2017-11-20 RRG Notes
- When should you believe that you can do something unusually well?
- How do you distinguish between a bad idea and an insufficiently explored good idea?
- Modest Epistemology
- Example: Eliezer noticed that the Bank of Japan's monetary policy wasn't creating enough money
- Why should we believe Eliezer?
- Even if Eliezer cites experts who are on his side, why should be believe those experts rather than the experts on the other side?
- Why shouldn't we go with the wisdom of the crowds?
- The wisdom of the crowds (or the efficient market hypothesis) only applies in a very limited set of circumstances
- Payouts conditional on actual performance
- Lots of data
- Fast feedback loop
- The process of exploiting an inefficiency corrects that inefficiency
- If these preconditions aren't met, then you can have more confidence in your ability to beat the market or conventional wisdom
- If there's no way to make money from something being broken, that thing will remain broken
- Going back to the Bank of Japan example - even though many people knew that the Bank of Japan's monetary policy was suboptimal, there was no way for them to make money from the fact
- There was also no way for them to take over the monetary policy and reap rewards from the increased economic growth
- Thus, Eliezer is more justified in criticizing the Bank of Japan's monetary policy, since the process by which that policy is decided does not have the preconditions of an efficient market
- We encounter situations in which the wisdom of the crowds doesn't apply in our everyday experience
- Eliezer found that a good treatment for Seasonal Affective Disorder was simply more light
- But why should he have believed that more light ought to work?
- There were no studies showing that more light would have worked
- He noticed that vacationing in Chile was an effective treatment
- He noticed that traditional lightboxes did not work
- He reasoned that the lightbox wasn't producing enough light - and that putting up a bunch of 60w LED bulbs with the appropriate color temperature would work
- So given that this was a fairly simple remedy, why weren't there studies that examined using more light as a treatment for Seasonal Affective Disorder?
- Humanity is not being maximally opportunistic about studying Seasonal Affective Disorder in the same way that it's opportunistic about taking advantage of arbitrage opportunities in the stock market
- If you want to outperform society, you should look for places where society isn't being maximally opportunistic and focus your efforts on those areas
- We need to distinguish between three concepts:
- Efficiency
- Inexploitability
- Adequacy
- Efficiency
- Efficiency merely means that you can't reliably predict the movement of a price in a market without additional information
- Doesn't mean the price is just
- Price fluctuations don't mean a market is inefficient - fluctuations are the mechanism by which the market integrates and communicates new data
- The problem with taking this view of the efficient market hypothesis is that you're assuming that there aren't systematic biases among market participants
- Moreover, efficient markets assume that everybody has unlimited capital, margin calls aren't a thing, etc.
- In practice, irrational noise traders can earn higher returns than "rational" investors, in a situation where all parties have limited capital and the noise traders outnumber the rational investors
- In short, "the market can stay irrational longer than you can stay solvent"
- In a liquid market, if you can predict that prices will move in a particular direction, you can make money off that fact and your making money helps correct the distortion
- See above - even in the most liquid market on earth, the American stock market, systematic mispricing can persist, due to the noise trader effect from above
- The efficiency of a market is relative to your own intelligence - a market may not be efficient from the perspective a superintelligent algorithm, but it is efficient from your perspective
- The efficiency of a market is relative to your own intelligence and capital - if you don't have any money, you can notice market inefficiencies 'til the cows come home, but you can't do anything about them
- Inexploitability
- Not all things that involve money are efficient
- The housing market has a lot more money in it than the stock market, but it's not as efficient
- It's not easy to make money from housing prices going down - no analog to shorting stock
- Therefore, even as a non-specialist, it's possible that you might be able to guess at which way the price of a house is going to move, which can be helpful if you're buying a house
- Therefore the housing market is inefficient - you can predict which way prices are going to move - but inexploitable - you can't make money from that movement
- It's the inexploitability that allows price distortions to persist
- Adequacy
- Adequacy is whether the low hanging fruit have been plucked in a given domain
- Example: seasonal affective disorder
- Eliezer found that hanging a bunch of LED lightbulbs up was an effective treatment for seasonal affective disorder
- So why aren't there companies selling wall-size LED arrays as treatments for seasonal affective disorder
- You can't just throw some 2x4s and a bunch of LED lightbulbs in a box and sell that as a product
- There's a huge jump from building something as a one-off project for yourself to scaling that up and selling it as a repeatable, cost-effective solution that's usable and installable by ordinary people in a wide variety of situations
- So how does inadequacy arise?
- Inadequacy arises when the incentives in a system aren't what you think they are
- Example: academic research
- Incentives for researchers are to publish in high-impact publications that get lots of citations
- Incentives for grantmakers are to fund research that gets a lot of prestige
- Neither side is directly interested in improving knowledge - that's something that happens as a side effect of the process
- It's not sufficient for one part of a system to be broken - there have to be two parts, so that a single individual can't fix the system without a massive amount of resources
- In the same way that inefficient markets are inexploitable, inadequate systems are unfixable
- Inadequate systems and markets both share the property of being in competitive equilibrium
- People are acting according to the incentives of the system, not according to what you think their incentives are
- It's the lack of "free energy" in the system that prevents the system from changing
- If people in the system had the resources to change their incentives, the system would no longer be inadequate
- But because the system is in competitive equilibrium, any attempt to change the incentives of the system puts the person attempting to make the change at a severe competitive disadvantage
- In fact the concept of inadequacy highlights an inadequacy
- If economics were adequate to applying the concept of free markets to other fields it would have come up with the notion of adequacy
- But it hasn't, so therefore economics is inadequate for discussing inadequacy
- Except that it isn't. Inadequacy is a really fancy word for "market failure". Or maybe a Nash Equilibrium.
- When Eliezer found that stringing up lots of LED bulbs was an effective treatment for seasonal affective disorder, he didn't go around trying to get researchers interested in it
- It's an obvious enough idea that if it were easy to research some researcher would have already studied it
- Therefore, the fact that it has not already been studied means that the system for researching treatments for seasonal affective disorder is inadequate
- And because it's inadequate, it's impossible to fix without enough resources to single-handedly change the incentive structure
- The interesting form of inadequacy is when everyone seems to know that the system is inadequate, but no one seems to have enough power to change the system
- It's very easy to come up with an a priori inadequacy argument about anything
- Before you conclude that the system is inadequate for solving the problem you're thinking about, do some research and see if the problem has been studied
- An inadequate system is one that doesn't even look at the problem, not one that hasn't solved the problem yet, but is actively working on it
- 3 categories of concepts for analyzing "adequacy"
- Decisionmakers who are not beneficiaries (i.e. principal-agent problems)
- Asymmetric information
- Systems being stuck in suboptimal Nash Equilibria
- The real problem is when you have multiple "inadequate" systems reinforcing each others' inadequacies
- Example: Parenteral nutrition for premature babies
- Premature babies born with short-bowel syndrome need intravenous feeding while their digestive systems finish developing
- It's best if the formula that these babies are given contains Omega-3 fatty acids
- However, the FDA approved formula contains mainly Omega-6 fatty acids, which is possibly harmful
- So why haven't we updated the formula to contain Omega-3?
- We need a large scale study to prove that Omega-3 is more beneficial than Omega-6
- The problem is that this large scale study would be a replication, and the current scientific incentives aren't aligned to make replication a priority
- Prestigious journals demand conventional scientific and statistical methods, so scientists design their studies so that they're publishable
- All of this is a result of inferior Nash equilibria
- Not all Nash equilibria are identical in terms of utility
- It's possible for some to be strictly worse than others
- Many of our civilizational problems can be explained by the fact that we're stuck in inferior Nash equilibria
- There's no global administrator choosing the Nash equilibrium that a system ends up in
- Total Market Failures
- Example: a doctor today does what ought to be three or four separate jobs handled by three or four separate people
- Eliezer has never heard of transaction costs, has he?
- The reason the medical profession hasn't split apart the role of "doctor" is because of regulatory capture
- Regulators have been captured by interest groups like the AMA, who have an interest in keeping the medical profession high status
- Publishing statistics would lead to a first mover disadvantage, so no hospitals publish statistics, nor prices, as the first hospital to do so would be hurt
- So why don't the people who know better go off and set up their own political entity and see if they can do better?
- Existing governments have a monopoly on land
- This monopoly is defended with force
- The fact that a system is in equilibrium doesn't mean that the equilibrium is good
- These equilibria may persist even though everyone wants changes, the process in question is a multi-stage one and success at the current stage of the process is dependent on acceptance by the next stage
- Something that's effective but unconventional at the first stage will never come to pass if it's rejected by the second stage in a multi-stage process
- So why don't people change their political systems to allow for more experimentation?
- Political systems are defined by the same inadequate Nash equilibria as everything else
- First-past-the-post-voting is a result of the same kind of inadequate equilibrium as everything else
- Eliezer is missing two big things: complexity and legitimacy
- Condorcet voting may have better outcomes, but those outcomes will be harder to explain, and, as a result, will be less legitimate
- Heck, the Electoral College is bad enough in this regard, even though it serves an arguably important check on the power of large, urbanized states like New York and California
- Another example of inadequate equilibria is Overton Windows
- Everyone is ready to talk about a particular change
- But because no one has started talking about it, everyone feels like it's taboo
- Then someone unusually brave starts talking about it and it opens up the field to it being a legitimate topic
