On Groups and Individuals
this is and isn't about behavioural science and policy
merry christmas. the next week is the best period of the year for reading, watching and drinking. proceed accordingly.
I read today that early 2026 will see the publication of the 2.0 version of the COM-B model of behaviour change.
Big news for behavioural science. This is the biggest model in the discipline. And if you’re an outsider, it’s a pretty solid example of how all frameworks used in behavioural interventions and policy tend to work.
This should be useful because I guarantee that as you sit there right now, you have at least one thing you’d massively like to change about yourself, and might even know how, but change never comes.
The eternal question is this: why don’t people, including you and I, do the things that are good for them? That they want to see themselves doing? And this framework replies…
C, O, M = Capability, Opportunity, Motivation. These are the three most important factors for determining whether or not a (B =) Behaviour happens. Whether you have the skills to do the thing, the chance to use those skills, and the desire to act when the chance arises.
You’re in a Department of Education that is funding free after school tutoring to underprivileged kids. Two years later their outcomes haven’t improved. Fire the tutors? No, turns out the underprivileged kids just aren’t turning up.
They have the opportunity to get tutoring, and their parents will (probably) have the motivation to take the government up on the offer. What gives? You go down to a local area, meet some parents at the school gates, and you work out where Capability has gone astray.
The parents don’t know what programme you’re talking about. No one told them. To be fair, you know they were sent a letter in the post. But now you realise the parent opposite you is unlikely to have the time and energy to read through all their mail. Or worse than that, they’ve developed a psychological aversion to official looking letters because the news is going to be bad more often than not.
Having the right information is key to being able to do the thing, so you COM-B your way out of it by approaching in parents in person (in a way that doesn’t publicly signal their SES to everyone else, i don’t know, just an example), and their kids actually go to the free tutoring this time. Two years later, their outcomes improve.
By focusing on individual-level decision environments, we meaningfully affect our population-level issues.
Awesome, shame it doesn’t really work.
Such is the implication of this paper from Nick Chater and George Loewenstein:
Individual-based interventions alone are likely to be insufficient to deal with the myriad problems facing humanity. Indeed, disappointingly often they yield small or null results. DellaVigna and Linos (2022) analyze all the trials run by two large US nudge units: 126 randomized controlled trials (RCTs) covering 23 million people. Although the average impact of nudges reported in academic journals is large– at 8.7%– their analysis yielded a mean impact of just 1.4%.
Why the difference? They conclude that selective publication in academic journals explains about 70% of the discrepancy.
DellaVigna and Linos also surveyed nudge practitioners and academics, to predict the effect sizes their evaluation would uncover. Practitioners were far more pessimistic, and realistic, than academics, presumably because of their direct experience with nudge interventions.
Selective publication covering up the reality that behavioural interventions barely ever work (and when they do, they don’t change much) matches my experience. Try your best to help people to do the things they claim to want and you’ll probably come up short. Null is the norm.
The question becomes what you do with that information.
Chater and Loewenstein conclude that if you want to change the outcomes of the world for the better, you’re better off focusing on the system instead. Its power is way too much for individuals to fight against:
Improving individual decision making has been the focus of i-frame behavioral insights. But often the most powerful way to help people make better decision is not merely to modify their “choice architecture,” but to fundamentally change the “rules of the game.”
Thus, eliminating conflicts of interest between professionals and their clients (e.g., in medicine or finance) is likely to be more effective than requiring disclosure, or educating consumers to detect potential conflicts.
Systems create outcomes that work for no one and the co-ordination required to fix them doesn’t naturally emerge from within. This is internet rationalism 101. Chater and Loewenstein say we should focus on that as the big problem, or at least as the channel of solution that offers the most alpha.
As far as I can understand the details they’ve released already, this relates to the primary justification of altering the COM-B model, so that the 2.0 version can have a framework for understanding both individual and group level behaviour change. Because this is how policy actually works and we should have been doing it this way in the first place. You make a big change to the environment (i.e. the rules of the game), if 20% of the population make a resulting behaviour change, maybe that’s all you need for it to be worth it.
Example: The Mayor of London wants people to cycle more. Healthier people, and less cars on the roads. They could try an individual-level intervention, like sending everyone a voucher for £20 off a new bike. Or they build 20% more bike lanes, resulting in 1 in X commuters thinking “hey i think i’ll try cycling in next week”. Let’s assume the Mayor goes with Option B and is right to do so.
Now you could come up with that idea within the original model, as you are adjusting Opportunity, but the relevance of the new model is that it has new math and measurement that allows us to trace that group-level issue from diagnosis to effect.
Does it work? I don’t know. It isn’t published yet.
Does something like that even count as a behavioural intervention? I don’t know. Let’s ice that question before I have an existential crisis.
Why are you talking about this at all? Well, let me show you.
Chater and Loewenstein rest their case with this:
Giesler and Veresiu (2014, p. 841) coin the term “responsibilization” to refer to processes “through which responsibility is shifted away from the state and corporations” and toward the “responsible consumer.” Giridharadas (2019) notes that we seem to have collectively lost faith in the engines of progress that got us where we are today – in the democratic efforts to outlaw slavery, end child labor, limit the workday, keep drugs safe, protect collective bargaining, create public schools, battle the Great Depression, electrify rural America, weave a nation together by road, pursue a Great Society free of poverty, extend civil and political rights to women and African Americans and other minorities, and give our fellow citizens health, security, and dignity in old age.
We see informing s-frame [their term for system level] interventions as the future of behavioral public policy. Behavioral scientists’ excessive enthusiasm for i-frame [individual level] policy has reduced the impetus for systemic reform, just as corporations interested in blocking change intend. We have been unwitting accomplices to forces opposed to creating a better society.
OK, was it very much in the interest of societal systems to allow behavioural policy to focus on putting the onus on individuals to improve their own lives? Yes.
Is that, in fact, the only way any of this could have possibly gotten off the ground in the first place? Yes.
That’s because complex systems and individuals have one big thing in common here. They both trend towards inertia. They don’t like changing themselves.
Just as the modern person struggles to break habits, and self-sabotages their way to things remaining the same, systems of people have a habit bouncing off any attempts to shift the tide in a different direction.
This is sort of how I think about companies, or at least how they feel from the inside.
Once any business grows enough to have a hierarchy to include middle managers, the whole thing tends to become a charade. You’ve created a class of employee that plays by the rules so hard that everything outside of them becomes a threat. Corporate inertia becomes almost impossible to break out from that point on. Even if, and this is crucial, everyone can see and agree that elements of the company are diseased.
Systems can have reason to rationally reject even changes that sound like great ideas, because once they are their own living entity, something needs to keep the blood pumping, and no one really knows how the machine works.
Companies again. Nearly all of them fail to get off the ground. Barely any make real money. If you work at one that’s lasted, the economy (which we don’t understand) has decided it’s useful for some reason. If you mess with that, with even the best of intentions, you’re putting all of your friends’ salaries at risk. You better be certain (you’re not).
Scale that up to the systems that run our society, and you have the same problem, but bigger. You want to change how the systems work? You’re going to affect the outcomes of a lot of people in ways you can never fully predict.
And this is where putting this conversation under the banner of behavioural ‘science’ or even ‘policy’ becomes meaningless.
Now, the question of how we go about attempting to create the change within the individual and the system, and eventually within the society, is a scientific one (even if your answer stretches the definition of ‘science’ to its very limit). But the question of what problems we should be attempting to solve is a political one, and you ain’t solving that within the bounds of the APA style guide.
Even the question of whether you think certain biases are irrational and people are better off correcting them is political. Not because we are dancing vaguely into the policy realm, but because this is at the root of any attempt to formalise psychology. There are infinite ways to deviate from how we expect individuals to behave themselves, they are all interesting, but we agree that some are acceptable and some aren’t. So when the DSM-VI is eventually released into the world, its contents are governed by the values of the few, theoretically put there by the many.
All that is to say, this is an inherently political game, and you wouldn’t have it any other way.
So do I agree with the politics of this as an emerging shift? Meh, not really.
Chater and Loewenstein say this:
We have focused on how behavioral scientists have inadvertently assisted efforts by corporate interests to resist systemic changes, but the idea that corporate interests craft the rules to benefit themselves is hardly original (see, e.g., Acemoglu & Robinson, 2012). Nor is the idea that commercial interests promote individualistic perspectives to avoid regulation. Giesler and Veresiu (2014, p. 841) coin the term “responsibilization” to refer to processes “through which responsibility is shifted away from the state and corporations” and toward the “responsible consumer.”
Which reminds me of this passage from a response paper they received at the time, ‘The I-frame vs. S-frame: how neoliberalism has led behavioral sciences astray’, where the authors state:
At the heart of neoliberal ideology is the principle of minimal government intervention (Ayo, 2012); the idea that markets, not the government, will solve societal problems…
…Following this logic, responsible individuals should consume these new goods in order to maintain health (Ayo, 2012). According to neoliberal thinkers, individuals are “free to choose” in the marketplace (Friedman, 1990). Health risks are thus individualized, and unhealthy behavior becomes a personal responsibility detached from the structural conditions shaping it (Spindler, 2010; Gollust and Lynch, 2011; ILO, 2011). As a consequence, lifestyles and related health outcomes can become subject to blame-shifting, stigmatization and discrimination (ILO, 2011) …
…In the words of Hayek (1960, p.71), one of the founding fathers of neoliberalism, “Liberty not only means that the individual has both the opportunity and the burden of choice; it also means that [they] must bear the consequences of [their] actions”
aside: for a paper that attempts to criticise the oppressive nature of the system, you’ll do well to find a clearer example of people mindlessly obeying and pleading to authority than adding needless sources to statements that are very much still them speaking (Kelly, 2025).
I will give you that the food environment in Western societies is terrible, and is worth fixing. But, and I am deadly serious when I say this, if within that environment you have become obese to the detriment of your health, and you aren’t personally responsible for that, then what on earth could you be held responsible for? What capability does any individual have to do anything?
Those systemic changes you want, how can the people behind them bear the responsibility of implementing them? Do they get credit?
That was pedantic. Sorry. My actual point is this:
This notion of we the many versus vague magical power that is responsible for all the problems (in the response paper above this is neoliberalism, in the C&L paper corporate interests, in the discourse a host of other names) has been a political strategy for the best part of my lifetime and it is clear as day that it has gotten us absolutely nowhere. It’s dead. Flatlined. It’s 2026 soon and the only people still flying that flag are gonna be Ash Sarkar and that guy from Rizzle Kicks.
And this is where the proposed shift from helping people to fixing the system becomes less about any political disagreements and more about the methodological problems.
The one glaring problem is this: trying to shift complex systems does not work.
Something really worth reading this year was Ed Bradon in Works in Progress magazine, with ‘Magical Systems Thinking’, although I think I can fairly summarise it for you here:
What we call modern “systems thinking” often fails because it treats complex systems as controllable machines, when in reality they resist intervention and generate unintended consequences. Large systems that work today (electric grids, supply chains, the internet) did not emerge from grand designs but evolved gradually from small, functional beginnings through trial and error. Attempts to redesign complex social systems from the top down, especially in government, frequently backfire, as seen in healthcare, welfare, energy markets, and large digital projects. Systems tend to “kick back,” grow more complex, and drift away from their original purpose.
The practical prescription is humility: instead of trying to fix broken systems wholesale, create small, parallel systems with clear goals, then let them scale if they succeed. Historical successes, from Cold War missile programs to COVID vaccine rollout and Estonia’s digital government, followed this pattern. Progress comes from simple, working systems, not magical thinking about total control.
What does that actually mean for us? That’s still claiming meaningful change comes from systems right?
To answer that, let’s talk about groups and individuals again. Because we’ve talking about them as almost separate things, when real life experience is dancing between the two constantly.
Groups move in the direction that is the aggregate of all the individuals contained within. I have my biases. You have yours. We both have expertise in something. Thus the group sails in a more accurate direction than either of us could steer it on our own.
But the big issue with that theory is this: when individuals encounter a group, they stop being individuals. TLP wrote about this a long time ago.
When people are asked to estimate unknown facts in the world (population sizes, crime rates, whatever), groups outperform individuals if everyone answers independently. Each person brings their own assumptions and errors, and when those different errors are averaged together, they tend to cancel out.
But this only works while people remain independent. Once individuals are shown what others have guessed, their estimates begin to drift toward one another. People stop relying on their own judgment and instead adjust their numbers to fit what looks reasonable. The spread of answers narrows, not because anyone has learned new facts, but because social pressure pulls everyone toward the same range. As this happens, confidence rises. Seeing many similar estimates makes people feel the group must be right, even though the group is no closer to the truth than before. The crowd becomes more certain without becoming more accurate.
This destroys the mechanism that made group wisdom possible. When estimates converge, individual differences disappear and errors no longer cancel out. Now the errors are aligned. A shared mistake is interpreted as consensus. Disagreeing becomes harder, because to disagree now means rejecting not just an idea, but the apparent agreement of everyone else. Even people who push back are still reacting to the group rather than thinking freely.
And this is how progress halts.
So, I am telling you now, when Chater and Loewenstein say this:
we seem to have collectively lost faith in the engines of progress that got us where we are today – in the democratic efforts to outlaw slavery, end child labor, limit the workday, keep drugs safe, protect collective bargaining, create public schools, battle the Great Depression, electrify rural America, weave a nation together by road, pursue a Great Society free of poverty, extend civil and political rights to women and African Americans and other minorities, and give our fellow citizens health, security, and dignity in old age
And they claim we have lost our ability to change the system, I guarantee that when you look into the success stories in all of those cases where people made the world better for all of us, you do not find a complex system that has been altered from within. You will find a newer smaller system that started working in parallel.
And when you look within that new system to see what is fuelling it, you will find the only true source of change we have in society.
Agentic. Individuals.
The people who can see a system working against them, or a consensus they don’t agree with, and not give a fuck.
So much suffering and stress people are dealing with is down to complex systems with crazy incentives and levels of inertia that serve to harm the public. Kyla Scanlon’s recent coverage should be enough to convince you of this.
But if you think we are going to fix that by going into those systems and attempting to reform from within, well, I think you’re gonna be disappointed.
Remember when they said “responsibilization” refers to processes “through which responsibility is shifted away from the state and corporations” and toward the “responsible consumer”?
Our salvation will be with thanks to those who heard that statement and thought “huh, sounds like a fun challenge”.
They create a new system, but once again, that system only works if it allows those contributing it to retain their individual biases.
Which brings me to our Christmas message, and the actual reason I wrote this whole dreary thing.
COM-B is adding group models because there are many ways groups and individuals are alike, and can be designed for and measured together.
But there is one core way that groups and individuals are not alike. And this makes all the difference.
Once they get big enough, groups don’t just dislike change. They are allergic to it. It becomes impossible.
But you. You can change entirely. That individuality you lost when you encountered the group. You can recover it. And that individual agency can move an entire society.
At a psychological level, the endpoint of Lacanian (*klaxon sounds*) therapy is to understand that The Big Other does not actually exist. To believe in the Other is to believe that somewhere there is a final authority that knows what you are, what you should want, and how your desire ought to make sense. The end of analysis is the moment when this belief in a final authority falls away.
The subject loses any guilt, shame, and resentment that come from waiting for validation that can never arrive. In this sense, growth aims less at self-knowledge than at freedom from the belief that someone else ultimately holds the truth of who you are.
In other words, a return to agency. An acceptance that you are the only one responsible for your own actions.
And it is that acceptance that makes any kind of change possible.
Chater and Loewenstein say individual level interventions have failed to improve society and they’re right, but by turning away from them entirely they condemn us to nothing changing ever again.
Systems are incredibly robust. They are bound by heaps of individuals who sacrifice wisdom for confidence and harmony, and have lost sight of a life outside them.
But they have a weakness. It’s the person who can be hit with the pressure to conform to that consensus and think… ‘so what’.
So, that’s the goal. That’s the frame.
Behavioural science fails when it treats people as passive units to be adjusted, and also fails when it treats systems as machines to be redesigned.
The day we create more real individuals will be the day behavioural science actually managed to improve the world. And we’ll also probably start getting less nulls along the way.