What is fun? Fun is a rather nebulous phenomenon. People have wildly varying ideas of what activities are fun, yet we all know what fun feels like. Are things that are funny also fun by definition? If not, why do both fun and funny things tend to incur smiles and laughter? Specifically, why is improv fun, and what is the experience of fun actually for? This week, I’ve found myself trying to answer these questions. I think I’ve done it. And furthermore, the answer I’ve discovered is striking enough that it’s going to change how I train people from here on forward.
My adventure in search of the origin of fun started with the books on behavioral economics I’ve been reading recently. (More on that in later posts!) A lot of the ideas I came across reminded me of topics I’d encountered in my work on machine learning over the years. Out of curiosity, I found myself trying to build a kind of map to explain human emotions and social habits in terms of machine learning goals.
To provide a basic example, we can compare human pleasure and pain with the positive and negative feedback signals that you feed into a neural network to tell it whether it whether it got the answer to a question right or wrong. If we believe that this approximation is fair, we might then describe fear as what a learning system experiences when most of the candidate models of the future it can build at a given moment entail receiving negative feedback. In other words, the system has a high expectation of something painful happening to it. Excitement, by contrast, is what a learning system experiences when the most likely future models it can build feature positive feedback. No big surprises there, but I’ve gone on to try to play the same game a whole range of human emotions, including anxiety, achievement, guilt, pride, embarrassment and frustration.
Intriguingly, there are certain kinds of human experiences that are a lot harder to reshape in AI terms than others. Love is a good example. The best explanation of love I’ve encountered is the one in Stephen Pinker’s ‘How the Mind Works’, where he describes it in terms of Game Theory. Romantic love, he suggests, is an abandonment of rationality that advertises willingness to engage in a pair-bond. General-purpose AIs don’t have much of a reason to engage in pair-bonds, so we can’t expect much love from them. However parrots bond for life and we see behavior that looks a lot like real love coming from them, even though they’re not human.
The most puzzling such emotion, though, of all those I explored, was fun. Clearly fun is something deep and important, because you see something that looks very like it in dolphins, chimps, dogs and the young of almost any mammalian species you care to choose. Similarly, it’s hard to define because if you ask a sample of human beings what’s fun, you’ll get wildly varying answers. For instance, a significant fraction are likely to include ‘shopping’ on their list, whereas many others would rate the same activity as a kind of torture. Fun doesn’t obviously map onto any kind of planning function you find in machine learning, and it doesn’t appear to maintain or moderate any kind of social behavior. It doesn’t even appear to be easily describable in terms of Game Theory. So what’s going on? The answer, I suspect, lies in who craves fun the most, and how they get it.
Clearly, fun has something to do with play, and play has something to do with being young. Play, also, has something to do with simulating experiences. Otherwise, why would children bother playing ‘Families’ when there’s usually ample experience all around them? Fun is also a signaled emotion. It comes with big obvious indicators that can be read between members of social species in an instant. Apes smile and dogs wag their tails. By contrast, achievement, also a powerful sensation, comes with much more subtle cues. Pulling an ‘achievementy’ face is perhaps possible, but a lot harder.
What I would hypothesize, then, is that fun is a specific, dedicated emotion designed for learning. It’s associated with play, and play is the process through which that learning is acquired. (It doesn’t look like I’m the first person to have this idea. In fact, a little web research suggests that Dr. Stuart Brown has been onto this sort of thing for a while. I have his book ‘Play’ on order from Amazon.) Play learning happens through experience simulation, and the primary characteristics of play appear to be that the costs and gains to individuals are reduced to a level at which new behaviors can be explored without the players damaging their social standing by taking part. Fun is strongly signaled to inform others that experimental behavior is being tried out and that normal social costs shouldn’t be applied. Additionally, fun takes different forms for different people because the skills each person wants to practice are different, and likely to line up tightly with the attributes that person values about themselves.
The same model can be used to explain why something is funny. Various philosophers have tried to capture the essence of humor, and their theories can be roughly boiled down as follows:
* Superiority: We laugh at something because we feel safely above it. (Hobbes)
* Incongruity: We laugh at something because it breaks a pattern. We experience ‘frustrated expectation’. (Schopenhauer)
* Relief: We laugh at something to break the sense of tension over an experience. (Freud)
The ‘experimental zone’ theory of play would seem to encompass all these. We decide that an experience is funny when we decide that the net social cost of a statement or idea is negligible even though we might not understand it. We flag that we’re safety-boxing that experience by laughing at it. Applying this model to the theories above, we get the following:
* Superiority: When we feel safely distant from an experience, we signal it.
* Incongruity: When we experience an unexpected pattern break without cost, we signal it.
* Relief: We signal safety to indicate that a tense experience is being considered trivial or experimental, and therefore no longer dangerous.
Laughter has the added bonus of informing those around us that they’re part of a safe group who’re parsing an otherwise dangerous experience in the same way. Hence its infectiousness. Laughter becomes derisive when it’s clear that the boundary of who’s safe includes some individuals and not others.
This model adds up with many personal experiences I’ve had from improv training. For starters, people don’t really start to take on new behaviors until they feel safe. And that safety is defined by everyone in the group agreeing through laughter signals that the experiences they’re having shouldn’t be taken seriously. Also, certain games invariably work better than others, and those that work well seem to be the ones that reinforce that sense of a ‘safe space’ in which nobody is excluded.
The implications of this for teaching, I believe, are significant. Most importantly, I’d recommend not expecting people to try out and take on new behaviors unless they feel they can do so safely. This is why so many ‘role-play’ sessions feel cold and cringey. You know when people feel safe because they’ll signal it to you. Furthermore, rather than being an indicator of ‘mucking about’, laughter and play are the indicators that the brains of the players have activated their dedicated learning mode and are ready to incorporate new data.
From the perspective of soft-skills training specifically, the lesson go deeper still. It’s always tempting as a trainer to illustrate key points to participants by having them directly experience the effects of their own behavior, or the effects of the assumptions they make, through striking, unexpected demonstrations. The ‘experimental zone’ theory of fun suggests that we should only do this so long as it doesn’t impair the sense of safety that we’ve built. Otherwise, the learning potential of the experience is likely to be lost, and it may be remembered for all the wrong reasons.
In short, there seem to be solid reasons to believe that fun should be our compass in guiding us to develop training that produces real, persistent, and valuable change. If anyone has any counter-theories, though, I’d be delighted to hear them. This topic continues to fascinate me. From a relatively innocuous starting question, the results so far feel surprisingly profound.