The Calculus of Newcomb's Problem

1TLW

2TAG

1TLW

2TAG

1TLW

1Heighn

New Comment

Newcomb's problem requires either that a) agents aren't Turing-complete (in which case of course agents can get sub-optimal outcomes) or b) Omega is super-Turing, in which case all bets are off. I'm not sure if it's worth focusing on as a result.

One finite, but otherwise Turing complete system can emulate another,. providing it's sufficiently more powerful.

One finite, but otherwise Turing complete system

There is a name for 'a finite analogy to a Turing machine' - it's called a finite state machine. You are correct in that one sufficiently large finite state machine can simulate a smaller FSM.

If agents must be FSMs, case a) applies. Your agents can of course get suboptimal outcomes, and many standard axioms of game theory do not apply^{[1]}. (For instance: a game's expected value can be *lowered* by adding another option^{[2]} or raising the payoff of an option^{[3]}.)

^{^}Or apply in modified forms.

^{^}Once the FSM is no longer large enough to scan through all outcomes.

^{^}Once the FSM is no longer large enough to be able to do the necessary arithmetic on the output payoffs.

In the previous post, we applied some calculus to game theoretic problems. Let's now look at the famous Newcomb's problem, and how we can use calculus to find a solution.

## Newcomb's problem

From Functional Decision Theory: A New Theory of Instrumental Rationality:

First, we need to come up with a function modeling the amount of money the agent gets. There's only one action to do: call this a. Then V(a) is the function for the amount of money earned for each action (one-boxing or two-boxing). However, as you might know, historically, thinkers have diverged on what V(a) should be.

## Causal Decision Theory

Causal Decision Theory (CDT) states that an agent should look only at the causal effects of her actions. In Newcomb's problem, this means acknowledging that the predictor already made her prediction (and either put $1,000,000 in box B or not), and that the agent's action now can't causally influence what's in box B. So either there's $1,000,000in box B or not. In both cases, two-boxing earns $1,000 more (the content of box A) than one-boxing. Then V(a)=B+1,000a, where B is a constant representing the amount of money in box B and a∈{0,1}, where a=0 is one-boxing and a=1 is two-boxing. V′(a)=1,000>0, and of course V(a) returns the most value for the highest value of a: a=1 (two-boxing).

This is all pretty straightforward, but the problem is that almost all agents using CDT ends up with only $1,000, as the predictor predicts the agents will two-box with 0.99accuracy and put nothing in box B. It's one-boxing that gets you the $1,000,000. Enter logical decision theories, e.g. Functional Decision Theory.

## Functional Decision Theory

Functional Decision Theory (FDT) reasons about the effects of the agents

decision procedure(which produces an action) instead of the effects of heractions. The point is that a decision procedure can be implemented multiple times. In Newcomb's problem, it seems the predictor implements the decision procedure of the agent: she can run thismodelof the agent's decision procedure to see what action it produces, and use it to predict what the actual agent will do. In V(a)=B+1,000a, this means that B and a aredependent on the same decision procedure! The decision procedure's implementation in the agent produces a, and the implementation in the predictor is used to determine what B should be. Let's write d for the decision procedure: d∈{0,1}, where 0 means a one-boxing decision and 1 means a two-boxing decision. Then a=d, and B=0.99∗1,000,000−0.98∗1,000,000d=990,000−980,000d. After all: if d=0, the agent decides on one-boxing and the predictor will have predicted that with 0.99accuracy, giving an expected value of 0.99∗$1,000,000=$990,000 in box B. Should the agent decide to two-box, the predictor will have predictedthatwith probability 0.99and only put $1,000,000 in box B if she mistakenly predicted a one-box action. Then the expected value of box B is 0.01∗$1,000,000=$10,000, which for d=1 is represented by B=990,000−980,000d. Great! We now have V(d)=(990,000−980,000d)+1000d=990,000−979,000d. V′(d)=−979,000<0. The lowest possible decision, then, wins: d=0, which gives V(0)=990,000, whereas V(1)=990,000−979,000=11,000.This outcome reflects the fact that it's one-boxers who almost always win $1,000,000, whereas two-boxer rarely do. If they do, they get the $1,000,000 and the $1,000 of box A, for a total of $1,001,000, but the probability of getting the $1,000,000 is too low for this to matter enough.