The Unspeakable Truth

The (edited) transcript for a recent interview:

Interviewer: one of your central themes is a critique of conventional training – you’ve described traditional Instructional Design as ‘mumbo jumbo’ for example – can you tell me why?

Me: Yes. These things are not on a firm foundation – I mean they don’t stem from a basic understanding of learning, so they are – for the most part – just a mix of folklore and artifacts.

Interviewer: can you give me an example?

Me: Sure. Take ‘learning styles’ for example – that’s just folklore. It’s like ‘digestion styles’. The reality is we all learn in the same way, with many small individual variations – just like we all digest in pretty much the same way. The basic mechanism is the same.

Interviewer: what about all the respected research behind instructional design?

 Me: The plural of research isn’t theory. We seem to have forgotten that the scientific method is a way of testing hypotheses derived from theory – it’s not a way of coming up with theory. Without theory the research can just be misleading. Imagine I decide to do some research into learning and lemons - maybe I am funded by Lemon co. - and I do a bunch of experiments and I discover that in some conditions lemons seem to improve learning. That’s perfectly plausible. Now lemons have made their way into the Instructional Design playbook – and it’s supported by research!

Interviewer: Can you give an example of something that falls into this category today?

Me: Sure. Repetition. Lots of people seem to think repetition is a good way of reinforcing learning. And it does work to some extent – but then so do electric shocks, bullying, threats of violence. Actually these are all abuses of our learning system, but because we don’t understand learning we can’t say why. The popularity of repetition stems back to Ebbinghaus who found that repetition could improve retention. What people missed was that it was retention for garbage: Ebbinghaus was using ‘trigrams’ – meaningless sequences of letters – in his experiments. The human mind is designed to forget garbage, so we don’t really learn anything about learning with a piece of research which shows how to force people to retain garbage. To see how silly this is, just imagine a learning system which required a person to be bitten twenty times by a tiger before they learned that tigers are dangerous. That kind of learning system wouldn’t survive natural selection.

Interviewer: but surely a theory needs to be supported by research?

Me: yes and no: in the beginning any new theory is strictly unsupported, since you first need to test hypotheses derived from that theory. Actually for a theory to be good it needs to have explanatory and predictive power – so sometimes a theory is just good because it explains the things that our existing theories can’t. This was true of the general theory of relativity, and the theory of natural selection.

Interviewer: so tell me about affective context model – the theory that you have proposed – can you explain it?

Me: I can try – although I seem to say one thing and people hear another. In essence the theory is that we don’t actually remember anything – we remember how things made us feel and then we use those feelings to reconstruct events. If I can tell a bit of a neuroscientific story: our sensory systems are bombarded with huge amounts of information. At the first level of processing we have neurons that do simple tasks – like edge detection for example – but beyond that we need a hyper-efficient way of crunching all that information down and storing the important bits. We do this using our affective – or emotional reactions. When we start out the reactions are relatively simple and primitive. As we develop these affective reactions become highly sophisticated and differentiated – a bit like our visual system – and we can use these patterns of reactions to reconstruct events. But of course one of the features of this model is that these ‘memories’ are always inaccurate to some degree – which is exactly what we find in real life. In fact if I asked you to draw a picture of something with which you are intimately familiar – like the front of your house – there would be shocking inaccuracies.

Interviewer: why do you think this is a good theory?

Me: it explains lots of things about learning that we know to be true – experimentally and personally – that we haven’t previously been able to explain. For example, Loftus’ research into eye-witness testimony which found that use of words with very different affective context – ‘smash’ and ‘bump’ – could affect memory at the time of recall. Bartlett’s ‘War of the Ghosts’ research, Brown and Kulik’s flashbulb memories – basically any research into everyday memory processes. It also explains why so much research is artifactual and misleading – primacy & recency and repetition for example. If repetition were so important I would surely still remember school hymns, and not a damning comment from person who passed me in the street when I was a teenager. I would remember more about brushing my teeth this morning than I do about the argument with a colleague several weeks ago. It also explains why we remember the things we do about our lives – the things with emotional impact. It explains why two people may experience the same event but remember very different things. It explains why storytelling is central to human nature. A crude way of expressing it would be that human memory is designed to store important things and forget unimportant ones – and this is completely different therefore from computers.

Interviewer: why do you think people have such difficulty understanding this theory?

Me: language gets in the way. We have gotten used to thinking of emotion as crude corollaries to our experience – happy, sad, angry – not as the building blocks of experience itself. Imagine you had to explain vision to an alien species and after listening to your explanation of the eye - rods and cones and so on - the alien said ‘so vision is detection of patches of red, green, blue?’. You would somehow want to say ‘it is so much more sophisticated than that’. Basically I am saying that cognition is comprised of subtle affective responses to our environment – every concept, every memory, every thought is comprised of affective responses. This is how we encode and process the world. This is why human cognition is fundamentally different from computation.

Interviewer: I can see how this might apply to episodic memory – but what about semantic memory?

Me: This is a great example of where a silly idea has been incarnated as research – like the lemons example. In reality there aren’t two different kinds of memory. For example if we look at people’s ‘episodic’ memory for their birthday parties they often draw on ‘semantic’ elements – like balloons and jelly – that weren’t actually present. And vice-versa. I guess a more rhetorical way of tackling this would be this: do you really think chimps have separate episodic and semantic memory? If not, can you really believe that given the relatively short evolutionary distance between us, that humans have a completely different way of remembering? There’s a more important point here I guess: many theories of learning assume some kind of radical discontinuity between ourselves and our biological cousins. My approach assumes continuity: chimps, dogs, mice all use affective context as the basis for encoding experience.

Interviewer: you started out in education, but these days you don’t tend to get involved in the education debate much – as opposed to the corporate learning debate – why is that?

Me: It’s become so byzantine. It’s like a world in which people are passionately debating the guidelines for preventing people from falling off the edge of the world. And you want to say ‘look – the world is round!’ but that would just negate all the crazy conversations people are having – about class sizes and standards and testing. What I really want to say is that unless you start with a firm foundation – an understanding of learning – then the debates just turn in circles.

Interviewer: how would education be different if it were based on an understanding of learning?

Me: A few obvious things: There would be a systematic mapping of people’s concerns on entry to the system. There would be a process of exposing people to different contexts – a kind of ongoing affective matching process. The general mechanism would be about presenting people with challenges of increasing sophistication together with the resources to tackle them. Finally there would be a blurring of education and work – people would simply be shifting the percentage of activity which they pay for vs what they get paid for depending on their level of competence. You would have a portfolio of such activities – you might do cyber-attack and cartooning for example. And of course it would be lifelong.

Interviewer: what are the biggest barriers to change?

Me: Convention – and conventional ways of thinking I guess. People tend to operate by doing what everybody else is doing. We’ve got to the point in learning where people sense that there is something wrong – they can see that our conventional approaches aren’t working, but because their underlying assumptions stay the same they can’t figure out how to fix it.

Interviewer: what are the underlying conceptions?

Me: that learning is knowledge-transfer. That somehow teaching, learning education is all about getting information from one place to another – from the teacher’s head – or the book – into the student’s head. This is starting to look blatantly ridiculous in the internet age – but we carry on doing it, because we can’t imagine what else learning could be. I call this approach ‘content dumping’.

Interviewer: so how do we apply this theory – Affective Context – in practice?

Me: Affective Context has basically two big practical implications – I sometimes say ‘there are only two things we can do in learning’. The process of learning design has to begin with understanding what people care about and what they don’t – since it is peoples’ patterns of concern that drive learning. In the first case – where people do care about something – then all we need do is provide resources. This is the origin of the ‘resources not courses’ mantra: if people care about something – for example not looking stupid when they start a new job, - then you can design a resource like ’10 mistakes to avoid in your first few weeks’ and people will use it. In the second case – where people don’t care about something – we have to build affective context if we want them to learn. This typically involves a different set of techniques: simulation, exposure, scenario, storytelling etc. These are all approaches which are low on information and high on affective context.

Interviewer: Do you think that learning and education will change?

Me: I worry that unless they do they will be ‘consumed’. I mean there are other areas of life – marketing, politics – where, in essence, people are starting to map affective context in order to get stuff done – sell products, win elections. They are doing it with an increasing degree of sophistication without really needing to understand the underlying theory. Everywhere where technology interfaces with humans, affective context determines the success of the interface. Look at ‘clickbait’ – or Netflix - for example. These areas where entrepreneurs have figured out how to get people to do things by mapping them affectively - and they will start to encroach on learning and education, if learning and education remains stuck in a paradigm of getting people to memorise information. Business are more interested in what people can do.