(you can find a brief video overview of the Affective Context Model here.)
For about five years I taught psychology - including learning theory, cognitive psychology, developmental psychology and comparative psychology. One of my main reasons for leaving teaching was that I wanted to put what I knew into practice. That might seem odd, but it’s a great deal odder to find yourself in a classroom writing ‘Piaget believed learning should be exploratory’ onto a board while students obediently copy it down.
Of course there are many ways of enriching the classroom experience – but I was also experimenting with web technologies and believed that as virgin territories they offered explosive potential. The truth is that there is very little about the classroom that lends itself to learning – it is generally only the teacher’s enthusiasm that brings it to life – and that learning in this formal fashion is merely a convention, and one which runs counter to what little we know about learning. Certainly people learn a great deal at school – but mostly outside of formal environments.
I spent the next several years working with teams of developers in an ambitious attempt to apply learning theory and cognitive science to create a form of ‘super learning’ – a method which was demonstrably more effective than other forms of learning. What I eventually discovered (and confirmed experimentally) was that it was perfectly possible to apply a great many theoretical approaches (such as learning styles, modes of representation, proximal development, meta-cognitive approaches) and build something that, ultimately, was no better than reading a book. In fact, the market was full of examples of really poor elearning content which nevertheless adhered to standard ‘instructional design’ principles.
What became perfectly clear was that context rather than the content determined learning efficiency: if the organisation to which you belonged could give you a compelling reason to study (such as a life-altering test) then it hardly mattered whether they gave you content at all – let alone what format it was in. People, it turns out, are resourceful learners. I subsequently developed a deep suspicion of learning theory which on closer inspection (and with the help of people like Donald Clark) I now believe to be comprised largely of discredited musings or research artefacts with little bearing on what actually happens in everyday situations. Take Kolb’s learning cycle for example – complete tosh. It disturbed me to think of myself as a charlatan; as someone who might go through life with no real understanding of the process that lay at the heart of my profession – though I admit that this is a fairly peculiar condition.
But it’s easy to criticise. Over the last several years the question of learning has continued to trouble me, and I have tried to put together a working theory to explain those things that I have found to be true.
The Affective Context Model:
I propose that learning is the process by which people attach emotional (or affective) sense to information. It might be said that learning is the process by which people attach significance to information, but this would obscure a central point, namely that the nature of human sense-making owes a great deal to mechanisms that have evolved under different evolutionary pressures to those we experience today. I’m calling this approach the ‘Affective Context Model’.
According to this theory the storage and subsequent processing of information depends on the broader intrinsic or extrinsic affective context. Humans ‘tag’ information in emotional terms; some provided principally by the learner, others by an external stimulus. To put it crudely: sometimes it really matters to people to learn, other times someone else makes it matter. When I have asked people what they remember from school – and I have asked this often – they will remember friends, girlfriends, school food, the terrible tedium and the teachers who inspired them with their enthusiasm or by taking them seriously. This model of learning addresses these characteristics, whilst most others ignore them. And I am not confusing the motivation to learn with learning itself, I am quite deliberately stating that the mechanisms are continuous – the motivational context is what is encoded along with the information – the ‘metadata’ which determines how we store and process that information. This approach is consistent with and subsumes other areas of research – such as context-dependent memory, distinctiveness and observational learning.
The headline theory is open to misinterpretation. We have a poor vocabulary for describing what I am getting at: when one talks about affective states or emotion people imagine that this is a plea to make all learning humorous, or that somehow I am invoking theories of emotional intelligence. This is not the case: there is great subtlety to our emotional responses; for example we each have networks of ‘mirror neurons’ which automatically mirror the emotional state of the people we observe so that the model’s emotional state may be encoded along with whatever it is that they are saying . The affective context for learning is also much broader than we often recognise: at school we may be trying to impress our teacher by learning, or impress our friends by not learning. Likewise, when we start a new job we want most of all not to make a poor impression with our colleagues and managers. In addition there are elements of the emotional context which are ‘hard-wired’ – no-one will forget the teacher who collapsed mid-lecture, and some students will remember the lecture on Alzheimer’s because they have first-hand experience of a relative with the disorder. The diagram below is a rough idea of how a model might start to build.
For too long learning has suffered from a computational paradigm which equates learning with ‘information-transfer’ and implies that what a teacher does is merely to relay data from one head to another (or from textbook to head). Why else would we persist in building online courses which are little more than powerpoint presentations removed from their presenters? Some time ago, Roger Schank pointed out that stories (or ‘scripts’) are an important vehicle for effective learning and information processing – but I feel he missed the bigger picture: stories are merely one way of attaching an emotional context to information. A good storyteller is also essential, and a story without pathos is hardly a story at all.
According to the ‘Affective Context Model’ an important distinction exists between ‘pull’-type and ‘push’-type learning (illustrated below).
1) Pull Type learning: In this type the affective context is provided by the learner; the learning already means something important to the learner. This type of learning is typical of informal learning, for example the learning that takes place when a person first starts a new job. In these circumstances it is usually important that the individual does not appear foolish to the new group and they will therefore use whatever means are available to adapt quickly in a ‘steep learning curve’ pattern. It is worth noting that under these circumstances the format of the content scarcely matters – in fact more sophisticated formats (such as interaction or multimedia) may obscure the core information. A learner who is desperate to figure something out (such as how to configure a wireless connection) will be quite content with short, textual information. This format accounts for most of the learning undertaken by users of Google, for example.
2) Push Type learning: is typical of formal learning situations. In this case the learner does not provide an immediate affective context themselves – in other words, they’re not especially interested in the information. In these circumstances it makes sense to provide this affective context by, for example, storytelling, relating the context to the learner’s experience, providing dramatic illustrations, conveying the content enthusiastically and in a way which involves the learner. In his book ‘Aquarium’ Vladimir Rezun (writing under the pseudonym Viktor Suvorov) describes a short film he was shown before agreeing to join Soviet Military Intelligence. In this film a traitorous Russian Colonel was burned alive. It is clear from what follows that this information not only stayed with him but continued to influence his behaviour for many years to come. Despite this, most ‘push’ learning remains ineffective, and retention of information follows Ebbinghaus’s familiar ‘forgetting curve’. This is because learning is usually presented to the learner at a time when the importance of it is not clear or imminent to the learner – it is ‘just in case’ rather than ‘just in time’ learning.
What I would like to say, when considering these two types of learning, is that if a learner can be persuaded of the significance of something then they will engage in pull-type learning and the need to deliver large amounts of information in formal contexts (such as classroom or online courses) largely disappears. A person who has seen someone fall whilst rock-climbing will probably prepare quite carefully for their own attempt. A student who fully understands the consequences of passing or failing an examination will probably take preparation more seriously (but only if there some outcome that is meaningful in their affective context). In practice, this means that organisational learning would do better to focus more on the affective context – the reasons why the target audience might care – than the informational content itself, especially in a world where information is freely available.
The model is also strongly supportive of ‘performance support’ approaches, where learning is provided at the point of need. In this way, when learning is essentially problem-based, learners learn only when they fail, and the affective context is therefore guaranteed (assuming they want to succeed). The theory would therefore predict that learners would learn better immediately after failing. Of course the main criticism of the Affective Context Theory of learning is that it is lacks experimental support. But like most good theories it is based on observations over a period of years and is able to generate hypotheses which can be tested. One could, for example, conduct an experiment in which two groups of learners listen to the same information in the form of a story – one of which is read in dull and emotionless fashion, the other in a lively and emphatic manner (the story being identical in other regards). Students could then be tested on recall. Equally, learning performance might be compared when there is no incentive (such as a test), when learners are amongst a group of strangers (and where test results are anonymous), and where learners are competing directly with friends.
Finally, the model above is not inconsistent with operant or classical conditioning. These theories are well-supported, and explain a wide variety of human and non-human learning. What they do not do is explain the mechanism for learning in the absence of immediate reinforcement or obvious association. In summary, I am optimistic that this model has predictive power, explains some common features of learning in everyday contexts and can help professionals to identify what constitutes effective learning in a range of circumstances. It also underpins more specific pieces of work, such as the learning design toolkit previously posted on this blog, which explains in some detail how this approach to learning might be applied in practice. Please let me know what you think.