This paper is designed to outline the current state of system 1 – how it is understood and measured in applied research (as opposed to academic research). Practitioner work has a very different emphasis to academia – we are obliged to use models and tools that can be shown to contribute to the resolution of client challenges. Dual process theory has captured the imagination of policy makers and strategists the world over. But just how well does it meet the needs of practitioners? This paper is designed to challenge the market to apply more critical thinking – both in terms of defining terms and developing effective measurement tools. Without these ingredients, practitioners will rapidly start to lose interest in this emerging area – we want to ensure that dual mode processing is properly developed and capitalised on. We believe it is currently at risk of failure as the lack of rigour feeds through to lack of value creation.
Surely by now, we are all aware of dual process theory – the way in which we at times process information in a fast and intuitive way (system 1) and at other times how we do this in a slow and deliberative manner (system 2). Although this distinction has very long historical and philosophical roots, Daniel Kahneman’s iconic book, Thinking Fast and Slow, cemented it in the collective consciousness of all those that wanted to better understand consumer decision making.
The promise that this distinction offers is significant. There are so many activities to do and decisions to make that if we pondered on them all in their fullness we would struggle to get through the day. As such, system 1 processing is helpful to us because we can operate in a semi-automatic way, quickly and efficiently making decisions and choices and generally getting on with our lives. As consumer researchers, we are interested in the way that many of the decisions we make in low involvement categories are more likely to be system 1 in nature. We are surely less likely to think carefully about our choice of cleaning fluid than the choice of school for our child.
There is a lot of interest and excitement about the potential of system 1 to provide new insights into consumer behaviour (and therefore to help drive change). However, there is also a lack of understanding and clarity about what the term means and what the measurement implications are. This document is designed to help clarify the thinking and pose a challenge to the industry for a more realistic and thoughtful approach to this area. There is currently a lack of critical thinking in this space which has potential to derail this important development in consumer research.
The need for an applied practitioner’s perspective on system 1
Much of the evidence for the importance of system 1 in shaping our judgements and decisions come from the academic literature. There are dangers here for practitioners as our needs are fundamentally different to academics. The reason for this is that academics are keen to demonstrate the presence of mental processes and as such will create experimental designs with this purpose in mind. These designs involve controlled conditions, fairly uniform populations (typically WEIRDs – Western, Educated, Industrialized, Rich and Democratic) and the tasks being asked of these participants often quite unusual. The emphasis on academics is to demonstrate the presence of a mental process in humans, not to really explore the implications of this in our daily lives.
The task of practitioners is quite different. We operate in environments where there are multiple influences shaping behaviours, a wide range of consumer types and a range of decisions being made. As such, we need to understand the relative importance of any influence on shaping behaviour. We need to be able to measure the variation with which these influences are present between consumers. We need to be able to assess the degree to which these factors are relevant in different consumer activities. And we need to be able to do this in a way that matches our clients’ geographies, categories and consumer groups if it is to be of value.
As such any practitioner framework is only valuable if we are able to properly define terms and operationalize it at scale. There is currently a lack of clarity about ‘system 1’ on both these counts. We deal with these in turn below.
Common errors people make about system 1
Whilst there are real benefits to the use of dual mode processing model for understanding consumer behaviour, we are also in danger of making mistakes by oversimplifying the principles. There are a number of traps it is easy to fall in when referring to this:
- Continuum, not a dichotomy: The notion that system 1 and system 2 are qualitatively different to each other is misguided. We might want to think of these as modes rather than types of processing, which means they are more interchangeable than is often assumed. For example, it is possible to carry out analytical reasoning both in a slow and careful manner as well as quickly and casually. Or indeed any point in between.
- Defining terms: there is a huge proliferation of labels used in this context – implicit/explicit, automatic/controlled, conscious/unconscious, reflexive/reflective and so on. All of these come with their own baggage and fuzziness but suffice to say that at times it is not clear precisely what distinguishes system 1 from system 2. Terms are often thrown around without fully considering the meaning or implications for measurement.
- Emotion: On the above, it is worth making the point that System 1 is often considered synonymous with emotion (whilst system 2 is considered to be rational). So if I am excited by loud music in a store then I am pretty aware of this (albeit there will be variations between individuals and indeed situations in the degree to which they are able to read their own emotional states). Similarly, habitual behaviour (system 1) is not necessarily emotional in nature. Clearly, there are logical inconsistencies here in the assumptions.
- Diversity: It’s fair to say that system 1 processing has at its core the proposition of autonomy – it makes minimal demands on mental activity. But there may be different ways in which this works. It could be through lower emotional regulation, instinctive evolutionary patterns, highly learned mental routines or associations. Immediately we can see that system 1 is a diverse range rather than one single form of mental activity.
- Interchangeability: It is clear that we often invoke system 1 processing, as humans tend to operate as cognitive misers. Nevertheless, we are clearly capable of interrupting this by invoking System 2. But what might spark this can vary from one person to the next depending on what is important for them. I may idly be choosing toothpaste but the recollection of my sensitive teeth could make me stop and consider my options more carefully.
Whilst it is not in dispute that dual processing is important, it is clear from the above that it is a far more complex and nuanced phenomenon than is often understood. We need to use a much broader and deeper understanding of the decision-making process if we are to make a tangible impact. Simply classifying decisions as ‘system 1’ or ‘system 2’ is not sufficient.
Challenges for measurement of system 1
There is a temptation to equate market research techniques with system 2 approaches and techniques such as neuroscience and behavioural science (such as biometric responses, reactions times, facial expressions etc.) as system 1.
This in itself reflects some misunderstanding of both the ontology (what dual processing means) and the epistemology (how we measure it). There are a number of points worth raising here:
- When people are asked to carefully think about their responses, there is a much higher correlation between measures relating to system 1 and system 2. In other words, a carefully constructed questionnaire can reflect the consumer attitudes, emotions and behaviours when they are in a low involvement situation such as grocery shopping. Consistent with this, consumers with a high need for cognitive deliberation (i.e. those who are able to self-reflect on their own actions, feelings, and thoughts) show a much higher relationship between implicit and explicit measures.
- ’System 1’ measures (such as IAT) have low internal consistencies, due it appears, to measurement difficulties. For example, when you aggregate the IAT’s many different tests, such as race, disability, and gender, it has a test-retest reliability of about r = .55. By the usual standards of psychology, this puts these IATs below the threshold of effective value in most practical, real-world settings. Research that has controlled for measurement error between system 1 and system 2 have found high correlations.
- Predictive validity: Despite there being explanatory value, there is little-published evidence that data collected in system 1 conditions are better able to predict subsequent behaviours than that collected in environments that require a more considered response. The IAT claims to be measuring implicit attitudes, but is this really the case? What if someone who scores high on an implicit attitude never actually acts in this way? Can a bias be a bias if it only exists in the context of a test result, but is never reflected in real-world behaviours? Both critics and proponents of the IAT now agree that the statistical evidence is simply too weak for it to predict individual behaviour. Of course, there may be relationships that have not yet been identified but the case still needs to be made with good quality research.
The point here, again, is not that implicit measures are not useful but that as practitioners we need to be careful, as simplistic solutions can lead us astray. Implicit measures can at times be useful but as practitioners, we need to be clear about what the boundaries and conditions are for our tools.
And when we collect implicit measures that appear to tell a different story to the explicit data, it is all too easy to construct a rationale for this. But as humans we are great storytellers – we will easily see patterns in the data where none exist.
Have we over-interpreted dual processing model?
Dual processing theory was never designed to be a coherent model for explaining how humans behave. Instead, it is a framework for describing different sorts of decisions, a useful means of categorising rather than predicting.
As such, it is a framework that has a different value for academics and practitioners. Academic psychologists are keen to identify the presence of mental phenomena, developing our understanding of them. But applied practitioners have a different requirement. Mental phenomena are only of interest if we can understand and ultimately measure their implications for real-world outcomes. Of course, in applied settings means of categorising decisions are also useful but we also need to be able to use it to resolve client challenges.
There is a danger that the dual processing model has been the victim of reification. This is the way in which we treat an abstract concept as if it were a real thing. Psychologist Gerd Gigerenzer was saying something similar to this when he said:
“What is system one and system two? It’s a list of dichotomies… Usually, science starts with these vague dichotomies and works out a precise model. This is the only case I know where one progresses in the other direction… What the system one, system two story does, it lumps all of these things into two black boxes, and it’s happy just saying it’s system one, it’s system two. It can predict nothing. It can explain after the fact almost everything. I do not consider this progress.“
Gigerenzer’s account is perhaps a little harsh but it nevertheless makes the point that the time has come for the thinking around system one and system two to move on.
The problem for the market research industry is that there has not been sufficient due diligence to establish the value of dual mode processing in applied settings. This is due to:
- definitional issues – there is not always enough clarity in what we mean by system 1 (e.g. Is it emotion? Is it the style of cognitive processing?). There is confusion whether system 1 refers to the nature of the decision environment (e.g. low involvement) or the decision itself (e.g. fast)
- measurement issues – it is debatable whether current approaches actually measure what we think they are measuring and the ease of undertaking measurement in the variety of environments of interest to us
- deference to the academic literature – not enough has been done to decide which literature is relevant (in that it can drive change in our clients’ organisations) and which is ‘nice to know’
A manifesto for going ‘beyond system 1’
A manifesto for action is urgently needed to provide a more nuanced understanding of dual mode processing in consumer behaviour. And to do that we need to answer the following questions:
- What are the circumstances that invoke a system 1 response that is different to our reflective attitudes? So, for example, can all apparently low involvement shopping activities always be ‘system1’ for all people?
- How strong is the relationship between system 1 responses and behavioural outcomes? At the moment the academic literature suggests a weak relationship but there is little relating to the prediction of consumer behaviours.
- Are all system 1 responses inaccessible by explicit means? Indirect question techniques have long been a staple of market research and more creative survey approaches may well be useful for accessing implicit attitudes.
- What is the nature of the relationship between system 1 and system 2 measures? Variations between measures will inevitably at times be due to design effects. We need to have clear quality controls in place so we can be confident we can separate out these design effects from genuine differences. And once we have done this we need a coherent understanding of when we find differences in measures designed to reflect system 1 and ‘system 2’ thinking.
In the rush to help clients gain new insights and value from system 1 we are in danger of not paying sufficient attention to our core understanding of how dual processing operates or offering sufficient rigour in the means by which we measure it.