## Monday, April 23, 2007 ... //

### Hartle & Srednicki vs. typicality

Jim Hartle and Mark Srednicki explain that all details of the assumptions about "typicality" - an inherent part of the anthropic reasoning - should always be explicitly stated and some of these assumptions lead to ludicrous conclusions.

They use the Bayesian reasoning to explain that a theory in which the observers like us are more typical are not preferred. If you click at the link in the previous sentence, you will get my article with a similar argument (search for "huge landscape" to get to the main assertion).

Some of their examples of ridiculous conclusions obtained with the assumptions of typicality are related to the Boltzmann brains. Because I clearly agree with their assertions, let me paraphrase the main points from their first page:

• Predicted atypicality of humans is not enough for falsification of a theory
• Deciding about validity of theories mustn't depend on counting of intelligent observers and their properties because this would kill objectivity
• All accessible reliable data may be used and should be used unless it can be shown that they are inconsequential
• Predicting a higher or lower number of copies of observers like us and observations like ours doesn't modify the degree of validity of a theory
• We shouldn't assume that something about us was created by a random process unless there is evidence for this assumption
• Bayesian inference may be used to guess measures that may be implied by a fundamental theory but don't have to
Some people assume that we are randomly selected from a class "C". Hartle and Srednicki call this misstep "selection fallacy" because its essence is to neglect some known data.

Selection fallacy is a special example of political correctness applied to the class "C". Needless to say, if you're a member of the complement of "C", e.g. a conservative, you will be in trouble. ;-)

Figure 1: Jovian atmosphere. Should the predicted number of exo-women of color in this environment influence the selection of the right compactification before we actually count them? ;-)

They say that a theory that would predict that life thrives in the huge atmosphere of Jupiter shouldn't be disfavored before we know whether they do. I added the "before..." part although they don't say it explicitly at the beginning of their paper. Indeed, I think that once we learn that there's no life on Jupiter, a theory predicting life on Jupiter should face some backlash.

The black list of people who are making the error of thinking that theories predicting localized life in Jupiter's atmosphere are already disadvantaged includes Page, Dyson, Kleban, Susskind, Bousso, and Freivogel. ;-) I agree with Hartle and Srednicki that all these physicists have used these logically flawed arguments.

Hartle and Srednicki also try to define "us". In fact, we are "IGUS" - information gathering and utilizing system. ;-) IGUS must be extended once we start to collaborate on scientific theories with aliens. It sounds funny but I think it is very true. We shouldn't be imprinting our properties into our theories because the resulting theories should be valid objectively, even for numerous aliens who may be living on very different planets. Our theories shouldn't be anthropomorphic even though some of the proposed theories are anthropogenic. ;-)

I've been saying it for some time.

Returning to the Jupiter example, they show how the wrong reasoning may proceed - remarkably similar to some reasoning used by the landscape community. For example, you could say that because both authors of their paper are human, theories that allow lots of non-human intelligent beings should be suppressed not by a small number "P" but even "P^2". That's clearly false because the probabilities of Hartle and Srednicki being human are not independent, for example because both of them share the same monkey ancestor. ;-)

Moreover, there exists a simpler argument to show that the typicality reasoning disfavoring theories with life on Jupiter is flawed. It's simply because the numbers of intelligent beings on the two planets are causally disconnected quantities - as everyone familiar with the history of the Solar System should agree - and there has been no thermal equilibrium that would try to compare the human quotas on the two planets. In their example, it is very clear that no random selection process that would try to compare citizens of Jupiter and Earth has taken place.

In some other examples it may be harder to show that such a process didn't exist - and it might exist, after all - but we should never assume that we know that the random process existed because it is not a fact.

The priors may be chosen in such a way that the theories without life on Jupiter will be favored, after all, but these priors seem to have the desired conclusions inserted as input which is a kind of scientific misconduct.

They also modify the example with counting beings on Jupiter to the case of counting cycles and time in various cyclic cosmologies or eternal inflation. Again, they argue that it doesn't matter whether a Universe with the observed properties is predicted to be more frequent or less frequent by a cyclical or eternal theory. The only thing that matters is that the theory allows the Universe with our observed properties to occur at least once. Whether there exist observers who see something else than we do is irrelevant for our procedures of validating our theories and for our predictions.

Finally, they present the different "measures" as different choices of the priors and argue that the right way to refine these priors is to present evidence and adjust them by Bayesian inference, not by inserting more personal preferences. One must carefully distinguish prejudice (priors) from logical deduction (including probability calculations) and from facts (data).

I think it's hard to disagree with them but I am sure that some people do.

And that's the memo.

#### snail feedback (0) :

(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){ (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o), m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m) })(window,document,'script','//www.google-analytics.com/analytics.js','ga'); ga('create', 'UA-1828728-1', 'auto'); ga('send', 'pageview');