An integrated approach to understanding behaviour

One of the key requirements for successful behaviour change is to understand what determines the behaviour you are trying to influence. This blog will describe five of the most important approaches to understanding behaviour (operant conditioning, cognitive behavioural models, psychodynamic models, rational choice and constrained optimisation, and behavioural psychology) and then will propose a model for integrating the different perspectives.

Operant conditioning

One of the most fundamental influences on human behaviour (as well the behaviour of animals) is operant conditioning (Skinner, 1948). Operant conditioning describes the impact of rewards and punishments on behaviour, with positive reinforcement describing a reward following a behaviour and strengthening it, negative reinforcement the removal of an unpleasant stimulus after a behaviour that again strengthens the behaviour and punishment being the application of aversive stimulus after a behaviour weakening it. What is often missed is that this effect does not need a conscious connection between the behaviour and the reward/ punishment, ie to train a pigeon to tap a particular piece of cardboard you wait until the pigeon randomly taps the right place and then reward them, you don’t communicate in some way with the pigeon to ‘tell’ it to tap the right place and then reward them. Once the pattern of behaviour has been reinforced, it will continue for some time even after the rewards/ punishments have stopped. Research has also shown that different reward strategies are more or less effective in establishing a pattern of behaviour, with intermittent random rewards establishing the behaviour very quickly and meaning it takes a long time to ‘extinguish’, a pattern that describes most commercial gambling operations very accurately. Indeed, when you interview drug or gambling addicts, there are points where they cannot explain changes in their behaviour such as escalating use, and these sound very much like the result of conditioning that happens at a non-conscious level (Barnard et al 2009a, 2009b).

Cognitive behavioural models

Behaviourists like Skinner, who emphasised that research into behaviour should focus almost exclusively on external, observable behaviour, held sway for many years but eventually the ‘cognitive revolution’ shifted the focus to understanding the effect of interpretation and information processing had on behaviour. Early challenges came from researchers who demonstrated that rats conditioned to find their way through a maze to get food, didn’t randomly try other routes when the first route was blocked off. Instead, they went to the route that was likely to be the most direct path to the food. The conclusion was that the rats weren’t only learning through trial and error, they were creating a ‘mental map’ of the space and that clearly influenced their behaviour.

Those early challenges led to a huge amount of research that continues to this day exploring the nature and structure of thought processes and how they affect behaviour. A key concept is ‘cognitive schema’ which are sets of information that organise thoughts and ideas. This is a subtle concept, because 'cognitive schema' sounds like a fancy way of saying ‘ideas’, but their power comes from the fact that they can be learnt without any formal training or study and accessed almost instantly. A good example from the field of language is the Mrs Merton joke ‘What first attracted you to the millionaire Paul Daniels?’. We find this funny because we have a cognitive scheme or set of ‘maxims’ that help us interpret language in the real world (Grice, 1975). One of these is ‘don’t add irrelevant information’. Because ‘millionaire’ is not needed to ask the question, the implication is that it is relevant in some other way and therefore implying that it is financial status rather than his inherent qualities that were attractive. But though people could probably articulate that if given enough time, it was never explicitly taught or learnt and no conscious thinking was required to apply the rule, illustrating the difference and power of cognitive schema as opposed to just a set of rules or information that one learns in other ways. A simple behavioural cognitive schema is the belief about what the different lanes of the motorway are for. Some people think there is a fast, medium and slow lane, while others think the outer two lanes are only for overtaking. These contrasting schema will lead people to exhibit very different driving behaviour and react in very different ways to others' driving behaviour. 

Cognitive behavioural therapy (Beck et al, 1979) builds on this cognitive model by observing that our responses (behaviours and emotions etc) to the world (stimuli) are mediated by our interpretations of them, including underlying values and related ‘if, then’ associations (eg it is important I am respect, if someone disrespects me then I need to retaliate). These processes can be distorted or biassed in various ways or generally be unhelpful, and broadly CBT aims to make this interpretive framework explicit and challenge unhelpful or inaccurate elements through ‘cognitive restructuring’. Also within the broad cognitive behavioural framework are the theory of reasoned action and the theory of planned behaviour (Ajzen et al, 1975 & 1986), which links attitudes, subjective norms (beliefs about what society holds about something), and beliefs about an individual's control over an action or behaviour, to the intention to carry that behaviour and the behaviour itself. A further development of this model (that also takes into account a range of other behavioural models), is COM-B (Mitchie et al, 2011), that groups influences on behaviour into the categories of capability, opportunity and motivation, thus recognising important contextual influences not emphasised in many cognitive behavioural models. 

Linked theories, which also recognise the importance of context, particularly social context, are social learning theory (Bandura, 1977), which links learnt behaviour to the observation of others behaviour and vicarious rewards and punishments; conformity, acquiescence and obedience (Milgram, 1974); the often overlooked but ubiquitous influence of impression management (Goffman, 1956), and the need to manage the tensions between one’s ‘actual self’, ‘ought self’ and ‘ideal self’; and attribution theory (Weiner, 1979), that models how people draw inferences from other’s behaviour based on assumed characteristics and the controllability of the situation.

However, while cognitive-behavioural models are powerful tools and have a substantial amount of evidence supporting them, it is also important to recognise their limits. Not all schema or cognitive evaluations are consciously accessible, and there are a range of examples where experiments have demonstrated that something affects behaviour, for example the presence of others on helping behaviour (Latane & Durley, 1968) and finding the solution to the ‘two string problem’ (Maier, 1931), but individuals have not been aware of the influence and in some cases actively denied that they were influenced. 

One key phenomenon to be aware of in this context is the influence of the ‘interpreter module’ (Gazzaniga and LeDoux, 2014), which means people tend to create causally plausible explanations for their own and others’ behaviour, even when those explanations are in reality not correct or even not possible. This means that while a huge amount of insight can be gained from exploring the reasoning behind people's behaviour, including eliciting the underlying cognitive schema that are informing their thinking and decisions, this may not be sufficient to understand all the drivers and influences that are operating and sometimes could be misleading.

Psychodynamic models

Though the cognitions in cognitive behavioural models are not necessarily conscious, a key assumption is that they are largely consciously accessible. This is not assumed within psychodynamic models, within Freudian theory in fact the ‘cognitions’ are often supposed to be camouflaged or hidden from conscious thought. While the substance of Freud’s theories are no longer seen as accurate, one of the key modern-day and well-evidenced inheritors of the tradition, attachment theory (Bowlby, 1969), retains the sense that the cognitive framework an individual uses to interpret the world is not necessarily accessible. Attachment theory holds that the nature of a parent’s or caregivers interaction with a child affects their 'internal working model' of relationships and the world, leading to ‘secure attachment’, where the caregiver provides a secure base from which the child can explore, or various forms of insecure attachment due to a maladaptive internal working model. Evidence of these internal working models can be seen in very young (pre-verbal children) in response to being separated from their caregiver. In broad terms, securely attached children will become upset but quickly be comforted when their caregiver returns. Insecure ambivalently attached children will become upset and take a long time to be comforted and insecure avoidant attached children will seem not to be upset at all.

These models are not stored at a conscious level and the young children who display the behaviours described above won’t be able to describe them. Equivalent patterns are applied to adults and affect adult behaviour and relationships (Waters et al, 1985) and are diagnosed using the ‘adult attachment interview’. But even for the adult, these models are not necessarily consciously accessible and the assessment of the interview is not based on the content of the interview (ie it’s not what people describe about their current lives and their past experiences) it is how they describe them, for example do they still 'entangled' with past experiences. So while psychodynamic approaches, like cognitive behavioural ones, go beyond just looking at external, observable behaviour, their assumptions about how the internal structures of the mind affect behaviour and the degree to which they are consciously accessible are very different.

Rational choice and constrained optimisation

The previous approaches to understanding behaviour are all rooted within the field of psychology. In contrast, rational choice theory and constrained optimisation provide the foundations for economic theory (or the science of understanding choices within scarcity, as many modern economists would define it). Rational choice theory doesn't really concern itself with the debate between behaviour and internal mental structures. Instead it posits a simplified model of the human as someone who places certain value (utility) on different goods and services and then acts in a way that is consistent with maximising their overall happiness. In this context, rational is more akin to 'consistent' than the common usage of the term rational, ie eating lots of pizza might seem irrational in one sense, but rational choice theory doesn't make a broader judgement on it, it just requires that one's liking for pizza is stable over time and doesn't shift around in arbitrary ways when compared to other goods. But while it is a simplified model, rational choice theory is based on some deep insights about human psychology and leads to an extremely powerful and accurate model of behaviour. 

One insight is that goods have a diminishing marginal utility, in other words, your sixth slice of pizza brings you less additional pleasure than your fifth. This means that if you have two goods, say pizza and champagne, they will form a convex 'indifference curves', which means you are equally happy to have five pizzas and five glasses of champagne as you are to have one glass of champagne and fifteen pizzas or fifteen glasses of champagne and one pizza. Added into this insight, is the modelling of attitudes to risk (with most people being risk averse) and attitudes to intertemporal substitution, ie that people tend to prefer £8 today to £10 next week (a phenomenon known as delay discounting). The brilliance of these insights really becomes apparent when combined with the idea of scarcity. People have limited budgets, so consciously or unconsciously undertake a constrained optimisation exercise which means they choose the mix of goods that maximises their happiness given their budget. This theory, based on a number of relatively simple assumptions and developed with mathematical rigour, can predict and explain an astonishingly broad range of behaviour. For example, it can explain why people work more hours when they get paid more up until a certain point but then work fewer hours as their pay increases. 

Behavioural psychology

Rational choice theory makes a number of simplifying assumptions and in recent years the new field of behavioural psychology or behavioural economics has challenged some of these assumptions on the basis that a more psychologically accurate model can make better predictions in some circumstances. One of these assumptions concerns time preferences. Behavioural psychology builds on delay discounting, which assumes a constant diminishing of value as the delay increases, but adding the concept of hyperbolic discounting. This assumes that £10 tomorrow is worth £8 today, but also that £10 next week is worth £8 today (ie the additional time doesn't reduce the value any further). This concept helps explain why people plan to start going to the gym next week, but don't follow through on the plan. Another key psychological concept added to rational choice theory is loss aversion, the idea that it is more painful to lose £10 than it is pleasurable to gain £10. This not only applies to money however; I observed a good example on the train recently. A man gave up his seat on the tube for a pregnant woman, but then a seat subsequently came free, and instead of taking it, he offered it to another young woman who had been standing when he previously was sitting. He didn't give up his seat to her when he was sitting, which would have been a loss, but was happy to give up the chance of a seat when he was standing, which would have been a gain, illustrating the power of loss aversion. Other important concepts within behavioural psychology include social preferences (we do things that benefit others as well as ourselves), limited attention (partly as a result of using mental shortcuts or heuristics), state dependent preferences (the value we assign to different goods changes depending on our states such as our mood), and utility from beliefs, the idea that believing something to be true brings a benefit even if it isn't true, which for example helps explain why people might not want to get information about genetic risk factors for disease as it is nicer to be able to believe you won't get a disease. 

An integrated model

Though behavioural psychology started to bridge the gap between rational choice theory and more subtle psychological insights, it primarily focuses on amending rational choice theory rather than undertaking a more fundamental integration. The key issue from a behaviour change point of view, is that rational choice theory starts from fixed preferences and doesn't provide insight into how those preferences are formed or could be changed. In contrast, psychological models focus on how motivations and intentions are influenced but leave a substantial intention-action gap. The integrated model below attempts to bring these two ways of thinking about behaviour together. The key step is to consider the outcome of models such as COM-B to be ‘preferences’ rather than ‘intentions’. This means that constrained optimisation can fill the ‘intention-action’ gap. The rational choice model is then augmented by having a systematic way of understanding how preferences are formed but also takes from psychology the evidence to suggest that both attention (Simon and Levin, 1998) and the ability to resist temptation (Mischel, 1972) are limited resources that need to be added into the constrained optimisation process. In addition, the focus in psychological models on capability and opportunity are added in the form of actual capability and actual opportunity (alongside perceived capability and opportunity that inform preferences). The model is set out graphically below. 

To illustrate how the integrated model can be used to understand behaviour and behaviour change, I’ll apply it to two examples, the first being going on holiday to Greece and the second going to the gym.

In order to get someone to go on holiday to Greece, one first needs them to be motivated to do so. This motivation will be influenced by their views of Greece (their cognitive schema), friends and others views of going to Greece (conformity and social norms) and their past associations with Greece and holiday habits (operant conditioning). It will also be influenced by their perceived capability of going to Greece (do they think they have enough money, can they manage the booking) and perceived opportunity (do they think there are flights available, is Greece allowing tourists into the country). If enough of these things point in the right direction, the result will be that they have a certain preference for going to Greece and it forms part of their constrained optimisation. They have limited time and money and have to decide whether to use that resource going to Greece or perhaps France instead (perhaps France is cheaper but Greece has ‘English’ pubs) or not spending their money on a holiday but instead saving it to buy a car or put a deposit down on a flat. This decision-making could be affected by things such as time preferences, perhaps they have money after Christmas and could spend it now on going out or save it to go on holiday in their summer. Perhaps they recognise that they will find it hard to save until the summer, so they put down a deposit on a holiday (using a commitment device that is needed because of the effects of hyperbolic discounting), so their preference has become a concrete intention. However, when it comes to actually going to Greece, they need to have the actual capability (for example they need to have a valid passport) and the opportunity (they need flights not to have been cancelled because of a shortage of ground staff or because of an erupting volcano sending plumes of ash into the flight path). Only if the implementation action is successful, will they board the plane and a few hours later actually end up in Greece, ie exhibiting the desired behaviour.

In the case of going to the gym, again people firstly need to be motivated to go, which will be affected by their views of it (it will help me lose weight, it will make me stronger), as well as friends views and their daily habits (do they habitually reach for the bottle of wine and the remote when getting home). They’ll also be influenced by their perceived capability (will they know what to do in the gym, can they find their way there) and their perceived opportunity (is there a gym close enough, will their need to earn a living and do home tasks or parenting leave them enough time to go). Once they’ve settled on a preference for going to the gym, it will become part of their constrained optimisation process, where its relative benefits are traded off against seeing friends, staying at home and watching Netflix, joining a running club, time spent with children etc, and taking into account their limited time, resources and energy. If the end result of this process is the intention to go to the gym, implementation of this will be affected by how easy it is to actually get to the gym, whether the gym is open, whether they can navigate to it and then use the equipment etc. They may also continually postpone implementation due to the influence of hyperbolic discounting. One technique that gyms have used to counteract this is by making thriller novels available on Audible only when people are at the gym, which acts as a commitment device. In contrast, having TVs available at the gym so you can watch the same programmes you do at home is feeding into the constrained optimisation process by changing the balance of costs and benefits. 

It should be noted that the model described above is a simplified version of reality, and it does not reflect the multiple feedback loops that exist within the system (particularly between the experience of trying to implement the behaviour, the behaviour itself and motivation). However, the model hopefully provides policy-makers and those designing interventions a big picture understanding of the factors that influence behaviour and where the barriers and facilitators. One implication of the model is that addressing a barrier in one part of the system may be necessary but not sufficient to change behaviour. For example, increasing people’s motivation to do something (which in itself is highly challenging) may not be sufficient because that new preference has to successfully pass through a constrained optimisation process as well as actually be implemented. Given the fact that many (and perhaps the vast majority) of interventions aimed at longstanding or entrenched problems are not effective, the importance of recognising and trying to address issues along the whole length of the behavioural pathway is critical to successful behaviour change.

The Integrated Model: Revised version

In the revised version, the choice environment is subdivided into a choice environment and an implementation environment, as shown below.

The context in which someone tries to implement their decision can have a major effect on whether they are successful in carrying out the behaviour. The research I’ve undertaken with people aiming to address their drug issues provides very vivid examples of this. In one case, a drug taker decided that he did not get one with the heroin substitute methadone so wanted to switch to the blocker Subutex. His case worker so annoyed by his decision, that he discharged him from treatment and put him back on the waiting list. In another case, a long term drug user, who suffered significant abuse as a child, had been released from prison and went to his GP to ask for help, and his GP told him all he needed to do was ‘pull himself together’. In both these examples, the individual had a clear intention to stop using drugs, but the environment in which they were trying to implement their decision had significant barriers that made it much more challenging.