Problem Class Dominance In Predictive Dilemmas

So I finally finished my honors thesis for Barrett. Check it out!

I think there is a video of the defense somewhere which might be very helpful for a lot of people. I'll try and get a copy to update this post with.

Morality and Economics

So I realized recently that I keep trying to make moral decisions but my model of morality is insufficiently precise to really do so on a consistent basis. I did realize that I had an intuitive model that I was using, but I wasn't really sure if it actually made any sense at all. So I decided to look into it further and try to tease out what was going on. I ultimately ended up at a concept of morality as a form of economic calculation (Please be aware that "economic calculation" means a lot of different things to different people, so that particular phrase might mean a very different thing to you than it does to me).

Take a minute to consider these two questions. I think they are actually interrelated:

First of all, what does it mean to say that "Bob should not murder"? Is it simply a statement of preference equivalent to "I would like it if Bob does not murder"? Is it something more? If it is something more, what more is there?

Second, why the hell should Bob care? Let's assume that it appears Bob would legitimately be better off if he murdered someone (Imagine he kills a random stranger with a remote button press. He never feels guilty about it and all his wildest dreams come true because he did it). Also assume that we cannot catch him (maybe the kill signal was scrambled through a bazillion nested routers or something). Should Bob ever behave morally against his own self interest? Why on earth would he?

Really give these some thought. Come up with a model of morality, make up some hypothetical situations and see how your answers hold up.  Even better, try testing them on a real world scenario. My goal here is not to force feed you the tentative answer I've come up with, but to get you genuinely curious about these implicit rules that govern nearly all the decisions we make.

Ok, ready? Here's what I came up with.

The meaning of "morality"
I think that we do mean more than a simple statement of preference when we say that Bob shouldn't murder. Otherwise we'd just say "I don't want Bob to murder" and leave it at that. Of course, there is the problem that sometimes when people want something they try to make it into a moral issue.

This reminds me that I'm getting hungry writing this blog post for you.

Give me a cookie or you are clearly evil.

See what I did there? Of course, if I was starving to death in front of you things might be different. Or maybe this is legitimately just my way of telling you that I really, desperately want a cookie right now.

So the problem we face is that normative statements are a lot like "love." In everyday usage, they are wonderfully and terribly ambiguous at the same time. We need something more accurate if we are ever going to be able to talk about this at the level of precision I feel would be useful. At this point I could try to forcefully redefine or even coin a bunch of new words to talk about things, but it turns out I don't feel that is at all necessary or helpful. Why? Because fortunately I decided to study economics and my predecessors have already made up a bunch of words (and even math!) to talk about preferences and their interactions within different systems.

Why Economics is Useful Here
 It seems to me that we generally mean that Bob is somehow making a mistake when he murders someone, and that if he knew better he wouldn't do so. That's why the statement that it is "wrong" for Bob to murder has extra force beyond just expressing a desire on our part. I believe this is because morality is a way of describing an ideal game of life (in the game theoretic sense), and that our moral intuitions are largely ways of approximating this solution.

I assume everyone is familiar with the prisoner's dilemma situation. It generally seems "right" for both people to cooperate, even though we know that using our current decision theory both will defect without enforcement or repetition. Likewise, the situation where all agents are forced to cooperate seems better than the one where all agents are (reasonably!) defecting. If I was in a giant game that was similar to the prisoner's dilemma in this way, I would be willing to pay to force everyone to cooperate all the time. If all agents are identical, they would also be willing to pay to force everyone to cooperate. Thus, a structure where everyone cooperates would be pareto-superior to one where everyone defects. The pareto-optimal structure for society as a whole is thus the moral one.

This is why Bob might want to refrain from murdering even though it seems like it is in his self interest to do so. He might benefit from murdering in this one particular case, but he would probably be worse off in the world where everyone murders someone when it is in their short run self-interest.

This seems to be what Immanuel Kant was getting at with his categorical imperative:

Act only according to that maxim whereby you can, at the same time, will that it should become a universal law.

Or Rawl's veil of ignorance, where we decide social rules behind a veil that ensures

...no one knows his place in society, his class position or social status; nor does he know his fortune in the distribution of natural assets and abilities, his intelligence and strength, and the like.

This also means that if we want to talk about what is moral, we are really talking about pareto-optimal game structures given everyone's existing preferences. Economics is the science and theory of value system interaction, and the optimal game structure under this model is morality. So the best way to tease that out (at least at the meta-level) is going to be using economic reasoning.

But wait! I've made a mistake in the above. I said: 

"the optimal game structure under this model is morality"

when I should have said:

"the pareto optimal game structure under this model is morality"

Do you see why this correction is important?

The problem is that there exists an optimal game structure for everyone to follow using a certain person's preferences as a criterion. This would be the game that benefits that particular person the most. A friend described the sort of maximizing reasoning that would lead to wanting this configuration very well in an email a few years back:

Human culture and morals as they exist today all just randomly happen to stabilize civilization. If they didn't, civilization wouldn't exist. It's the same argument as "Why isn't the universe lifeless? Because if it was, nobody would be thinking that." The ideal situation for any given individual is to encourage the rest of humanity to follow these restrictive morals while personally rejecting their principles. Breaking the rules is only beneficial if everyone else follows them. Therefore, one must maintain the appearance of respecting sociological laws. The result? Psychopathy. Embrace it people. Or rather, follow your social traditions and let me manipulate you. 

This is very different from the moral game structure which is pareto optimal. The pareto optimality ensures that we account for everyone's preferences in a balanced way. It means the best game structure that everyone could agree upon. 

A Useful Distinction this Model Makes

"the rich should give more money to the poor"

Let me say first of all that this statement drives me crazy. Particularly when it comes from someone in the first world who isn't even making an effort to practice effective altruism. However, the main reason it drives me crazy is a great example of a place where this moral model might be useful. Should the rich really give more money to the poor? Or is this person just expressing that they dislike rich people and would be happier if we took rich people's money and gave it to the sympathetic poor people? We can isolate the moral part by asking the question "Would income redistribution be a pareto optimal policy (in the broadest sense) to enact" and thus separate it out from the mere preferences that haven't been processed with consideration of what everyone else cares about.

Problems and Further Considerations

One interesting problem that I think might be solved by recent work on decision theory and newcomb's problem is why anyone would behave morally when there is no enforcement mechanism. In particular I recommend Yudkowsky and Drescher if you are interested in this.

Another interesting problem is where you draw the line on who is a moral agent and who isn't. This is a VERY good candidate for Hardest Philosophical Problem Ever. Take a look at Non-Person Predicates to get an idea of how insanely difficult this can be. A more everyday scenario could be whether eating meat is ethical. 

There are always the widely understood but still fascinating problems of "how do we find the preferences?", "how consistent are human preferences really?", and "if they aren't how do we model things taking that into account?". My account of morality above is a very broad and high level one, but the details really matter a lot more. They are what actually do the work in constraining actions and systems. 

Finally, there is the problem of shared wants. If many of us value others to some degree, there is going to possibly be a lot of recursion involved in the preferences. I haven't seen any economic models where I can take increases of your utility function as an input into my utility function and I don't really know what the consequences of that might be. 

Finally finally, morality is definitely not a solved problem and this all could (and very likely is) totally wrong compared to the correct model that's out there somewhere. However, it does seem to be substantially better than anything else I have right now, but I'll keep looking because scholarship is important! There is a lot of philosophical grey area for me. For example where do we start the initial positions from for our pareto calculation? I do think that Rawl's approach takes this very seriously, but I haven't read far enough into moral philosophy to even know all the major positions, much less the minor ones and nuanced details. 

Also, I really recommend reading Nyan_Sandwich's recent sequence on LessWrong. Morality is Awesome is indeed quite awesome. Pinpointing Utility also talks about ethics quite extensively using economics as the launchpad. 

Darwin and Newton

In my Human Event class this semester we've now gone over the two most celebrated thinkers in the history of science, Newton and Darwin. There was also an ASU Origins debate on the definition of life, and my friend was an actor in the premier of a fascinating new play, Dreaming Darwin. In light of all this, I've definitely got some interesting ideas floating around in my head.

The contrast between Newton and Darwin is what I find most fascinating.

Both theories provided a comprehensible mechanism to explain some of the great mysteries of their time.

Both waited a great deal of time before they published. Newton did merely because he had it figured out, was a little bit crazy, and apparently didn't find it all that important. Furthermore, he was worried about protecting his invention of calculus. The ideas themselves had nothing to do with the delay. The legwork had already been done by Copernicus, Kepler, and Galileo. He was merely fleshing in and unifying the details of the system, instead of discovering a radical new view. If anything, Newton's work allowed people to go back into an older style of belief with comfort. The applecart of the heavens had been upset, but it turned out the apples all fell in nice, orderly, logical rows. Newton's theory is inherently appealing to the human mind.

Darwin on the other hand, delayed publishing because of the torture he was undergoing within his own mind. Yes, he saw that his theories would be controversial, but more importantly he was living the controversy. For though Darwin's system of natural selection was elegant and powerful, it was absolutely brutal. It created order, yes, but in a way that was so discordant it rails against the built-in preferences of the human mind. We do not want to see order out of chaos, we want to see chaos come out of order. We want to believe that the universe has a benevolent God, not a thousand blind and contradictory ones who operate in ways absolutely foreign and repulsive to our intellect. We want to believe that we are rational beings, that we are not mere precipitates of a great irrational process that optimized us for things we do not value. As if that's not bad enough, the same process even gave us the contradictory values and very repulsion we have to it.

Darwin's breakthrough forces us to question everything we know, everything we care about, and everything we believe at a level that no scientific theory ever really has. That's why it leaps and bounds above Newton's breakthrough as an idea. Newton's mechanics comes like the light of an angel from above guiding the way. Darwin's natural selection is a monster from the deep telling us we are meaningless in our sleep. Rebuilding the light out of that darkness is a monumental intellectual challenge, and so it is no small wonder that so many people today would rather attack than embrace it.

Towards a New Republic

I find it amazing, the way that people can synchronize and create incredible experiences like these together. My question is whether this is something that can be brought about not just in isolated incidents, but over the course of life as a whole.

A lot of it has to do with a common goal. The difference between a team and a group is that a team has a common goal. Because of this, the individual members are willing to depend upon and work with each other to achieve it. A group merely has members who are pursing their individual interests.

However, that last sentence should throw up a flag, because the members of a team are pursuing their individual interests as well. They just realize that the best thing for them is for the team as a whole to perform well.

It's a pretty simple distinction really, but the moment I really thought about that, I realized how simple so much of what I've been working towards really is. Nearly all meaningful portions of my life have been about getting the people around me, and ultimately all of humanity to work as a team instead of as a group.

My synthesis paper last year was my first huge attempt to figure out a coherent belief system that covered all aspects of my life. I started out with a very strong individualist foundation. I felt that all belief systems had to have a justification from self-interest, but I had to try and figure out what that meant for me. What really was in my self-interest?

I initially started out trying to determine what my actions and concerns would be to best achieve my existing interests. The more of these I wanted to achieve and the more thoroughly I wanted to achieve them, the more my actions and concerns extended outwards into the world. If I really wanted to get everything I wanted, I was going to have to change the entire world anyway. Even more importantly however, was a nagging thought that came into my head. Why do I want these things I want? I began to realize that many of them were completely arbitrary.

This was heavy, and I had to seriously reconsider. I could either set out for an objective basis for what my goals should be, or try to subjectively harmonize them so that I could at least live a happy life if they were all ultimately just arbitrary.

I figured that the only way there could be an objective end goal is if there is an ultimate point to the universe (because individual lives in a pointless universe necessarily don't have a point outside of themselves). The only conclusion I could come to on this line of thought is that if the universe has no objective point, then the only action that could possibly have a point at all is searching for one. Obviously this puts a big priority on avoiding the destruction of humanity, because it cannot continue the search if annihilated (existential risks). Likewise, it puts an emphasis on the search for new discoveries that might bring us closer to finding a purpose (science), and the constructive tasks needed to support, nourish, and inspire everyone as we pursue the endeavor (industry, culture).

On the front of subjective harmonization, I found that having an overarching purpose for humanity led to a much more satisfying life with much less contradictory goals. It allowed me to consider situations in a detached way, and bring about much more joy and happiness in my life. I know where I am going. I know that I have reasons for everything I do. I never become bored, and enjoy a nearly constant state of productive flow. I've found that many complex dilemmas people find themselves in, particularly social ones, are like Gordian Knots that can be cut through most expediently with an integrated value system.

Now, of course, those values of happiness and harmony are probably ultimately set by evolutionary trends, as there is no particular reason that a mind would come to value and be fulfilled by them better than any others. We know that a "paperclip optimizer" is not an invalid possibility, nor do we have any standard to determine that it would be wrong to optimize the universe for paperclips while optimizing for our values would be right. A perspective of objective value fails on that front.

However, I think it is possible that almost every human on the planet shares the core values that can justify adopting a certain kind of value system. Thus, even if there might be no goal that can integrate the value system of every possible sentient mind allowing them to successfully cooperate on a team, I think there exists a real possibility that all minds which could possibly occupy the humane portion of mindspace may have values for which this is the case.

The task that lies before us then, is that of finding a way to harmonize all humane value systems so that they can achieve the fullness of their potential. There are two main elements of this endeavor. The first is understanding how the cognitive evaluation systems of minds work, and particularly how they work in humane minds. The second is understanding how the interpersonal relations between minds can be structured so that their values can be achieved. In short, the tasks of the cognitive-behavorial and the socio-economic sciences.

This is not a new goal, it's been around for a long time. Plato's Republic is the most notable attempt that comes to mind. He was trying to figure out the complex and often contradictory nature of human value systems so that he could create an optimal society. While many of his methodologies may be flawed, his science dismal, and the results unworkable, this is cause for hope rather than despair. We can use the findings of cognitive science and economics, both of which didn’t even exist in the time of Plato. Likewise, we can postulate the achievement of goals using technological means that Plato could never even dream of.

Certainly there is an element of hubris here. Utopia means "no place" for good reason. However, we have learned from past attempts to not take rigid structures of society as the ideal. We have learned that we do not know everything, and cannot ever know everything. However, while better understanding our limitations, we can simultaneously identify with much more precision and scope the basic laws and principles that underlie cognition and social interaction. We know the powers of deontological rules that allow for the development, evolution, and emergence of complex systems. We also have new ways of communicating, entirely new sciences, and breakthroughs across the board.

There have certainly been people since Plato who gave their best shot at the project, but I have yet to see anyone who looked at it in quite this perspective. I believe it not only holds the potential for outstandingly fruitful results, but also deep edification for those who pursue it.

Life as an Optimization Problem

I've always been very concerned with the concept of what constitutes a good life. Having good goals is difficult enough, but actually achieving them is at least several levels of difficulty above that. The way I see it, there are two primary kinds of optimization that help to achieve goals. One is highly personal, building up the patterns of thought, behavior and action that lead to a greater capability for growth. However, this is not enough. We are both limited and empowered by our environment, by the people and situations around us. Even the brightest person cannot grow in a vacuum, and the good a single person can do unaided is extremely limited. Thus becoming extraordinarily adept at both aspects of optimization is vital, and being acutely aware of the ways they play off of each other is a huge part of that.

The two best resources I can think of for individual optimization are the posts and sequences on less wrong and economics, particularly Austrian Economics. The tools of rationality and meta-cognition on less wrong can work wonders for anyone if well understood and practiced. Seriously, it's brilliant stuff. Economics studies the most efficient use of scarce resources and thus is obviously going to be useful for optimization. The Austrian School is a goldmine for individual optimization, because the focus is entirely on how individuals make decisions to best achieve their goals. Mises called the science of human action "praxeology". Combining an understanding of praxeology with the knowledge of epistemology, rationality, and biases on less wrong gives you the analytical capabilities to really make some progress on optimizing your life and attaining your goals.

Optimizing everything around you is in many ways more challenging than merely optimizing yourself. Some things simply can't be changed, or would require more effort than it's worth to be changed. Recognize these things and let them be. Wu-wei is your friend. When you are building yourself, being smart about things matters, but raw focus and determination will probably bear the brunt of the work. However, when dealing with the outside world, particularly social situations, it is necessary to be more clever. Communication theory and practice is always helpful. You can also study cases of success in whatever area will help with achieving your goals. For example, I read biographies of great economists that focus on the factors that contributed to the development of their ideas. I also read things on the sociology of intellectuals.

Reflection is absolutely crucial the entire time you are on this endeavor. If you aren't lucidly aware of your own mental state and abilities, don't know what is happening in the world around you that affects you, and don't carefully consider the options available to you, failure is pretty much guaranteed. In fact, despite any pleasant reassurances, failure is the norm even if you do all those things. Think about it, how many people really, truly, absolutely achieve their goals? Quite possibly none. How many come close? Very few. True success is not only hard, it's perhaps the most difficult thing in the universe.

If at this point you are ready to give up your goals, do so. Because if you are ready to give them up they didn't really mean anything to you anyway. If they really matter, it doesn't matter how hard they are to achieve. It doesn't matter if the probability of success is only a shade above zero. There are goals like that. Ones that matter that much, and if you don't have one yet I suggest you start searching.

Subgoals are fluid. They change based on probabilities. If your goal is to help everyone, you might think the best plan is to go to Africa. It might turn out that going to Russia actually has a greater probability of helping people so you recalibrate your subgoal to go there. On further reflection, it might be more effective to become a CPA, work a standard desk job, and donate all your spare income to good charities. The point is that regardless of the subgoals that most effectively optimize for your main goal, your main goal is not going to change. Even if people tell you that there's a 0.00001% chance of helping everyone, you're still going to try. Otherwise that goal is not your true main goal, just a really important subgoal.

So for people who have goals that truly matter, the fact that they are almost impossible to achieve only means that more effort should be put into doing so. Every applicable resource, every ounce of intelligence and effort, every single conscious moment will be put towards increasing that sliver of probability, and the smaller it is the more herculean the effort will be. They will optimize like their life depends upon it. This post should be beneficial to anyone, but it is to those who really have something that matters to them that I hope it proves most useful.

A Question on Property Rights

This was a question I wrote to kick off a recent Society of Economic Thought meeting:

In many ways the very reality of an economic system depends upon how ownership is structured. It is the existence of property rights and the way they function in a society that determine who gets to make decisions and how those decisions are made, and since economics is fundamentally about decisions, this of course affects how the entire economy will work.

I think everyone in this room is acutely aware of the fact that there are some huge differences in how we all think economies should be structured, and I think that most of these differences ultimately boil down to questions about how ownership should work. There are an incredible amount of viewpoints on the matter. There are people who think that everything should be owned by the state because it's efficient, like in socialism or fascism. Communists think that property should be abolished completely in its current form, and everyone should be able to use everything. Some people think that workers should own the capital they use to produce goods, which is called syndicalism. Noam Chomsky is a famous modern advocate of this view.

Many people think that individuals should privately own property and that's basically your conventional capitalist viewpoint, but even within that viewpoint there are many shades of gray as to how firm individual property rights should be, as well as a multitude of ways that that private ownership of property can be done. You can have it as a trust from the state to individuals because that's the most productive way to generate resources for the state, which is effectively the mercantilist viewpoint. Or, you could make it so that one king effectively owns everything but leases it out for the same reason a mercantilist state would... because it increases his wealth and power. Sometimes people hold that individuals are the fundamental owners of their property, but that there can be exceptions to that if the situation requires them. In other words, many of us believe that ownership isn't absolute. And even if we do hold that individual property rights are absolute, which system do we use to justify that and determine who owns what?

Does each person own their property by mixing their labor with it, like Locke argued? Is property ultimately just about who has the biggest guns? Or like my uncle says, are we kidding ourselves because possession is 9/10 of the law? Is all property just theft? How far back do property rights go? Should we be concerned with say, giving the property we stole from the Native Americans back to them, or is there a certain point where we just accept the current distribution of property because it’s expedient to do so? Are property rights ultimately subject to a sort of practical truncation after restoring them becomes costly and difficult? I mean, these are only a few ways to think about and justify individual property rights, but I think you get the idea… it’s complicated. What even qualifies as property has changed over time. For example, people used to be able to own other people. You can’t do that anymore. And whether people should be able to own ideas is still an extremely contentious issue today.

So what I want you to really think about is.... how historically has property been structured, and why did past societies structure it that way? How is property structured in today’s world, and why is it structured this way? And most importantly, how should property be structured, and why should we structure it that way?

Harmony

Harmony is something that never ceases to amaze me, because it involves the synthesis of almost everything in a very deep and fundamental way, including the different motives of people. It is really remarkable how much we can get to fit together and align, and the deeper the analysis is, the easier it usually is to fit things together. That deep feeling seems to stem from structure, because of the power good structure lends to concepts. Maybe that is the core of sound synthesis, there has to be some kind of structure there for it to happen. Otherwise things can't fit together. At the risk of sounding crazy, it's sort of like jello. You can't really fit together two pieces of jello --it's just awkward and never quite happens right. Yet you can fit together puzzle pieces with clearly defined edges in a much more satisfying way. However, sometimes you just have to fit the pieces of jello together in a fuzzy way, even if that doesn't give the satisfaction of a deep conceptual fit. Because, ultimately, to fuse the jello correctly, you would have to go to the underlying structure of the atoms, and that is too complex for a human mind to comprehend the particulars of, so you have to just be happy with the sketchy union between the two ideas. This is the problem, that harmony requires knowing so many things at such a deep level that it is impossible to properly do it with many things in the human mind. Specialization is easy, and can allow us to bring greater levels of harmony to little parts of the universe, but harmony over broad swaths of subject area, which is perhaps the most valuable to have, can only be achieved pretty superficially, even by the most powerful of minds. People who can develop and present a deep structure of some element of reality in a very lucid way are thus serving an incredibly valuable service to those who seek harmony. They give a way for us to stand on the shoulders of giants and synthesize ever greater parts of our life into powerful structures that allow us to act much more effectively with the world around us and with each other. When two things seem irreconcilable, go deeper into their structure because that's where the solution will lie.