Review of "Rationality: AI to Zombies": Religion as a Framework

If you prefer to listen rather than read, this blog is available as a podcast here. Or if you want to listen to just this post:



Or download the MP3

---

Last week I started my review of Rationality: AI to Zombies (RAZ) by Eliezer Yudkowsky, with a post comparing Bayesian Rationality to Talebian Antifragility. My conclusion was, that though they both represented very useful frameworks for understanding the world and making decisions under uncertainty, I concluded that antifragility was better at working with the world as it actually was, and especially, with handling the impossibility of predicting the future.

This week I’d like to compare the framework of Bayesian Rationality to yet another framework for making decisions under uncertainty, religion. And interestingly enough Yudkowsky’s book, despite its disdain for religion, gives us an interesting jumping off point for showing exactly why religion is necessary.

To begin with, the book’s ostensible purpose is to educate the readers in the aforementioned framework of rationality. And both the A to Z title and the length of the book (equivalent to 2393 pages) gives one the expectation that they’re in for a fairly exhaustive education, at that. If you start looking into the history of the book, there is one other reason to expect a step by step education in the art of rationality. When these essays first appeared they were called the sequences. And they formed the backbone of the website LessWrong.com. It was in this form that I first encountered them, and, in my completely subjective recollection of things, they were spoken of almost reverently. Only later, were they compiled into a book. And despite the preponderance of atheists among the ranks of the rationalists I don’t think calling this collection the rationalist bible is that far from the mark.

Perhaps you’re not willing to grant it the role of “rationalist bible”, even so, you have to admit that the word “sequences” is evocative. It certainly suggested to me that if I just read the articles in “sequence” that they would teach me everything I needed to know about Bayesian Rationalism. Well I did, and they didn’t.

In the end, as I mentioned last time, it didn’t ever entirely rise above the fact that it was still just a collection of blog posts. Now don’t get me wrong these are some great blog posts, and when grouped around a specific focus they do a good job of explaining many things, but I don’t think, when the book is considered in its entirety, that they ended up being as “sequential” as I had hoped, nor were they as comprehensive. There was a lot of fluff in there, as you might imagine.

You may be wondering when I’m going to get to religion, and there may even be some atheists out there who already think they know the connection. Having mentioned that the book represents something of a rationality bible, (even going so far as to toss in the word “reverently”) and knowing, as we all do, that rationality is one of, if not the, primary belief system of atheists. They expect me to do that thing where I mention that both rationality/atheism and religion are just systems of belief, and now that I’ve show that rationality has a bible, and a group of followers, and a bunch of beliefs which will perfect them, isn’t it, then, just one more religion? And, in fact, this is precisely NOT what I’m going to do. No, that would be far too lazy. Which is not to say the comparison is entirely without value, particularly when you start talking about Transhumanists and “The Singularity”. But that’s not where I’m headed. Rather, what I want to do is compare how rationality does when compared with religion in terms of educating people in their respective frameworks.

If, as I claim, RAZ’s ostensible purpose is to provide an education in the art of rationality, how well did it do? Well, I can only speak for myself, but having read it, I don’t feel very educated. There were some interesting bits that one couldn’t get just from reading Thinking, Fast and Slow, by Daniel Kahneman,  but fewer than I expected. Mostly I feel like, while I do have a few new tools, I didn’t get much in the way of new ideologies or frameworks. And if I had decided to embrace everything in its entirety I’m not sure I would have been able to describe exactly what it was I was embracing. (Winning? Or is that Charlie Sheen’s Philosophy?)

Now I have said before that I’m a bear of very little brain, so it is possible I’m missing something, or that it took me long enough to read the book that there were chapters in the earlier sections of the book which precisely answer this objection and I have since forgotten them. But humility aside, I did read the book, how many people can be expected to do that? All of which is to say, that if, after finishing a 2300 page book, I still don’t feel like I have a framework to draw on, how much harder is it going to be for the vast majority of people to acquire this framework who are never even going to hear about the book, let alone finish it? Rationality is not some system which naturally fits in with human biases and desires, in fact it’s the exact opposite, the rationalists want you to overcome your biases. Absent some technological “rewiring” of humanity, how effective could this ever be? To put it more simply, if only 1% of the population (and probably much less than that) can understand and incorporate the sequences, then what do we do with the other 99%?

And this is where I want to start my comparisons between rationality and religion. As I have said, both are frameworks for behavior. They both want to take the default individual and improve them. And if you’re on the outside looking in, the obvious question is: which one is better? There a ways to examine this question and varying comparisons we could make, but let’s start by just looking at things from an individual level.

Rationalists, particularly those with atheist or agnostic leanings would almost certainly argue that an efficient rationalist is a better person than an obedient Christian. (I am using Christianity as an example, because that’s what I know best, but I think most of my points would apply to the other religions as well.) I expect that as part of their argument they might make the claim that the efficient rationalist would do a better job of allocating resources, for example: choosing which charities to support. I assume they might further argue that the rationalist would be free from the harmful prejudices of the obedient Christian, for example: they would be less likely to be homophobic or racist. I’m sure that on top of the few obvious things that they would argue much more besides. I actually don’t think this is true. I would contend that the very best Christians are probably better people than the very best atheists, even if one was to use impartial standards (like charitable giving, or community involvement.) To begin with, I’ve met some truly amazing religious individuals, and secondly, recall that because of how many religious people there are, as compared to the number of rationalist atheists, that even if the percentage is smaller, the absolute number will still be much greater, and with that bigger pool, the best example is probably much better.

But let’s set all that aside, and for the purposes of this particular argument grant that the very best rationalists are better than the very best Christians. That if both are given their purest expression, that rationalists are better people. Even if we imagine this is so… So what?

I believe a lot of rationalists and atheists and secularists of all stripe, believe that because something appears to be working for them, that it should not only be equally effective at working for other people, but that if extended to the society at large that everyone would be better off. I certainly see where they get this idea, but there’s no law that says that ideologies have to scale.

There are in fact several factors involved in creating an ideology that works for an entire society. One which doesn’t just result in a few good people, but in good people and good outcomes at all levels. The first and most obvious of these factors is that it has to work. It has to produce better individuals than you would get otherwise. Many of the non-religious individuals we have been talking about are going to claim that we don’t need to go any farther because religion doesn’t even do this. That whatever else may be said about it, it doesn’t work. Are you sure about that? You shouldn’t be.

In fact, far from being a flawed framework which needs to be replaced, there is significant evidence that religion has a broad positive influence. Which is not to say there is no disagreement, but when outlets from Forbes to The Huffington Post tout the advantages of religion, it’s not inappropriate to ask if maybe we already have access to a framework that’s working. Which would make sense. As I have discussed in the past, many people want to view religion as a collection of meaningless superstitions which are either actively harmful or of no benefit whatsoever, but I find that argument entirely unconvincing, particularly when many religious traditions appear in nearly identical forms across nearly all cultures, regardless of how different they are in other respects. Also religion is so ubiquitous, particularly if you go back a few decades that are you sure you know what people are like in the absence of a religious framework? Finally, on top of all the other benefits which people claim for religion, I would like to add the fact that religious people have more children, which is a long term guarantee of success which doesn’t get nearly enough attention. (As they say the future belongs to those who show up for it.)

The second factor to consider, and closely related to the first, is how well does it work on average, ignoring the exceptional cases? For our purpose we need to ask, what does an average rationalist look like? Here rationalists are at a severe disadvantage. To a reasonable approximation, there aren’t any average rationalists. To even consider rationalists as a category, we’re already restricted to looking at only exceptional individuals. (Whether they’re just exceptions from the norm or truly exceptional I leave as an exercise for the reader.) Thus any evaluation of society-wide impact has to start by determining how common they even are. What standard would we use to declare someone a follower of rationality? Are these people who’ve read the entirety of RAZ? People who’ve ever visited LessWrong.com? People who’ve read Thinking, Fast and Slow? The only remotely authoritative numbers I can find would be for the last category. But my best guess for each would be five figures, six figures and seven figures respectively. Which puts the percentage of rationalists at either less than 3/100ths of a percent, 3/10ths of a percent, or 3 percent, depending on which standard you want to use. With the number certainly being on the lower end of this range, since even if you were so bold as to declare every reader of Daniel Kahneman a rationalist, the number of books which were read only loosely correlates to the number of books which were sold, particularly for books like that. (In fact this exact book showed up very high in a list of books that people started reading but never finished.)

Let’s compare this to the percent of religious people: 70% of Americans are Christian to one degree or another with 53% of all religious people declaring that religion is very important. Meaning by any conceivable measurement, and no matter how optimistic you are about the number of rationalists, they’re completely overwhelmed by the religious. This should give you a sense of how large of a task the rationalists face before their society-wide impact can even begin to approach the society wide impact of religion. To look at it from a different angle, if you can improve religion by 1% it will almost certainly have a greater impact than increasing the number of rationalists by 100%.

All of which takes us to the final factor I want to consider, and the one I started, with how accessible is the ideology/framework of rationality as espoused by RAZ? I think it’s already clear that my opinion is “not very”. But having already looked at the exceptional individual in our exploration of the first factor and at the average individual in our exploration of the second factor for this final factor I’d like to look at how the two frameworks work at the very bottom of the spectrum. For this group let’s use prisoners as our representative sample. Here I would expect very little argument that religion is more effective here than Bayesian Rationality. But once again data of all kinds is particularly scarce, though it is interesting to imagine how many prisoners would fit into the three buckets I mentioned earlier. I’m guessing the number of people in jail who’ve read RAZ might be in the single digits. Particularly since it’s only available electronically.

The one piece of data we do have is somewhat ambiguous. We have some limited evidence that atheists are less common in prisons than in the general population. And if we use atheists as a rough proxy for rationalists than that might give us something to work with, but it cuts both ways. One could view it as proof that atheism keeps people out of jail, or alternatively, that the majority of prisoners, having hit bottom turn to religion for redemption. I obviously favor the latter view, though it’s a wonder there isn’t any more data on this. I would think it would certainly be interesting to know if very religious inmates end up with a lower recidivism after release. Or if giving prisoners intensive courses in bias detection and statistics would make any improvement. (I suspect not which is why I favor the religion as redemption view.) One thing I have noticed as I’ve been looking into this subject, is that research on the positive social effects of religion has fallen out of fashion with most of it being done decades ago.

In any event, we could wish for better data, but given the enormous amount of anecdotal evidence, I choose to assert that religion is uniquely effective with those at the very bottom of the heap. You could assert otherwise, and you could choose to believe that all we need to do is have every inmate read a 2000 page book and they’d be far better off than with any religion, but I’m not sure even Yudkowsky believes that.

The point of all of this, is that for a framework to be successful it can’t be something that is usable only by the elite (unless you’re really in favor of a tyrannical oligarchy.) It’s not enough for the very best follower of your ideology to better than the very best follower of a competing ideology, it has to be something which can be understood by and which resonates with everyone from the the lowest prisoner to the greatest king. And religion does this. People criticize religion for being simplistic, but that’s a feature not a bug.

As I said in beginning, if only 1% of the population has the intelligence and inclination necessary to understand Bayes Formula, or the Availability Heuristic or why Yudkowsky feels so passionately that the many worlds interpretation of quantum mechanics is superior to the Copenhagen Interpretation (and that’s more of a 0.01% thing), than Bayesian Rationality with RAZ as it’s bible is never going to succeed. Compare this to religion which has rules like don’t kill, only have sex with your wife, whatsoever ye would that men should do to you, do ye even so to them, etc. Things which anyone can understand and which also lead to good outcomes.

Is religion, or specifically Christianity perfect? Far from it. Indeed it could be said that religion is the worst framework for correct behavior, except for all those other forms that have been tried from time to time, to adapt the old Churchill quote.

In short, what I’m trying to get at is that we have various frameworks for trying to get people to behave well, to avoid bad things and do good things. Rationality may have better rules than religion, and it may even have better outcomes than religion, but after having read RAZ, I feel I can state with certainty that 99% of people are never going to bother with it. And if that’s the case then how much better is it really?

If it’s not clear, this post is less a criticism of RAZ than a defense of religion. Creating a framework to encourage good behavior and good outcomes is hard, and RAZ and the associated ideology are interesting attempts. And to a very great extent Yudkowsky and his co-ideologists should be applauded for even attempting it. But it is worthwhile to consider that maybe, just possibly, all of the religious people who preceded us, the 99.9% of all humanity that was religious, were just perhaps, not all low-IQ, bigots, who were full of hate, but rather doing the best that was possible. And that when everything is taken into account, and when all of the factors are considered, that religion is already the best ideology for creating good outcomes and good people?

---

You know one way of proving you’re a good person? Donating to blogs you enjoy. No, really.