Friday, February 10, 2023

What Is To Be Done?

We are constantly faced with decisions about what to do, at every scale. Oranges are on sale: how many should I buy? Fossil fuels are creating climate havoc: should we switch to nuclear fission power?

In the simplest situation, one can foresee with sufficient accuracy the results of each alternative action, and choose the one with the most positive result. This formula outlines some major dimensions of a decision-making situation. One needs a set of alternative actions from which to choose; one needs to understand the results of each possible action; one needs to evaluate each of these possible results.

Commonly enough it is not possible to predict accurately the results of actions. We must decide in the face of uncertainty. We might have a pretty good idea about the probilities of the possible results of each possible action. For example, in a card game, we can calculate quite accurately the probabilities for each combination of cards we might draw from a well-shuffled deck. When developing a financial plan for living in retirement, actuarial tables can give reasonable estimates for survival to whatever age. Comparing the uncertain results of various possible actions is quite difficult. Given a choice between one action whose result is a certain $1, against another action whose result is $0 with probability 99% and $100 with probability 1%.... the expected value for each action is the same, $1. Whether to buy a raffle ticket for $1, that is a choice where expected return is not going to sufficient information to make a decision.

Many real world situations do not permit probability calculations with any realistic promise of accuracy. Probabilities are applicable in situations that repeat. Of course situations never repeat exactly, but a large number of situations can be similar enough so that the outcomes of each possible action can be tabulated to provide guidance for what to do when the situation occurs yet again. But sometimes situations don't repeat with any reasonable similarity. What's the probability that Donald Trump will be elected President in 2024? Of course one can assign this whatever probability seems appropriate, but there is no way to check this number against the facts. In situations like this, one can look at the set of plausible outcomes of each possible action. An action might turn out well, or might turn out badly. How well? How badly? Comparing these sets of plausible outcomes is not simple or mechanical, but that's what's required for deciding on what action to take.

Sometimes a decision involves a significant action that takes place essentially at a single point of time. For example, if I am considering a major purchase, at some point I have to signal my decision to complete the transaction. But oftentimes what is called for is an ongoing series of actions. There is deciding what to cook for dinner tonight, and then there is deciding on my diet, on my pattern of meal selection. I don't have to plan out my meals for the rest of my life; I can decide on meals more or less on the spot, depending on my schedule, my activities, the availability and prices of various food items, etc. In a game like chess, there is no way to plan out the full sequence of moves one should make in order to win. Each move must take into account the preceding moves of one's opponent, which cannot be predicted with anything like sufficient accuracy. One can, however, potentially decide on a strategy. A plan is a sequence of actions. A strategy is like a table of possible situations that might arise in the future and what action to take in each situation. Market orders versus limit orders in the stock market would be an example. A market order is the decision to buy or sell some number of shares. A limit order is conditional: whether any shares are bought or sold depends on the market price. A market order is a plan, a limit order is a strategy.

Deciding on a strategy can be very difficult. It can be impractical or impossible to tabulate all the possible situations that can arise in the future. And when one encounters a situation in the future, one might choose a quite different action than whatever had seemed the wisest back when one was contemplating future possibilities. Our understanding of actions and outcomes evolves: we are always learning, or at least we can be learning. So an effective strategy for action is one that enhances the quality of one's future decisions, by providing opportunites for learning along the way, and leaving open as wide a range of possible actions in the future as possible.

We should not be planning to imprison ourselves; we should be planning to liberate ourselves.