Sunday, July 27, 2014

The War on Terroir

FINALLY, people are starting to avail themselves of the pun opportunities given the similarity of terror and terroir.

I for one have long suspected (and our foreign policy similarly ignored) the possibility that a team of suicide vintners from Bordeaux would infiltrate Napa. But for those of us with refined palates, this worry is but one of many crosses to bear.

Monday, June 30, 2014

The Endowment Effect as a Rational Strategy

The endowment effect is well-studied: people assign higher worths to their own possessions than to same things owned by someone else. That is to say: you wouldn't pay more than $30 for that couch on Craigslist, but when you list your own - same model, same condition - for some reason, you ask (and somehow, actually expect!) $75.

To many of us there are two central features of cognitives biases that make them so interesting. The first is that they're cognitive biases, not cognitive stupidities. We aren't just all over the map due to limited brainpower, as bounded cognition models would suggest; we consistently make mistakes in the same "direction". This consistency (even if it's consistent incorrectness) leads to the other interesting aspect of cognitive biases, which is the extent to which they might actually be instrumentally rational shortcuts in disguise, either being seen out of context or profitable only over the long run. That is, perhaps when we were hunter-gatherers, they were useful. For example, hyperactive pattern detection (type I error) is a terrible thing in the modern age when we're looking for incoming nuclear warheads, but during the Pleistocene, we were stupid and it couldn't get us into too much trouble. Sure, maybe you might end up thinking the gods struck the mountain with lightning because your children cursed a lot that morning, and a thousand other strange things, but we couldn't really do too much about our strange beliefs anyway - but if that one time you saw a shape in the grass that looked like a hyena, and it really was a hyena - well, you still came out ahead.

There's a lot in behavior and medicine like this that only makes sense in evolutionary context, for instance fever. Fever is not something that pathogens do to us, it's something that our bodies actively choose to do to them. In the modern age it's really hard for us to understand how fever can be beneficial, until we remember that merely two centuries ago when, without medicine, we could (and often did) die of fevers. But this is a kind of base-rate fallacy. Fever is a way our body shakes off pathogens, and before modern medicine, that cut in your foot might have a 70% chance of going septic and killing you. If the fever only has a 65% chance of killing you, you still come out ahead of any competitor who didn't have a fever response.

So how could a completely judgment-clouding bias like endowment effect ever be helpful? Finding a group of people who don't appear to have it might point us in the right direction. And indeed, Apicella et al found that Hadza hunter-gatherers in contact with markets show the endowment effect just like the rest of us, but Hadza way out in the bush away from markets do not.

This is pretty amazing - there's a difference in a known bias, even within the same group of people. So we've already learned one thing: the endowment effect is a learned behavior. So either we post-agricultural types are stupid, or we're getting something out of this bias. What's the benefit? And what exactly is different about the two groups of people that makes one group adopt this strategy?

A simple model of markets assumes information symmetry between participants. But in reality, specialization of labor means that the person approaching you to buy your possessions will almost always have better information than the aproachee who's selling because the buyer buys (cars, computers, art, etc.) all the time - and if they're initiating the transaction, they are even MORE likely to have better information. Therefore, the endowment effect may be a learned behavior whereby we value our own possessions more highly than is justified on the open market as a defense against information asymmetry.

Here's a concrete example. Imagine you're selling a car. You seek out an offer, and you find another individual buyer. Are you more comfortable that you're getting a fair deal from them than the car dealer? Of course you are. Now imagine you're approached out of the blue by someone who buys cars for a living. Sure, you'd consider it, but only at a very high price where you're sure you're not getting swindled. (This isn't to suggest that the endowment effect occurs consciously, but you can see how when we calculate consciously, we might behave in exactly the way the endowment effect would influence us.)

Consequently, you can imagine the endowment effect as the cushion you need to come out even, when someone with better information than you buys your possessions, especially one who's initiating the deal. And returning to the Hadza hunter-gatherers, we may have identified the relevant difference. Hunter-gatherers are less specialized. If you're a hunter-gatherer, you know just about everybody you could possibly buy or trade something from, and there's not much special knowledge to allow information asymmetry - i.e., everyone is equally smart about the relative value of gazelles and axe heads - so there's no cause to develop such a bias.

This theory makes several testable claims.

1) The more information asymmetry that exists between possession-owner/seller and the buyer, the more the endowment effect gap should be. You think your car is worth more when the car salesman runs up to you on the street initiating and offer to buy it, than when you're selling to someone online with apparently equal knowledge about cars.

2) The endowment effect gap should be larger in markets where there is poorer information, there is less trust, and for goods that are more difficult to value. Complex goods are more difficult to value.

Not central to the theory, but relevant to the fact that this appears to be a learned behavior: children at some arbitrarily young age should not show the endowment effect, because if this differs between humans, it's learned behavior; kids haven't been swindled out of their possessions enough times. (And what is this arbitrarily young age?) Non-hunter-gatherers who nonetheless live in small groups of people with little interchange with other populations should lack this bias. Hunter-gatherers should develop this bias after they come into contact with markets.

There is also a kind of transaction cost argument, separate from the theory. Yes, you might have just gotten a good deal for your used Toyota, but now you have a Nissan and there's a utility cost to you of learning how to operate a new car. If that's ALL the endowment effect is, then each individual's ability to learn new behaviors should completely predict the entire strength of the endowment effect in each individual, at least with complex things like cars.

Saturday, June 21, 2014

Passivity and Usefulness of Information Do Not Positively Correlate

Previously I posted that we shouldn't be surprised that media information that we receive passively isn't necessarily useful to us. It costs money for all the information to get transmitted so broadly, and the reason it happens is because the people on the other end think it is benefiting them, usually financially. This of course doesn't correlate positively with that information being useful to the recipient, and might even correlate negatively. If you put in effort to find information you used to make a decision, it is more likely to be useful to you. Robin Hanson approaches this same idea in his post Why Info Push Dominates.

Monday, May 12, 2014

Anarchist Conference Devolves Into Chaos (Not from the Onion)

Story, such as it is, found here. I wonder if anyone asked "I thought we were an autonomous collective?"

Friday, March 21, 2014

Crime and Punishment: If We Could Create Hell, Should We?

This is also posted at my religion and morality blog The Lucky Atheist.

Sam Harris has in several places discussed his moral thought experiment of a pill that would make prisoners think they'd been tortured when they really hadn't. This is why a recent piece in Aeon about length of punishment was so interesting. First it asks if it's ethical to extend someone's life so they could serve a thousand year sentence; then it asks if it would be ethical to change their consciousness so they thought they had served a thousand years, or least much longer than they actually had. (Star Trek fans will recognize an episode of Deep Space 9 here, where exactly this happened.)

More broadly, the moral issue they're getting at here is whether some kind of hell is permissible or even to be encouraged, once we can create it. (Again science fiction has been there: Iain Banks readers will recognize the central question of Surface Detail, where virtual reality hells are used to punish brutal dictators.) It would also be interesting to ask theists about this. Assuming a situation where everyone (theist and atheist) agrees that a particular person's actions are suitably heinous, shouldn't we begin their punishment as soon as possible? If Hell is a real place that bad people go to, why wait?

Leaving aside the more mundane questions that we can ask of our prison right now - who wants to feed a bad guy for a thousand years, and does knowledge of a thousand year sentence actually deter behavior (utility calculations don't seem to affect bank robbery rates, for example) - I think it's obvious that we should focus on how to prevent terrible behavior in the first place, instead of how to make punishment after the fact more severe. Instead of creating virtual hells, shouldn't we focus on creating a real heaven? If you have the science to make someone think they're living a thousand years, what about the science to keep them from becoming a serial killer in the first place? Here people might object that if we choose that path, surely there's something scary about tampering with human nature and free will, to which I answer: you better have a very, very clear argument then. If we weigh the very concrete suffering experienced by Ted Bundy's victims and their families against this amorphous threat to human nature, I think engineering better behavior wins hands down.

In a way, up until now we've been fortunate, because this hasn't been a choice that it's within our power to make. But once it's within our grasp, if we don't face the facts head on and make an explicit decision based on our values - if we muddle through pretending we still don't have that power, or we just let inertia be our guide - then we're being profoundly immoral. Maybe not this particular decision, but many decisions like it, are going to have to be made already in this century.

Wednesday, March 19, 2014

Wealth and Coastline, Part 2: The United States

Previously I showed that the average income in the set of 138 countries for which I had income data, countries that do have some saltwater coastline is $13,562, vs $9,040 for those with no coastline. This is not surprising, to the extent that you'd expect places with globally linked economies established by their historical presence on the seas to do well, as you would expect for places that were non-extractively colonized (they'll often be the same places, i.e. the U.S.) So it might be worth asking, is there a similar effect in the U.S.?

Yes. In general, the coastal states have higher per capita incomes than the landlocked states. But the scatter plot is interesting. There's a U-shaped trend to this data, of distance of each state capital from American saltwater, versus per capita income (U.S. Census). I say "American saltwater" because there are two states (New Mexico and Arizona) where it's faster to go to the Gulf of California. Even including that as the closest saltwater doesn't change much.



You notice the income minimum at 607 miles from saltwater? It's not just a curve artifact - if you look at the states with capitals 500-700 miles from saltwater, they have a lower per capita among them ($24,141) vs the country as a whole ($28,051). These 7 states with capitals 500-700 miles are the sparsely populated and therefore small service economies of the northern Rockies (Utah, Montana, Idaho), as well as part of the Bad Stripe in the Appalachians (Tennessee, Kentucky, Indiana), and post-industrial Michigan. Is there something worse about being 600 miles from the sea, as opposed 1000 or more? (6 states: Iowa, Colorado, Wyoming, Minnesota, South Dakota, and North Dakota, the latter the furthest at 1,235 miles.) But it's possible that it's just a historical accident, as with the supposed effect of rainfall at 100 degrees west longitude causing the population to drop off toward the west. In the case of 100 degrees west, Americans had settled to about that distance inland when the rail lines were completed. (Read the history of towns in western Texas and you will usually see a discussion of which rail line it was on; true for most of the Frontier Strip.)

Also of note: I used driving distance to saltwater, but this is not a great approximation. Even if we were going to use driving distance to saltwater regardless of country, you couldn't include Hudson Bay as North Dakota's closest, because you can't drive there! And just because you can drive to coastline doesn't mean there's a port there, or geography that allows ports to be built. Furthermore, the Great Lakes states are de facto on saltwater because canals and locks connect shipping all the way through to Minnesota.

If you count states in the discrete category of whether they're landlocked and how landlocked they are (1 state away from saltwater? 2 states?) it's clear that the median income of saltwater states is higher - $29,551 vs $26,073. However, among the landlocked states, the U-shaped curve appears again: for the land-locked states, there's no trend for more isolated = lower income; if anything it's the opposite (isolated means the more states you have to go through to get to saltwater from there). The only state 4 states away from Saltwater is Minnesota which at 30,656 has a higher PCI than the median of the coastal states., the better its income. (1 state away, $25,275; 2 states away, $26,695; 3 states away, $26,545.)

Wednesday, February 26, 2014

What Do the Patterns in Alternate Histories Say About Us?

For the most recent alternate history (a Buddhist Colony in Ptolemy's Alexandria) go here. This post is also cross-posted to my speculative fiction blog.

Alternative histories have tended so far to be about events in European (or Western) history, because they're written mostly by European-descended people, and mostly by English-speakers at that.

But that's kind of obvious. So even forgetting the (so far) Western focus of what-if fiction, there are two clear patterns that betray some of our assumptions. And the first obvious pattern is: alternate histories are about violence transpiring differently. How many alternate histories involve policy decisions or inventions happening out of order? It's not usually: what if Newton and Leibniz had not developed calculus, or antibiotics had never been invented (or been invented in pre-Mongol-sack Baghdad)? No, the large majority of what we see are the effects of different outcomes of battles and wars. What if Hitler won, what if the North lost Gettysburg, what if Alexander or Jenghiz had not turned around at India and Austria. And this is depressing; because it means either that history is mostly determined by violence, or (even if it's not true) it means that we at least believe that history is mostly determined by violence.

Yes, there are a few stories about political decisions (what if the Ming had not called back the treasure fleets is a favorite) but it's really mostly about if people had killed and dominated each other in a different way.

You could also argue that what we see commercially is not an accurate reflection of our beliefs about history. It's the same answer for why in speculative fiction, dystopias far outnumber utopias. There's clear conflict and thus they're easier to write.

The test at Sitio Trinidad, 1845, Jornada de Muerte, Nueva Mexico. The test was witnessed by governor Manuel Armijo and president Santa Ana, both of whom lost their vision as a result. (It seems both fission and high speed cameras were developed in nineteenth century Mexico but not dark glasses, go figure.) Despite this setback, the dreaded bomba santabarbara was smuggled in wagons and assembled in Austin and San Antonio and led to the end of the decade-long revuelta tejana and ultimately, the conquest of Alto México (previously the "United States".) Sure it did.

The second pattern, or really observation that students of history can make about this sub-genre: even with battles won or lost differently, it's very hard to find believable changes that affect the outcomes. Sun Tzu was right, the battle usually is won or lost before it begins, and even if the tide on one battlefield had turned, the currents were running one direction or another. For example, so what if Lee won at Gettysburg? A big setback for the Union to be sure, but the Confederacy was screwed from the start in terms of their population and economy. To really get big changes, you have to make major shifts long before the obvious change - the kinds of changes that would have given the South a fighting chance would have had to begin many years before the actual war. Case in point: someone once asked me what might have happened if, in the Mexican War, the Mexican factions had unified against the U.S. as an external threat, and the U.S. lost its first major foreign war and gained no territory. (Or if the U.S. had decided to actually press its 54'40" claim.) A suddenly unified and organized 1840s Mexico is (unfortunately for patriotic Mexicans) only marginally easier to imagine than the first atomic bomb being engineered a hundred years early at Los Alamos by Mexican scientists. Such a story might actually make for some really interesting Latino steampunk fiction, but might as well also include unicorns.