Home > Less Wrong and decision theory > The Smoking Lesion: A problem for Evidential Decision Theory

The Smoking Lesion: A problem for Evidential Decision Theory

This is part 3 of a sequence titled “Less Wrong and decision theory”

The previous post is “Newcomb’s Problem: A problem for Causal Decision Theories”

In the previous post, I discussed a flaw in Causal Decision Theory due to Newcomb’s Problem which started off a lot of discussion on whether it was rational to one or two box on Newcomb’s Problem. I believe that it is the majority view on Less Wrong that one boxing is the rational decision. I also believe it is the majority view elsewhere that two boxing is the rational decision.

As this is a post for Less Wrong, I will continue to act as if a decision is rational if it leads to the best possible outcome. If you disagree with this then you could think of this exploration of decision theory as an attempt to explore what decision theory should motivate an artificial intelligence – in this case it seems more clear that we would want the agent to make the decisions that will deliver the greatest benefit. This leads us in to an investigation of another decision scenario: The Smoking Lesion.

The Smoking Lesion

The Smoking Lesion Problem is as follows: In the world of the Smoking Lesion, smoking is correlated with cancer but does not cause cancer. Instead, there is a genetic lesion which, if present, increases a person’s chance of smoking and their chance of developing cancer.  The Lesion is either already present or not present. You do not know if it is present in you. The question is this, if you like to smoke (but strongly dislike having cancer), should you smoke?

The desired answer in this situation seems to be that you should smoke – doing so does not increase your chances of developing cancer and it gives you pleasure. What do causal and evidential decision theories do here?

Remember what Evidential Decision Theory looks for – it asks whether a decision would act as evidence for a possible outcome. So smoking would be evidence that you were more likely to have the genetic lesion. Thus, evidential decision theory would advise that you don’t smoke. Causal Decision Theory, meanwhile,  would say that smoking has no causal effect on cancer and so you should feel free to smoke.

The failure of traditional decision theory

So this is a situation where Evidential Decision Theory fails. The previous post noted that Causal Decision Theory fails to provide the correct answer in Newcomb’s Problem. So due to these decision problems, it seems that traditional approaches to decision theories are flawed. These problems, and others like them, are the reasons why people on Less Wrong have focused so much attention on trying to develop new decision theories. The next post will explore other scenarios that can test these new theories.

Appendix 1: Other resources

“The smoking lesion” – entry on the Less Wrong wiki

Timeless decision theory and metacircularity – A post by Eliezer that deals with the Smoking Lesion problem among other topics

The next post is Decision theories and strange scenarios”

About these ads
  1. Perplexed in Peoria
    August 23, 2010 at 12:37 pm

    Causal Decision Theory, meanwhile, would say that smoking has no causal effect on cancer and so you should feel free to smoke.

    I might have worded this differently. CDT doesn’t actually say that smoking has no effect on cancer. Rather, it permits you to make this assumption when setting up the problem, just as every version of decision theory permits you to assume that smoking is pleasurable.
    Once you have fed in your assumptions, CDT lets you do two things. The first is model-theoretic. It allows you to feed in statistical data regarding the frequencies and associations of smoking and cancer, and it will build a consistent model including the (unknown) frequency of the lesion and its causal influences on smoking and cancer. The second thing you can do is decision-theoretic. Given the causal model and your assumptions regarding smoking and cancer preferences, it will prescribe (i.e. recommend) your decision as to whether or not to smoke.
    My main point here is that CDT goes beyond EDT by allowing you to feed in a kind of assumption (“S does not cause C”) which EDT doesn’t even understand and can’t deal with. CDT, for example, lets you express the assumption in a Newcomb problem that your decision whether to one-box or two-box can have no causal influence on your mental state at the time of Omega’s decision, though there may well be an influence in the opposite direction. EDT doesn’t even let you talk about things like this.

  1. August 18, 2010 at 9:51 am
  2. August 18, 2010 at 9:51 am

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

%d bloggers like this: