The Learning Relationship

-->

An individual can be thought of as the total set of concerns for a specific organism. This is true whether the individual is human or not. Like humans, a bird may be concerned for its offspring, like humans a dog may enjoy spending time in the park. Human concerns reach higher levels of specificity due to their cognitive flexibility – it is unlikely that either dogs or birds pay much notice to Opera for example – and this feature permits much greater individuation in the species. Whilst sharing many basic concerns, humans very greatly in their learned concerns.

We can define a concern as anything which elicits an affective response (a ‘reaction’) in an individual. A person with a passion for architecture may react strongly to a building, whilst someone without that interest may not even notice it. Our concerns govern not only the encoding of experiences – it is the reactions themselves that are encoded, and form the basis of cognition.

The question I wanted to consider here is the way in which these concerns develop – in the process we call learning – specifically in interactions with others.

A fundamental principle is that new concerns always spring from the old: for example all creatures share a concern for painful experiences. This enables a child that touches a fire to learn to fear fire, developing a new concern in the process. A more subtle concern might be an innate concern for status (and the desire to avoid embarrassment). Over an extended period of time this is shaped into a concern for fashionable clothing.

Learning, in this sense, happens quite naturally and does not require the intervention of other individuals.

Humans, in common with many species, are social. Social species have certain mechanisms which ensure the harmonisation of concerns with an ‘ingroup’ – and this extends so far as a culture. These mechanisms act to suppress the expression of individual concerns in the interests of maintaining group cohesion. For example, it may take you a long time to discover that a work colleague is an avid Opera fan. This is because the work context implicitly proscribes a set of behaviours that members of the group are expected to conform to (norms) – and it will likely be outside of a work context that one really gets to know a colleague therefore.

Human interaction is also characterised by conversation, to a degree that is peculiar among mammals. During conversation, two or more participants explore the respective pattern of concerns that make them individuals.

This is a complex process, but the basic dynamics is simple: each party takes turns sharing their concerns. In general this proceeds by establishing shared concerns, then exploring finer levels of detail. The parties then go back and forth, taking turns to share a little more of what makes them an individual. Overall human relationships are governed by a relatively simple rule: similarities good, differences bad. Two people who quickly discover they have nothing in common will likely go separate ways.

The difficulty with this account so far is that is leaves no space for learning: too much similarity and we become entirely homogeneous (cultures vary in this regard).  For this reason (and to a degree that varies between individuals and cultures) we do indeed seek out novelty – typically we like to establish a shared set of concerns as a foundation for the relationship, which in turn provides a platform for exploring new experiences which build on these. From the above explanation you can see why this is so important – if you and I share a passion for fine food, I might be persuaded by you to try a new restaurant, since I can safely assume that my new concern can be built on the old, just as was the case with you. Music recommendation engines ape this process – ‘if you like this, then you might like this…’ – the process is simple: track the musical tastes of people and then compare those with significant overlap.

This latter point is the heart of our question: when we are merely ‘establishing rapport’ in a relationship, the focus is not on learning but on discovering things that we both care about – a shared taste in music, in holiday destinations, in food, in football, in movies – this is the foundation on which we can build a relationship.

When it comes to learning, we are trying to shift the concerns of an individual – this is the essence of learning whether we are talking about fruit flies or humans. Even if we take a simple case of classical conditioning, we are building one concern atop another – we shock the animal each time we ring a bell, and it learns to fear the bell. It develops a new concern (fear of a bell), based on the old (avoidance of pain).

In humans, the role of social interaction in learning is considerably more complex – but the same basic mechanism applies. Here we enter into a relationship not merely to establish rapport but to develop an individual.

The first thing to notice is that development cannot take place without sharing the concerns of the learner. Historically humans have developed sophisticated methods for doing this: observation and conversation for example. A skilled mentor will spend time talking to a person, understanding who they are and what matters to them before offering guidance.

In recent years we have seen powerful technological approaches, developed by companies wishing to sell their products but without the additional complication of getting to know each person through conversation. Instead, they are able to map what each of us cares about via our online behaviour. In turn this enables them to sell in an ‘if you like this, then you might like this’ fashion – in a way which follows a similar pattern to learning.

Educational systems have yet to grasp the basic operation of learning. In a future system, the starting point will be this ‘concern fingerprint’: a mapping of concerns, accomplished personally or digitally. It will be near impossible to move an individual in the direction of some new concern if they do not believe that the system/individual shares their concerns deeply.

This is also why children who are the product of extreme or abusive early childhoods present a far greater challenge to a learning system – since it is far harder for another person to share their concerns, and therefore to build the kind of foundation on which to develop a learning path. Someone, at the outset, has to say ‘I know how you feel’ and the child has to believe them.

We can only speculate regarding the role that may be played by machines in this process. A machine cannot share the concerns of a learner, since a machine is not designed to react to any experiences in the way that an organism does. It can, however, accurately map the concerns of similar users at scale – in the way outlined in our music recommendation example. It can say ‘this individual likes experiences A, B, C. Other individuals that liked A, B, C, also liked X, Y, Z with a high degree of probability. I can recommend experiences X, Y, Z to this individual therefore.’

This is significant because it suggests that a person may be able to learn from an interaction with a machine – knowing full well that it is a machine – purely because it is a proxy for human concerns. In other words, it is fine for me to know that the machine does not care, if I know that it is merely summarising the cares of other humans similar to me. Something similar happens when we read a book - strictly speaking, we are learning from the author, not the book.

There is an important difference between a marketing recommendation engine an a tutor however - a recommendation is relatively agnostic with respect to new concerns (at least within a given product domain). It may say 'if you liked this product, here's another of ours you may want to buy.' With education, we have some destination in mind. We probably don't want to create a population entirely comprised of ballerinas.

With education we want to 'steer' learning to some degree. We want to understand how to achieve one of a number of outcomes where an individual is happy and productive - rather than just guiding them down commonly-trodden paths.

This requires a much higher degree of sophistication: a 'learning path' proper has to be a series of stepping stones which take the individual from the things they care about today to the things they will care about tomorrow - without any of those steps being too great a leap. This is the great challenge ahead of us, with respect to education.

But today we are at ground zero. Our first step must be to find out what people care about. Only then can we begin a learning journey.

*photo: Spela Jambrisko