than rolling out a large-scale public health campaign based on pre-designed models, a complexity-informed approach might start with several small, experimental interventions in different communities. The goal would be to learn quickly from what works and what doesn’t, adjusting the approach as new patterns and responses emerge. This strategy allows for flexibility in the face of the unexpected—a flexibility that is often absent in more rigid, rationalist models of behaviour change.

This doesn’t mean that structured approaches like Michie’s are useless. Far from it. In simpler environments, where the variables are fewer and more easily controlled—such as a smoking cessation program or a workplace safety initiative—the CoM-B model can be highly effective - at least initially. It works well when the problem is well-defined, and the factors influencing behaviour are relatively stable. But as the complexity of the environment increases, so too do the limitations of this approach.

A More Modest Vision of Behaviour Change

So, where does this leave us? Can we change behaviour by adopting the Behavioural Science advocated by Michie and others? Yes, but perhaps not as reliably or as universally or as sustainably as we might hope. In some cases—especially where systems are relatively simple or complicated—structured, scientific interventions work. Michie’s models provide a useful toolkit for thinking about behaviour change in these contexts, and they’ve certainly shown degrees of success in areas like public health.

Nevertheless in complex environments, where human behaviour is shaped by an unpredictable array of factors, the limits of this rationalist approach become clear. Snowden’s complexity science offers an alternative: a more experimental, context-sensitive approach that embraces uncertainty and learns from the system as it evolves. This is not as scaleable as COM-B purports to be, but it is honest about the world we often encounter. It’s not about abandoning science but rather expanding it to include models that better reflect the messy, dynamic nature of human life.

In the end, the science of behaviour change is not a failure—but it’s also not the magic bullet it sometimes promises to be. We can influence behaviour, but we cannot control it with the same precision as we might control a machine. Instead of seeking absolute control, we might do better to adopt a more modest vision of what behaviour change interventions can achieve—one that respects the complexity of human systems and approaches them with humility, flexibility, and a willingness to adapt.