(no subject)

Dear Joi,
Unsurprisingly, it has taken me a long time to try to construct a more precise model of tradeoffs between money, energy, information, love, etc. (Unfortunately, I have lots of work to do for my day job.) Here you go, though. Much of this will be impressionistic. I'll flag with a * the things that are mathematically precise or can be made so (I won't put in any math, though!).

I would like to make more precise the part of our conversation that had to do with narrative. You made the point that classical economics is based on a rather bald and unconvincing narrative (to paraphrase Bertie Wooster) that the only thing that is good is money, and the only thing better than money is more money. Only a very few obsessed people operate with this principle as the =ole basis for their narrative, however. Most folks construct a narrative =n which to base their behavior out of a variety of different principles =nd 'sub-narratives.' (Mind you, I'm a little uncomfortable with the word 'narrative' itself: I may have mentioned to you during our =onversation that I was negatively impressed with a remark that Edward Said once =ade to the effect that 'The problem of the Palestinian people is that they don't have a narrative.' Perhaps more to the point, they don't have food or land or schools.) But the stories we tell ourselves do form the basis for the decisions we make: this is where classical economists fall down in thinking that decisions are made with only money in mind. And in =act, if we look upon narratives as the basis for the thought process behind the making decisions, then there is quite a lot of mathematically precise stuff that can be said about them.

Suppose that someone (a person, a dog, a computer) is faced with a yes-no question: Do I buy this suit? Do I pee on the rug? Do I crash? (I'll let you decide who asks which question.) To make any such decision requires weighing a number of factors, or =nputs, in the process of making the final decision. For example, I need a =uit, but I don't have much money this month; I like dark suits and this is =n the light side; on the other hand, the geometric pattern is great; =uch a suit is unlikely to show up again. Or: I really need to pee, but =11 get in big trouble if I do it on the rug. Or: I am overwhelmed with conflicting requirements and many tasks to be performed =imultaneously; is there a way I can find to schedule them, or not? • One can think of the reasoning process that goes into attempting to make a decision as the process of constructing a reasonable narrative whose conclusion is the result of the decision. (E.g., I can't make it outside in time, so the rug it is.) From the perspective of the person making the decision, the =onclusion to the narrative (the yes or no) is not determined until the narrative itself has been constructed. (From the perspective of someone else, of course, the narrative may have a foregone conclusion. That damn =og: it always leaves it until too late. Stupid computer! Seth is =ncapable of buying a suit.) • This feature --- the being that makes a decision can not in =eneral know the answer beforehand --- is a reflection of a mathematical fact: * no physical system, whether human, computer, or dog, can model its =wn full behavior any faster than the behavior itself. One can construct fragmentary, incomplete models of oneself that capture some aspect of one's behavior. But to construct a full model of oneself requires at least as many physical resources (atoms, energy, time) as one =ossesses. In other words, the only complete model one can make of oneself is oneself itself. • This self-referential conclusion is the basis for a number of famous mathematical theorems, including Goedel's incompleteness theorem and the halting problem. But its primary expression in everyday life is the undecidability of decisions before they are made. • What does this have to do with the existence of conflicting narratives? Well, one of the main reasons that a classical economist makes a highly oversimplified model of human behavior is to render that behavior predictable within the model. Once one takes into =ccount love, religion, a sense of duty, sheer orneriness, and the rest of the sub-narratives and features of human existence out of which we construct our behavior and which form the basis for our decisions, then not only can we not predict what we're going to do, we can predict hardly any human behavior at all. • Indeed, I think that it can probably be established =athematically that *any theory that is sufficiently simple to allow the easy =rediction of human (or animal, or computer) behavior, is too simple to be =redict that behavior correctly; and *any theory that allows the correct prediction of behavior is sufficiently complex that its predictions cannot be evaluated in a closed form. This is basically a consequence of Goedel's theorem: mathematical theories have a threshold of =omplexity; once you're above the threshold, the theory contains statements that =re true but cannot be proved to be true, statements that are false but cannot be proved to be false, and statements that can be taken equally to be true or false, with no proof either way.

In the case of human behavior, this intrinsic undecidability arises from the construction of narratives out of a variety of =ifferent subnarratives and inputs; many possible self-consistent narratives can be constructed from what one knows and what one believes. But these different narratives, though each internally self-consistent, need not be consistent with each other. Indeed, the different possible narratives thus constructed can wildly contradict eachother and lead to radically different conclusions.

OK, that's probably enough abuse of mathematics in the service or disservice of social observation for today. I hope that you are doing well and look forward to having lunch again one of these days.

Yours,
Seth

Thanks Jeffrey. Jeffrey was talking about you in the context of Bounded =ationality which relates to an ongoing conversation I've been having =ith him. I was digging through old email and I found the following.

Now that I'm at MIT, I'd love to reconnect and see how your thoughts =ave evolved in the last decade and share some of my thoughts.

• Joi

seth ,joi, i think you two will like each other