Thursday, November 24, 2016

Avoiding the "Smug Liberal" Trap

Background: This is the fifth piece in a series on understanding conservative intuitions and communicating effectively with conservatives, for liberals. Please read the previous post BEFORE reading this one. The SSC zombies post is also strongly recommended. Earlier posts are optional.

Trigger warning: brief discussion of racism and college rape discussions.

Last post talked about signalling moral superiority. We used the example of "LGBT" vs "LGBTQ" vs "LGBTQIA" vs "LGBTQIAPK" vs "LGBTQQIP2SAA". This post talks about how that sort of thing can spoil conversations with conservatives.

For now, let's forget about conservatives for a moment. Let's say you're a badly stereotyped liberal talking to a fellow badly stereotyped liberal, and you bring up the topic of LGBT rights - maybe something like "hey, there's an LGBT rights march in the city next weekend, are you interested?". Your fellow liberal says something along the lines of "Yeah definitely, I should really be more involved in the LGBTQIAPK community.".

At best, that's going to raise an eyebrow. At worst, a response like that comes across as someone trying to show how morally superior they are, by memorizing all those extra letters, despite apparently not being all that involved in the relevant community. Depending on context, it could seem extremely pretentious (and often disingenuous to boot).

Obviously this is an exaggerated example. But when engaging with conservatives, it's important to pay attention to this sort of thing. Even conservatives who support LGBT rights are going to find it slightly pretentious if you talk to them about LGBTQ rights - not because they don't support queers, but because they feel like you're just trying to one-up them.

If you want to actually get through to someone, teach them something or change their mind, then you need to build some rapport. Making them feel like you're one-upping them is basically the opposite of building rapport. If you want to get through, don't try to seem morally upstanding; instead, play up your own moral weaknesses. On the other hand, remember that conservatives usually don't try to signal morality to nearly the extent that liberals do. So if a conservative seems to be signalling against something, remember that they might just not be signalling at all.

Some examples:
1. Racism is a great area to apply this, because practically everyone is at least a little bit racist (source: implicit association tests). Maybe say something like "look, I get a bit nervous when I'm walking alone at night and I pass a big black guy. That's a pretty ingrained instinct in our culture. But that doesn't mean that instinct is right, and we certainly don't want cops to go shooting people just because of that instinct...". You might try saying something like this even if you don't actually feel nervous.

2. Ambiguous cases make good signalling opportunities. College rape is a good example of this. There are (lots of) obvious clean-cut cases, like date rape drugs. But arguing that we need to crack down on date rape drugs is not very useful from a signalling perspective - practically everyone will agree with that, so arguing for it doesn't say much about you. On the other hand, there are people who argue that we should believe virtually every accusation, or that it should count as "rape" if a woman decides in hindsight that sex was a bad idea. These are more questionable. And because these ideas are questionable, arguing in favor of them sends a very strong signal - no one would argue in favor of ideas that extreme unless they were really concerned about college rape.

If you want to communicate effectively with conservatives, then try to notice this pattern. Even if you really honestly think that we should believe virtually all rape allegations, you should still recognize that reasonable people can find this pretty questionable, and it will require pretty strong evidence to make a case (you can find this evidence if you look - but most people haven't). Just because people don't agree with you outright does not mean they're bad people who hate women, it might just mean that they're not signalling as hard. And you still need to build rapport - open with something like "I know it sounds surprising, I thought it sounded ridiculous at first, but I looked at the data and...". Again, try saying this even if you didn't question it at first.

2b. Not necessarily a signalling thing, but as a general rule of political discussion etiquette, try to avoid the "mott and bailey" trick. I've seen this come up e.g. when discussing feminism - going from "feminism just means supporting equal treatment for women and men" to "opposing abortion means opposing equal treatment for women and men". The former is easy to defend. The latter is harder to defend. They are not logically equivalent.

In practice, mott-and-bailey arguments frequently accompany ambiguous-case signalling. If someone calls you out on using a mott-and-bailey, just admit that your argument was wrong, but make it clear that just because the argument was wrong doesn't make the conclusion wrong. Contrary to instinct, people will be far more likely to believe you later if you admit you were wrong at some point, so don't be afraid to really ham up an apology.

3. Remember that economics is a thing. There's the classic cautionary tale of rent control: New York City originally adopted rent control to protect lower-income people from exorbitant rent. After that, developers stopped building low-cost apartments, because rent control meant that prices couldn't adjust to demand, and tenants would hold rent-controlled apartments for decades. To this very day, there is a huge shortage of low-cost apartments in NYC. But there is an abundance of "luxury" apartments - apartments past a certain price point are not subject to rent control. In short, rent control accelerated gentrification.

Why do I bring this up in a signalling discussion? When someone argues against rent control or minimum wage or the ACA or what have you, there's a temptation to interpret it as signalling. It feels like they're against poor people or against sick people. Don't blindly trust that feeling; they may just be economically literate.

4. If you think someone is working against their own self-interest, that's a major red flag. I occasionally hear liberals talk about how lots of conservatives vote against their own interest because they're conned by Rupert Murdoch. That sort of thing should be considered unlikely a priori, so you better have a strong case for why people think they're following their own self interest. "Because fox news told them so" is not, by itself, a strong case. The rent control example is a strong case - something that has bad consequences long term seems good at first glance. If you want to argue that people are voting against their interest, then you should also figure out why it looks good at first glance.

Assuming that you know what is good for a person better than that person may be the fastest way to get labelled a "smug liberal". Of course, there will be cases where you do know what's better, but that does not mean it's a good idea to say so! The relevance of signalling is left as an exercise to the reader.

Closing reminder: please remember that the point here is how to talk to conservatives, not who's right. Just because I suggest avoiding an argument, does not mean that the thing the argument supports is wrong.

Wednesday, November 16, 2016

The Inevitable Post about Signalling

Background: This is the fourth piece in a series on understanding conservative intuitions, for liberals; here are the first, second, and third. Second is optional, and just discusses current events.

The zombies piece hypothesized that liberal instincts are evolutionarily adapted to a world of abundance. It also mentions "what would no doubt be the main pastime of the people of this world: signaling". In order to talk more about how conservatives see the world, and especially how conservatives see liberals, we need to cover some background info on what that means. In particular, we need to talk about moral signalling as a memetic penomenon. Hopefully, this will help you understand some of your own thoughts, even before next post's discussion of conservative perceptions.

Let's start with a rich person, who wants to signal their wealth. Oldest signalling problem in the book: the rich person buys something very expensive, and utterly useless, which no one would possibly bother buying unless they had lots of money laying around.

Now let's crank it up a notch. The first rich person was the standard nouveau riche, flaunting their money about. The old money families want to signal that they are NOT the nouveau riche; they want to signal their social superiority. Thanks to years of expensive schooling, they are in no danger of being mistaken for poor, so they signal their old money by NOT buying useless expensive things. This is the classic example of countersignalling.

But it doesn't just stop with signalling and countersignalling. Perhaps the "old money" is only old by American standards, and European aristocratic families want to signal their superiority to the both the nouveau riche and American old money by buying even more expensive things. And then there's people even higher up...

We can extend to a whole continuum of wealth or status. At each point on the continuum, people try to signal slightly higher status by acting like the people ranked just above them. If we remove external constraints, e.g. wealth, then it becomes purely a game of acting like the people a few steps above you on the social ladder. The people at the very top act as trendsetters, and whatever trends they decide to adopt wind up propagating down the line. If the people at the top start wearing funny hats, it propagates down the line, and pretty soon everyone is wearing funny hats. Now imagine that status is not a simple line, but a complicated web, perhaps even an *inter*web, and you start to see how signalling games can propagate around memetically.

But we're not here to talk about games of signalling wealth. We're here to talk about games of signalling moral superiority.

Once upon a time, there was the LGBT community.

One day, someone came along and said "You know, we should signal our support for queer and questioning people too. How about, rather than just LGBT, we use LGBTQ?" Well, certainly nobody wanted to signal a lack of support for the queer and questioning communities. "LGBTQ" was clearly more inclusive of marginalized groups than "LGBT", and therefore morally superior. And so "LGBTQ" began to propagate through the memespace.

Now some people came along and said "But wait! There are other alternative sexualities too. What about asexuals?" And so "LGBTQA" entered the lexicon.

Around this same time, "intersex" was also added, and after a brief issue with a race condition, "LGBTQIA" appeared.

When I was writing this, I thought about writing "LGBTQRSTUV" as a joke. Fortunately, I did a google search beforehand, and found out that "LGBTQIAPK" is somewhat standard, although competing standards include "LGBTTQQIAAP" and "LGBTQQIP2SAA". Unfortunately it does not appear that NIST has addressed this yet. I'm poking fun a bit at the end, but I don't want to criticize too much - an awful lot of value comes about from moral signalling. The goal, for this post, is just to provide a little intuition for what moral signalling memetics look like. In a sense, it's intuition for your intuition.

Tuesday, November 15, 2016

Understanding Conservatives, for Liberals: Goals

Background: A lot of my liberal friends have suddenly realized over the past few days that they live in an echo chamber. Many of them want to know what typical conservative beliefs look like, beyond the small minority of fundamentalists and nationalists usually portrayed in liberal media outlets.

The first few posts in this series were mostly about current events. The remaining posts will shift focus. This post is a mid point; it's a good opportunity to lay out some objectives.

My main objectives for this series are as follows:
1. I want to provide liberals with a framework for understanding how conservatives actually think/feel. It's all too easy to imagine one's opponents as evil goblins who sit around thinking about how best to stir-fry babies. But most people, most of the time, do not see themselves as evil; almost everyone thinks they're the hero of the story. In order to engage productively in the real world, we need to understand how conservatives see the world and themselves, rather than imagine cartoon conservatives who hold meetings on how best to heat up the planet and fleece poor people.

(Side note: How best to heat up the planet and fleece poor people are actually very interesting theoretical problems. Unfortunately it's difficult to find people, either conservative or liberal, who share my interest in this sort of thing.)

2. With an understanding of how conservatives think/feel, we can talk about how conservatives' emotions and intuitions connect to conservative political positions (and liberal positions, too). The goal is to anticipate how a conservative will react *intuitively* to a new policy proposal. In practice, people make all sorts of logical arguments, but these only rarely change minds. More often, a person's stance on policy is determined by knee-jerk, intuitive reactions, and the arguments are filled in later. So if we want to engage productively, we should structure our arguments to fit the intuitions of the audience. Those intuitions will then work with us rather than against us, and the audience will argue *themselves* into agreement.

3. Address relative advantages/disadvantages of both liberal and conservative viewpoints. Arguing more effectively is one thing, and I'm sure many people reading this series are just interested in that aspect. But personally, I sometimes think about effective policy. In order to design effective policy, it's important to recognize the relative advantages on both sides of the political spectrum, and ideally integrate any relevant insights from both. Heck, it might even turn out to be useful for arguing more effectively, too.

Fortunately, (1), the framework for understanding, has already been done by a much better writer than myself: it's the zombies piece linked from the first post in this series. The series itself is mostly about applying that framework.

Sunday, November 13, 2016

What Might Trump Actually Do?

Background: This post is part of a series designed to help liberals get inside the heads of conservatives. My own views on Trump are here, and an overview of mainstream conservative beliefs is here. This post in particular mostly tries to evaluate likelihood of actions without getting into how good/bad those actions would be.

Remember where Trump comes from: he's a real estate developer who does megadeals. That identity has consumed him, it's who he is on deep level. He tells the investors whatever they want to hear. He hires competent people. He wheels and deals and greases palms, and he gets the project done (which may or may not be the same project he promised).

Those habits are not going to go away.

Step one to guessing what Trump will do is to recognize that voters are the investors. He's been saying whatever people want to hear for over a year. If we want real predictive power, we need to basically ignore anything that came out of his mouth. But anything with the word "contract" on it is probably reasonably reliable... bearing in mind that the exact wording of that contract matters, because it's likely to be intentionally misleading in places.

What follows is a brief tour of Trump's Contract with America. Before we dive in, it's important to have some idea of what's plausible, and where to be more skeptical. This document is the closest thing Trump had to a platform during the election. It's meant to pander to his base, and various specific political groups. That said, Trump's still a con man, so we should keep an eye out for misdirection, especially in the pandering parts. Also, there's a number pandering proposals which will certainly be struck down by Congress. On the other hand, in areas where Trump isn't clearly pandering to anyone specific and doesn't make promises which Congress will certainly strike, the "contract" probably lays out actual plans.

Let's dive in, section by section.

Section 1: Term Limits, Lobbying, etc...
This section opens with a congressional term limit proposal. Needless to say, that's unlikely to ever actually pass, but hey, he's welcome to propose it.

Next two items are a federal hiring freeze, and a rule that two old regulations must be removed for each new regulation added. Both of these are feasible, as they would require only an order from the sitting president. The exemption categories for the hiring freeze are pretty broad, but the regulation rule could have some real teeth.

Finally, there's a few items around lobbying and fundraising bans. It's not clear whether these could be implemented without congressional support (congresscritters are unlikely to support banning themselves from lobbying).

Section 2: Boo Trade, Yay Jobs, Fuck Da EPA...
The first four items in this section are all trade related, and include renegotiating NAFTA and the abandoning the TPP. All of them are feasible and do not require immediate Congressional support. Given Trump's background striking big real estate deals, I suspect he's eager to try negotiating a trade deal like NAFTA. However, the language in these items is ambiguous in places, and leaves plenty of room for Trump to maneuver as he gets to know the trade landscape.

The next two items involve trade-offs. In both cases, the Obama administration made trade-offs in favor of the environment over the economy, and Trump wants to reverse those. With republicans in Congress, this is feasible and will likely be a priority for congressional leadership as well.

Finally, Trump proposes dropping funding for UN climate change programs, and redirecting the money to US infrastructure. As grandstanding goes, this is almost comically conservative. Relatively speaking, the numbers involved probably aren't huge (for either the US or the UN). The current main targets of US funding to the UN are not climate programs, and the US is a significant but not top donor to UN climate programs.

Section 3: Boo Immigration & Miscellaneous...
First up, Trump promises to cancel every "unconstitutional" executive order issued by Obama. So basically just yelling "boo Obama" and then not really promising anything.

Next up, a promise to appoint new justices who will "defend the Constitution". Definitely signalling judicial restraint. Conspicuously absent here is any mention whatsoever of abortion or marriage - this strongly suggests Trump will completely ignore those issues.

The last three items are all immigration related. These sound pretty hard on immigration, but look closer and the most telling item is a promise to remove "two million" immigrants. Turns out, two million is roughly the same number the Obama administration removed, per term. Funny coincidence, that. Sounds like Trump is trying to con some voters.

Page Two: Proposed Bills for Congress
Since these are all items which involve Congress, Trump's team has a lot more freedom to pander here, with the expectation that Congress will strike down the bills. That means we need to be a lot more skeptical of these proposals, especially if it's unlikely to pass Congress.

2-1: Tax cuts. Republican president, republican Congress, there's going to be tax cuts. No surprise there.

2-2: Tariffs. Not sure how this one plays out in Congress; republicans haven't historically been very protectionist, but politics could change that.

2-3: Ambiguous, supposedly revenue-neutral infrastructure plan. Likely to turn into a pork-barrel Christmas tree once Congress gets their hands on it.

2-4: Charter schools etc. Long list of education-related ideas here; at least some are likely to pass.

2-5: Repeal Obamacare, streamline FDA approval. Again, an obvious target for a republican Congress.

2-6: Childcare/eldercare tax cuts: Likely to pass.

2-7: The wall. The wall's kind of a red herring (mostly symbolic), the proposal's mostly about stricter punishments. The most aggressive parts are unlikely to happen simply due to prison capacity. Also, republican congressional leadership has been fighting for years to gain ground among hispanics; I'll be surprised if they give up on that.

2-8: More money for law enforcement. Likely to pass, unlikely to change anything, probably going to be pork-heavy.

2-9: Security. Controversial part is screening immigrants for "compatible values", which could morph into anything from stricter background checks to blatant discrimination against muslims. Again, not sure how congressional republicans will respond to this, considering how many voters it would piss off.

2-10: "Clean up Corruption in Washington Act". Lol, yeah right.

A few things to note, right off the bat: there is absolutely no mention of marriage or abortion. He does mention appointing new justices, and says "defending the Constititution" will be a priority. That sounds like avoiding judicial activism, but it doesn't sound like overturning anything.

There is some serious talk about immigration. The promise to deport 2 million sounds flashy, but it's a red herring - that's roughly the same number Obama's administration deported, per term. Cancelling all federal funding to sanctuary cities also sounds serious, but remember that little federal money goes straight to towns anyway. The 2 million number in particular makes me think that the immigration talk is all a straight-up smokescreen; Trump's just going to keep the status quo.

I'm sure some people will argue that it's not what's said here that's worrying, it's the horrible things Trump might do that he hasn't said. In principle, sure. But in practice... I think if Trump were planning to really go extreme on immigration or abortion or gay marriage or what have you, he probably would have mentioned it in a document specifically designed to pander to his base (i.e. the document we're talking about here). If anything, in most of the scary areas, this document sounds like it's trying to appease the base without actually doing most of the things liberals are scared about.

The one thing I'd be most worried about is discrimination against Muslims looking to immigrate. That said, it's not something Trump can do without Congressional support, and it would be a very politically risky move for republicans in Congress.

Also, if you're worried about climate change, then keep worrying. Trump is definitely going to undo some environmentally friendly policies. There will be another post about that a bit later in this series.

Saturday, November 12, 2016

Conservative Views, for Liberals

Background: A lot of my liberal friends have suddenly realized over the past few days that they live in an echo chamber. Many of them want to know what typical conservative beliefs look like, beyond the small minority of fundamentalists and nationalists usually portrayed in liberal media outlets. This is the first in a serious of posts on the subject. There is also a prequel on my own (definitely not typical conservative) views here.

Part 1: The broader conservative viewpoint on Trump

This Slate Star Codex piece about zombies is a hard prerequisite for this post. Don't even bother reading this post until you've read that; it's better written, more useful, and the next few paragraphs assume you've read it. It's also medium-length, so give it a bit of time. That piece will give you a very good idea of how conservatives see the world, and why they act the way they do. The rest of this post will talk about those ideas in the context of current events.

Done with the zombies piece? Cool. Hopefully you have an interesting new perspective on conservatives in general. Let's talk about how that applies to this year.

This year in particular, elitism was a major additional dynamic. From the perspective of most conservatives, the zombie apocalypse is imminent... for themselves. But there's also a bunch of urban, coastal, college-educated elites who seem insulated from the zombies. Those elites don't need to worry about factories or mines closing down. Those elites don't need to worry about the rapidly shrinking demand for workers without a college degree. Those elites, from the safety of their ivory towers, would rather argue about what bathrooms to use than address the immediate economic problems of half the country. Those elites want to tell the conservatives what to do, but they don't care about the conservatives' economic problems. That's the feeling. (Although that's not the language - the "elite/anti-elite" phrasing is itself somewhat leftist. But that's the feeling.)

Hopefully you can see where somebody like Trump - low-brow, often crass, decidedly not a gentleman or scholar - might appeal to people with that sort of feeling.

And it's not just white males. Trump outperformed across the board among people without a degree - in fact, compared to 2012, he even slightly outperformed Romney among the latino/latina community. Apparently, one in three latino/latina voters either don't put much stock in Trump's racial rhetoric, think his anti-elitist attitude outweighs it, or both. This should not be surprising. By the numbers, minority communities share those zombie-apocalypse economic problems.

But don't think that anti-elitism was the only factor. As the examples below illustrate, standard conservative positions were still the main driver in the election. Elitism was more relevant than usual this year, but that doesn't mean it was the dominant factor for most Trump voters.

Part II: A few conservatives I know well...

a. My family
My family, including both parents, four aunts/uncles, and grandfather, are all self-employed. They all support LGBT rights, they all want free immigration, but what they really care about is regulation and taxes. Every one of them watched their health insurance premium go through the roof this year - that's what happens when insurers can't turn people away for pre-existing conditions. What they wanted most this year was to get rid of the ACA, and go back to the insurance they had before. They want regulatory reduction, they want spending cuts. Fun fact: self-employed people pay twice as much of their income into social security. No one in my family ever expects to see any of that money again. They've written off that loss - they just wish they weren't still stuck paying into it.

b. A few friends in the military
They're not big fans of gun control. Also, there's always a certain amount of tension between the military and the state department, so of course there's a list of gripes about things Hillary did as secretary of state. But I think what really matters for many of these people is group identity: many of them went straight into the military after high school, and I think Trump's anti-elitist identity really resonated with them (even though they might not use fancy words like "resonate"). Some (not all) of them were fairly racist back in high school, but that turned around pretty quick once they joined the military. Turns out there's a high fraction of racial minorities in the military.

c. Several other friends from high school
Health insurance hikes were a common theme, and anti-elitist identity was popular. On the democratic side, there's ongoing talk about making college more accessible to all. A number of my high school friends never went to college, don't want to, see most degrees as useless, and certainly don't want their hard-earned cash taken away to subsidize other people getting useless degrees. The phrase "underwater basket weaving" comes up from time to time. There's a number of jokes about what all the democrats do while the republicans are at work. I remember an article at one point about hipsters with various useless college degrees using food stamps to buy free-range chicken at Whole Foods. That's how my republican friends from high school view the democratic party.

Next post in this series is on what I think Trump will actually do

Friday, November 11, 2016

Why I Like Trump... AND Hillary

First, some general principles which underlie my view. I feel that the vast majority of the impact of a president comes from day-to-day work, mostly outside the spotlight, which is mostly unrelated to popular political issues. Furthermore, most of that day-to-day work is highly specialized, requiring special knowledge and skillsets. I believe that most presidents, regardless of party, perform far better during their second term primarily because they've acquired the requisite knowledge and skills.

On the flip side, the public mostly supports very bad ideas, on both sides of the isle. The vast majority of people, including myself and probably you, knows very little about economics, foreign diplomacy, legal procedures, or bureaucratic feasibility. (Democracy is still better than anything else we've tried, but it has major downsides.)

Put two and two together, and the ideal political candidate is somehow who tells people exactly what they want to hear, gets elected, and then completely ignores everything they told people.

In this election, Hillary was certainly the better candidate in terms of having the requisite knowledge and skills, more so than any other first-term candidate in recent memory. Unfortunately, Hillary also seems conscientious enough to be at least somewhat honest. She had to make promises in order to be electable, and she would have carried through on some or all of those. Now, this where you and I might disagree: you might think most or all of the things she promised would be great. I'm not going to bother digging in to that. But I will say that I thought Hillary's competence made her a great choice this year, and I would have loved a Hillary presidency.

Now back to Trump. If ever there was a complete huckster, a con-man who is master at the art of schmoozing and suckering, it's Trump. One of the two things that I really like about a Trump presidency is that there's no way in hell this guy is keeping his campaign promises. Maybe we'll get a two-mile "wall" on the Mexican border, with lots of fanfare, but there won't be anything with real impact. The other thing I really like about Trump is that he has an established reputation for bringing in the most competent people to do the actual work. That's not as good as having all the right knowledge and skills himself, as Hillary does. It's a higher-variance choice. But if Trump's presidency goes anything like I expect, he'll be offloading all the work to extremely competent people, and he'll spend his time going around blustering and bullshitting and generally telling the public whatever they want to hear. In the best case, the competent people will get a great deal of freedom to do what needs to be done, while Trump bullshits the media. That could be even better than a Hillary presidency. In the worst case... well, worst case is that Trump does what he said. Like I said, high variance.

On the bright side, the worst case is that Trump turns out to be completely honest and does exactly what he said. So not very likely. Reasonable to worry about, but unlikely, especially considering the political views he espoused back before running for president.

To wrap it up, I should mention that roughly 1 in 3 latino/latina voters voted for Trump (slightly more than Romney in 2012), so it's not just safe white guys who think the risk isn't too high.

Thursday, November 10, 2016

How to Implement a National Popular Vote

There was a project for a while around implementing a national popular vote without a constitutional amendment (I think this is it). Since states are free to apportion electoral votes as they see fit, a national popular vote just requires that states controlling a majority of electoral votes commit to give all of their votes to the winner of the popular. Under the project setup, states individually pass bills committing to go along with this scheme as soon as enough other states agree to it.

I think that project has the right general approach - tackling the problem state-by-state, achieving a simple majority, and routing around the challenge of a constitutional amendment. The problem is that the states with a relative advantage under the electoral college probably won't want to give that up. And you will need the support of those states - precisely because they control an outsized portion of electoral votes!

What we really need is a setup that makes the national popular vote a Nash equilibrium. States with extra electoral votes need to be able to gain an advantage if they agree to a popular vote, but other states don't. So the trick is to set up a deal between two (or more) states, which puts them at an advantage relative to other states. I haven't figured out a perfect way to do it yet, but I may have a solution which is good enough.

Here's the idea (there are others): rather than a state allocating their electorals to the popular winner iff enough states agree, how about a state allocates all of their electorals unconditionally to the winner of the popular vote among states which have agreed?

Say, for instance, Alaska decides to form the "Alaska voting block". All of Alaska's electoral votes go to the winner of the popular vote within the block. Any state can join the block by agreeing to the same terms: all of their electoral votes go to the block winner. So now California comes along and is like "hey, we can pick up 3 electoral votes basically for free by joining Alaska's voting block. Let's do it." And then Texas takes a look and says "you know, California's popular vote is actually pretty split... we could swing the popular vote of the Alaska block if we join." And then New York hops on board. And so on. Pretty soon, the Alaska voting block is the dominant factor in an election, and all states need to join in order to have any say at all in the election.

To make this happen, you only need to pass a single bill, in a single state, and it doesn't even matter which state. Plus, whichever state is first gets to name the voting block!

Saturday, November 5, 2016

Game Theory vs The Halting Problem

The usual presentation of the Halting Problem goes something like this. Suppose we have some computer program H, which takes in the code for another program P. H(P) outputs "yes" if P, when run, will eventually stop (a.k.a. "halt"), or "no" if P will run forever.

Then we take the code for H, and use it to build a tricky program called G. G first runs H on itself - it gets H(G). If H(G) is "yes", then G immediately enters an infinite loop and never halts. If H(G) is "no", then G immediately halts. So, when H runs on G, G will halt if and only if H(G) is "no". But the whole point of H is that for any program, H says "no" if and only if the program does not halt.

In short, G breaks H. Given any supposed halting-predictor program H, we can easily write a program G which breaks it. Thus there is no perfect halting-predictor program: the halting problem is "undecidable".

Using the halting problem as a starting point, theoretical CS classes go on to prove all sorts of other problems are undecidable. I'm going to go in a different direction, one which I haven't seen before: I want to argue that the standard halting problem asks the wrong question.

First, let's rewrite the halting problem in game-theoretic terms. Rather than two programs, the game has two players: H and G. G has two choices: halt, or don't halt. H's goal is to predict whether G halts or not; H wins if it says "yes" and G does halt, or if it says "no" and G doesn't halt. G, on the other hand, want trick H: G wins if H says "no" and G halts, or loses if H says "yes" and G runs forever.

Here's a table to show the rules:

halt run
yes H wins G wins
no G wins H wins

Now, when computer scientists say the halting problem is "undecidable", what they're really saying is that this game has no equilibrium in pure strategies. What does that mean? Well, a "pure" strategy means that H and G always do the same thing. In this case, if H always says "yes", then G always wants to run forever; but if G always runs forever, then H always wants to say "no"; but if H always says "no"... You see the issue. There is no equilibrium in pure strategies, because either H or G will always want to change their choice.

Remember the movie "Beautiful Mind", about John Nash? Well, John Nash's original claim to fame was a proof that any game with finite players and finite choices has at least one equilibrium. (In fact, for the simple two-player zero-sum case here, Von Neumann proved existence of an equilibrium even earlier.) But there's a catch: the equilibrium isn't always in pure strategies. In order to find an equilibrium, players may need to randomize their choices. Strategies with randomized choices are usually called "mixed" strategies.

For the halting problem game above, there is one very simple equilibrium in mixed strategies: H picks "yes" or "no" randomly by flipping a coin, and G chooses whether or not to halt randomly by flipping its own coin. Half the time, the choices will match, and H will win. Half the time, the choices will not match, and G will win. As long G splits 50/50, H cannot do any better by picking "yes" or "no" more often. As long as H splits 50/50, G cannot do any better by picking to halt or not halt more often.

In general, no matter what program H is playing against, whether it's G or any other program, there is always some optimal strategy for H. For some programs, especially programs which aren't actively working against H, a pure strategy will work. For other programs, like G, a mixed strategy is needed. (Technical note: I'm brushing some things under the carpet here, specifically involving jump discontinuities and limits in strategies, but these are the sorts of things which can usually be worked around.)

So let's reformulate a game-theoretic version of the halting problem. The new goal is to write a (nondeterministic) program H, which takes in another (possibly nondeterministic) program P, and returns either "yes" or "no". For any program P, H should select "yes" or "no", possibly randomly, in order to maximize its win rate, i.e. maximize the probability that H correctly guesses whether P halts.

Is this problem solvable? I don't know yet. Maybe it will turn out that this new halting problem is still undecidable. But the good news is, we're at least asking a reasonable question this time, not just searching for pure-strategy equilibria. And if it turns out that this game-theoretic halting problem is solvable, then the consequences would be significant: the undecidability of many other problems is based directly on the halting problem, or proven using the same method. The same technique should provide a practical approach to a wide variety of other problems, if the new problem is indeed solvable.

Finally, for anyone reading this who is familiar with both game theory and the undecidability proof of the halting problem: this should all have been fairly obvious. There is no way that this hasn't been thought of twenty times over. So why can't I find any literature on it? Am I not searching for the right terms? What results are known?