Skepticus: Prudence, my friend, there was one other dream that the Grasshopper recalled to me, which may shed some light on our speculations.
Prudence: Please do tell.
Andy Clark: Hello, Grasshopper.
Grasshopper: Hello traveller. What is your name?
AC: Andy Clark. I am Professor of Cognitive Philosophy at the University of Sussex.
G: Andy, I enjoyed your paper with Chalmers on the extended mind. Very playful, I might add.
AC: Quite. I’ve come to see you because I think that some speculations you have engaged in with your students are germane to my work. Am I correct to say that you have argued that life is a game we are playing, and that a game is the voluntary overcoming of unnecessary obstacles?
G: That is correct. The rules of a game must also be adopted knowingly, in order to make it possible. This, I believe, is both necessary and sufficient for a game.
AC: That raises the question of what goal life aims at (which you call the prelusory goal), and what rules are adopted.
G: Also correct.
AC: One of the biggest problems with your account is to identify what the goal of life could be, if life is a game.
G: It could be that life is a game, whose goal is different for everyone.
AC: I worry this risks tautology. Just about any behaviour can be interpreted as goal-directed, if you consider arbitrarily specific goals. Even if a person is acting entirely without direction or purpose in life, we might say that she is (without knowing it) acting toward the goal of achieving XYZ mental states in such-and-such order.
G: I see. You are saying that the more complex are the goals, the more we will see goal-directed behaviour (and therefore possibly games) where they aren’t really there.
AC: Precisely. We therefore have strong reason to believe the goal of life is simple. And we also have strong reason to believe the goal of life is shared by many (if not most or all) people. Bespoke goals for each individual are what allows for these arbitrary and ad hoc goals. Finally, I think the goal of life should not be opaque. When people play games, they (normally) know what the goal is.
G: I agree these are desirable properties, even if not strictly necessary. The trouble, of course, is finding such a goal.
AC: Grasshopper, I’ve come to talk with you because I’ve found a candidate. In fact, I believe it’s the only possible goal that life could have which rescues your theory. Karl Friston and others have found that a remarkable range of behaviours can be explained by the theory that the brain has a mental model of the environment, which minimises the difference between anticipated and actual sensory inputs. Life is a game, and the prelusory goal is the minimisation of surprise.
G: I can immediately see the attraction. That goal is simple and shared by all. Your theory also accounts for why the goal is opaque. If the goal of life is to enact a relatively abstruse neurological process, then it’s clear as day why most people would not realise they are playing a game. If Friston’s work came to be more widely known, the goal would become more widely understood.
AC: Interesting – I admit I had not thought of that myself.
G: And what is the rule?
AC: The rule is that the updating of anticipation obeys Bayes’ theorem and the probability calculus.
G: This brings up two interesting considerations. The first is whether the rule you suggest makes it harder to achieve the goal. It is not obvious that conforming with Bayes’ theorem makes it harder to minimise surprise. The second question is whether mathematical rules can ever be thought of as constitutive rules. It would be absurd to say that “the probabilities that a player wins the foot-race, across all players, sum to one” is a constitutive rule of a race.
AC: A Bayesian model clearly makes it harder to achieve the goal compared to some alternatives. Your brain might never experience surprise, and even trick you into thinking that you were always perfectly predicting sensory inputs. I agree that mathematical theorems should not be considered constitutive rules, as opposed to mere background assumptions. But it’s fiercely disputed which mathematical results apply to human belief, and how. If the probabilities of each runner winning the race are ‘out there’ in the world, then, yes, they must sum to one. But as you know, there was a famous debate between Keynes and Ramsey over whether a rational agent must have subjective probability credences which obey the probability calculus. Views on that question are far from unanimous. The question of whether rational belief-updating must occur in a Bayesian fashion is even more contentious.
G: And what is your view?
AC: I am a thorough-going Bayesian. But my point is more that the answer to this question is far from obvious. There are many philosophers, smarter than myself, who believe Bayesianism isn’t necessary for rationality. Ironically, that I don’t defer to them more is rather un-Bayesian of me. I am not such a partisan that I think Bayesian belief-updating is an inalienable fact of the universe, and not a contingent rule that games may have.
G: I am unconvinced. What about the other conditions on the rule – that they are adopted knowingly and to make the game possible?
AC: As for making it possible, if Friston is right, this ‘predictive processing’ is necessary for any complex cognition, game-related or otherwise. So it clearly makes the game of life possible. As to knowing we are adopting the rules, I agree this needs more work. Perhaps we subconsciously know the rules. You recounted to your students the story of Ivan and Abdul, who tried to play a game with no rules. They eventually realised that, by agreeing on a starting time for the game, they were unknowingly adopting a rule. Similarly, we might one day wake up and have explicit knowledge of these subconscious rules.
G: Friend, why do you think that minimising surprise is the only goal that life could have?
AC: If life had a simpler goal, I think it would have presented itself to us by now. From everything you and your students have said, and the surrounding writings, no goal has been proposed which properly minimises the number of problematic cases while preserving the spirit of games. Some have proposed that the goal of life is death, and the rule is ‘no suicide’. I think this fails for the simple reason that players generally want to achieve their goals. Pace Bernard Williams, I would be immortal if I could. Of course, there may be some goal I haven’t thought of. But, if my account works, it does not matter; life would still have been shown to be a game. Life could be multiple overlapping games. We haven’t even discussed the possibility of life being a multi-player in addition to a single-player game.
G: Speaking of minimising surprises, where am I?
AC: You are in a dream, Grasshopper.
G: How splendid… but come to think of it, don’t dreams disprove your theory? If anything, dreams seem to maximise surprise. It is not every day that an ethereal Andy Clark talks to me about my theory of games.
AC: Quite the opposite. Since you love a challenge so much, Grasshopper, can you figure out why?
G: Let’s see… I don’t yet follow why all players in the game of life don’t seal themselves off in a dark room with no inputs. If there are no sensory inputs, there will be no uncertainty either; your model of the world would be perfectly accurate. It seems to me that whatever the solution to this problem is, is also the solution to the problem of dreams.
AC: That’s right.
G: And if this is a dream, then, almost by definition, I already know everything which you can tell me. I must be playing a game with myself, in which this dialogue is a more challenging way to recall facts and arguments which I already know.
AC: Didn’t Plato teach us that all learning is re-learning?
They both laugh.
G: I think I’ve figured it out. In your account, the purpose of dreams is to generate experiences which are varied and unusual enough to minimise the overall rate at which your models of the world fail to correctly anticipate the senses.
AC: Yes!
G: The man who seals himself off in a room with the lights off – or who does not dream – fails to play the game of life.
AC: Exactly; he is like a boxer who, with the goal of causing a knockout, resolves to shoot his opponent in the head.
G: If your account is correct, it follows that dreams should end whenever they become too predictable, and stop testing the limits of your model of the world.
Andy Clark gives a smirk, and the Grasshopper pops out of existence.