Meteuphoric

Oftentimes, the correct response to an argument is ‘your argument appears after cursory investigation to make sense, however the fact that many smart people have never mentioned this to me suggests that there are good counterarguments, so I remain unconvinced’.

I basically never hear this response, which suggests that there are good counterarguments. Or alternatively that it is unappealing to respond accurately in such cases. The latter seems very plausible, because it suggests one cannot assess any argument at the drop of a hat.

If so, what do people actually say instead? My guess is the first argument they can think of that points in the direction that seems right. This seems unfortunate, as the ensuing discussion of that counterargument that nobody believes can’t possibly resolve the debate, nor is of much interest to anyone.

Effective altruism is all the rage. And as we all know, a highly effectively altruistic thing to do is bring about more effective altruism. However, the optimal mixture of activities for promoting effectiveness is not identical to that for promoting altruism. They are upheld by different norms, and require different efforts to expand and arguments to proliferate. Which brings us to a need for prioritization. So, which is better: effectiveness or altruism?

By this I mean, which is the better recipient of marginal promotion effort?

For background, I’m visiting the Oxford Effective Altruists at the moment. Conversations here made it salient to me that my implicit views on this question have changed a lot since I was a teenager, and also differ from others’ in the EA movement. So I thought I’d lay them out.

When I was younger, I thought altruism was about the most promising way to make the world better. There were extremely cheap figures around for the cost to save a human life, and people seemed to not care. So prima facie it seemed that the highly effective giving opportunities were well worked out, and the main problem was that people tended to give $2 to such causes occasionally, rather than giving every spare cent they had, that wasn’t already earmarked for something more valuable than human lives.

These days I am much more optimistic about improving effectiveness than altruism, and not just because I’m less naive about cost-effectiveness estimates. Here are some considerations:

Past success

So far, effectiveness has bought more than altruism for EAs, arguably. That is, giving an ordinary amount highly effectively will buy more good than giving a highly altruistic amount in an ordinary way. The difference between a normal amount of giving and a highly altruistic one is around 20x by my reckoning. I’m not sure what the mean charity is, but my impression is that it is less than 1/20th as useful as the best things we know of. Better data on this welcome.

Furthermore, perhaps the quest for producing altruistic value can learn from the more popular quest to produce selfish value. There have been efforts both to encourage people to be more selfish (see Ayn Rand) and to find cost-effective ways to direct selfish efforts. As far as I can tell, the former might even produce negative selfish value for people, whereas mankind has become reasonably good at finding selfishly valuable investments. But you might reasonably retort that the quest for selfish value has less to gain from changing values, given the ubiquity of selfishness. Also, selfish people have little reason to spread selfishness. The spread of useful understanding for prioritizing selfish actions is perhaps largely because information is reusable and hard to contain.

Neglect

Effectiveness is much rarer and altruism more common than I assumed when younger. Altruism is actually done to extremes reasonably often. Perhaps more importantly, pushing for effectiveness is much rarer than pushing for altruism. People are neither naturally very effective in their altruistic endeavors, nor driven to look for effectiveness and praise it in others. Altruism on the other hand is universally liked, and celebrated in others. It is searched for on first dates, and praised at funerals. Quite a lot of effort has been put into encouraging it.  

ROOM FOR MORE FUNDING (marginal value)

There appear to be high value ways to encourage effectiveness, such as by doing or funding the research others need to be effective, and making it appealing to use. For instance, it seems that producing trustworthy prioritization of high level causes (e.g. poverty alleviation vs. technological progress) could easily produce more value from redirecting others’ funds than one competent person’s lifetime of giving to AMF, and I think would take less than one competent person’s lifetime. So I think such prioritization is a better investment than AMF, which makes it a good investment.

In encouraging altruism, I don’t know of any particularly promising directions. The most effective I have seen is probably Peter Singer‘s method, of writing for a popular audience, encouraging readers to resolve their moral inconsistencies in favor of altruism. I’m not sure if you can do such a thing again however; the low hanging fruit may be eaten, and Singer may have been especially well equipped to pick them.

UPSIDE

There’s only so altruistic a person can be, and the limit is within about an order of magnitude of where a lot of people are. We know what this would look like: about ten times better than what a Giving What We Can member’s achievements looks like. It seems on the other hand that there are much higher reaches of good prioritization that we haven’t seen – it isn’t clear how good the best visible opportunities could be, if we could spot them.

Flow-through effects

Effectiveness probably leads to altruism more than the other way around

An intuitive story is that if a person really cared about helping, they would be naturally driven to find exceptionally good ways to do so. The fact that ‘effective altruism’ is a thing suggests this is not so, for whatever odd/cynical reason.

On the other hand, learning that one can can help others more cheaply than one thought, and having more confidence in the reliableness of the estimates producing this belief, seems to encourage altruism. People are moved by Peter Singer’s story of the child in the pond in a way they would arguably not be if the passerby had to make a very substantial sacrifice and was less confident about the outcome of his action. On the other hand, on the way to improving effectiveness, we often learn that things are less effective than we thought, which will presumably undermine altruism by this reasoning.

Economics of information

An abstract consideration in favor of expecting higher returns to encouraging effectiveness is that much progress on effectiveness is in the form of producing insights and other information, which can be shared, reused, and built upon.

Effectiveness is more valuable when applied broadly

My impression is that in every part of the world, not just in philanthropy, better systems for assessing opportunities and making good decisions promise great gains. Altruism is not the only place where where contemporary efforts are primitive. Progress in prioritizing altruistic endeavors should often be applicable to this larger worthy project. Progress in effectiveness often comes with the accumulation of insights and other information, and information can be shared, reused, and built upon in a way that a pile of altruists can’t be so well. If people weren’t altruistic at all, but were basically cooperative (for selfish reasons), and they could discern of good policies from bad as perfectly as publicly available information would allow, I think this would make the world much better.

Altruism too appears to have many benefits outside of philanthropy. If people were more considerate of one another in general, they would avoid many harms from externalities and failures to coordinate. My sense is that within the plausible ranges of change however, these gains are smaller.  If everyone was extremely well meaning but as just as uncertain about how best to act as we are today I would also expect things to go better but not nearly as much.

One thing that contributes to these intuitions is my impression that the difference in social effects between a person who is nice enough to have enjoyable relationships and one who is particularly kind at cost to themself is smaller than the difference in effect between a person who has a good sense of how to achieve what they want, and one who acts without regard to strategy.

This intuition also depends on one’s answer to the question of why things go wrong in the world. Some ills are intentional malice, some are unchecked side effects of selfishness, and some are from ignorance and error. Altruism makes a dent on malice and selfishness, whereas progress in effectiveness diminishes the costs from ignorance and error.

I’m inclined to think that ignorance and error make up a large part of the problem, though not obviously most. My guess is selfishness makes up most of the rest, and malice is small.

Naively, avoiding ignorance and error should be an easier sell: everyone wants their own choices to be better, whereas I’d expect more hesitance about increasing their willingness to sacrifice for others. However after effort has been put into both, this may not be true on the margin. Also, there are many ways to make people more altruistic in effect that don’t involve changing their values per se, e.g. arranging for them to feel better when they are being altruistic, or to have social support for it.

Measurement

Ironically, cost-effectiveness seems hard to measure. You can go by the current value for money estimate of your best cause, but this has a tendency to decrease as the quality of the estimate increases. This makes it a less good target for improvement, as it is harder to know when you are improving it, and easier for standards to decline without anyone noticing (this should go under the next section). On the other hand, there are relatively good (if imperfect) measures of altruism, such as ‘what fraction of your income do you give to charity?’.

Long term durability

One last consideration is how changes in altruism and changes in effectiveness will last. I am not optimistic about changes in values lasting well. At first glance, if it was easy enough for you to shift others’ values, you might expect them to shift again the next time someone pushes on them. Perhaps the situation is more complicated though. For instance, perhaps altruism is stickier than other values: it just inherently makes more sense, so once you have it you keep it. Or perhaps it spreads more easily: only altruists care about spreading their values to others. But in these kinds of situations, spreading altruism is still not as helpful as it looks, because everyone will end up there soon enough anyway.

The same goes for people valuing ‘being effective’ highly. However as I mentioned earlier, changes in effectiveness often involve insights that can accumulate. This makes it harder for progress to be completely reversed. Humans have come to not care much at all about hand-weaving, but still our accumulated insights on it mean that whenever anyone wants to weave stuff by hand, they can do it very effectively.

***

In summary, I think effectiveness is relatively neglected, though it has a better track record of success, more room for funding, better upsides, and produces better flow-through effects.

There is considerable disagreement over which things are roughly good or bad for the world. Encouraging same-sex marriage, speeding the cogs of ‘progress’, compelling would-be immigrants to live in third world shanty towns, and building self-replicating nanobots are considered by different parties to be either substantially positive or substantially negative. Such questions are sometimes loudly and fiercely debated.

There also seems to be much disagreement over how good different good interventions are. Is promoting sustainable agriculture a crucial last-ditch countermeasure in the face of impending societal destruction, or harmless feel-goodery? This type of disagreement seems very widespread, given the huge variety of endeavors people support with their money and time. However such disagreements are much quieter; for the most part people politely get on with their own causes. You are much more likely to be handed a brochure explaining the virtues of buying beads from women in the developing world than one earnestly arguing that this isn’t particularly valuable.

This seems like some evidence of a tragedy of the commons that one might expect prima facie. Even if you can hand out don’t-expend-effort-on-veganism brochures more cost-effectively than the vegans can hand out theirs, you can’t capture the efforts of the would-be vegans for your own preferred cause. The efforts you divert will probably be spread across many other causes.

Ultimately then, the public discussion of causes should become full of hooks to harvest goodwill and efforts, with few working to rescue people from bad causes and replenish the stock. The main demand for counterarguments should come from the people who wish to avoid joining causes themselves. But they don’t need strong cases, just some excuse to say as they decline to join. And indeed, the retorts used for such purposes seem rarely strong enough to change a person’s mind.

One big exception to this tragedy of the commons situation would be causes where people are quite divided – where leaving the pro-X side means joining the anti-X side, and where the anti-X supporters would be happy enough to diminish the X side anyway. These causes tend to be of the first sort I mentioned: where the disagreement is over whether X is excellent or hellish, rather than good or better. For these issues there are no tragedies of the commons, and we should expect to see elaborate cases made for both sides. As of course we do.

Another situation where you wouldn’t see this behavior much is one where most of the causes are considered similarly good by most people, but there are still a few much worse ones. Then a person who understands this state of affairs can redirect most of the efforts they divert from the bad causes toward causes that seem best to them. So if the tragedy of the commons account is right, it seems we are not in that situation. People either think a relatively narrow range of causes are quite good, or that so many causes are quite good there aren’t bad ones around to divert people from.

I don’t know that this is most of the explanation, though it seems it must happen to some degree. People may also prefer not to speak against anything that seems ‘good’ – and measure good against some consistent zero point, not against the opportunity cost. It could also be that the easiest way to compare things is for different parties to make a pro- case for each, then to compare those. Though it seems in practice that would leave out many negative details about each cause. Any other theories? Any reasons to believe or disbelieve this one? Note that the heroes of this story would include the selfless people who go out on the internet night after night, with little to gain by it, tirelessly assaulting any position that shows weakness.

Users of languages improve them by adding new words. Often languages would also be improved by taking away words – not so much such that nobody ever uses the word, but just so it is not necessary to convey a certain meaning. For instance since it is very rare for a claim to be meant with one hundred percent certainty, it would be more efficient if a speaker had to add ‘certainly’ in that case, rather than adding ‘probably’ or ‘I think’ for every other case where they are only claiming a best guess.

Robin Hanson explains the retention of such extra words by people being stuck in a signaling game: if you don’t add the extra words in the normal cases, then people mistake you for those who really are making extreme claims, since everyone else adds the extra words when they are making the ordinary claim.

That seems correct. But why would this begin? How did everyone ever come to use the long, complicated sentences for saying ordinary things and the succinct ones for claims that almost nobody ever wants to make? Perhaps we just especially notice the half of times when the signals are assigned the inconvenient way around?

Another plausible theory is that the usual cases are complicated, whereas simple cases are unusual and extreme. Adding qualifications allows you to say something much more specific and nuanced. And we usually want to refer to specific, nuanced state of affairs. However for uncontroversial issues nobody seems to have much trouble interpreting simple statements as the obvious more complex ones. If you say ‘roofs in England have steeper slopes than roofs in Australia’ most people would interpret this as indicating some sort of general tendency for steeper roofs in England. If you want to say the extreme thing, you have to say something like ‘literally every roof in England is steeper than every roof in Australia – no, I mean it literally. I know, it sounds insane! Look, I’ll send you these seven review articles summarizing the research.’ Whereas if a person says ‘women are less interested in engineering than men’, listeners seem to often interpret this in such a way as to make raising counterexamples a reasonable response.

So it seems closely connected with controversy. It appears that people making uncontroversial claims are much more scared of being misunderstood and thought controversial than people making controversial claims are the opposite. Or at least they are willing to pay more to avoid being misunderstood. So the controversial people can snap up the short, efficient signals without worrying about confusion, leaving the uncontroversial people to run away to awkward, qualified corners of signal space.

But it’s not obvious why uncontroversial people should be more afraid. The probability of misunderstanding seems to be much lower for them, because there are so few controversial people. A random apparently extreme claim is much more likely to be an error of speech than a random apparently moderate claim is. So to make the risk from potential misunderstandings higher for people making moderate claims, the cost of being misunderstood needs to be a lot higher for them. Why would this be?

Honesty

One answer is that more controversial claims are just more costly to make – everyone would prefer to make non-controversial claims, but some people are forced to make controversial claims by honesty. If they are misunderstood, all the better for them. This seems to have an element of truth, but doesn’t account for the apparent glee many people exhibit while making controversial claims.

Taking back

Another thing you might wonder about such models is why anybody should be much worried, given the possibility of clarification. Nobody fears being too misunderstood about roofs, in part because if the audience appears to be taking the statement the wrong way, it is cheap to clarify. Most conversation seems to involve plenty of this. Yet it seems to me at least that if you say something apparently controversial you are much less allowed to take it back than if you say something non-controversial. The audience is suspicious. Perhaps your slip of the tongue is  taken as telling, regardless of your conscious corrections. Whereas saying something uncontroversial first is not taken as much of a sign about you. If you correct your non-controversial statement to a controversial one, listeners are generally willing to allow that you are controversial.

Selection

Another possibility is that people are basically selected into making extreme claims or not according to whether they are terrified of other people criticizing them. Naturally the uncontroversial people are more scared of being misunderstood as controversial because this is what being uncontroversial is about.

Signaling

When talking about controversial topics, your words are taken as a sign of what side you are on. Perhaps to show that you are on the side considered more virtuous, you need to pay higher costs. People who believe controversial things might prefer to look like those who don’t, but their lack of passion for the cause means they can’t be bothered adding all the qualifiers, or feel more uncomfortable doing so. The people who truly believe the ‘virtuous’ claims can be bothered, and do so to set themselves apart.

This would also explain why listeners wouldn’t allow you to take back controversial sounding claims – if you could say things cheaply and only elaborate when you were caught, it would be much cheaper, so may not signal your devotion correctly.

It doesn’t however explain why you would have to do this particular costly act, instead of some other. But that is rarely explained I think.

Status

I discussed this with Robin, and he suggested a status explanation: making controversial claims is generally a bid for status. If roofs were controversial, then by claiming that roofs in Britain are steeper than roofs in Australia, you are taking the initiative to ally your group with ideas that the current powers consider enemy. So you are making a bid to increase your own importance in the group, as well as trying to change the status of other groups and ideas. People respond strongly to bids of status. If you want to do something that looks a bit like a bid for status without being beaten down into your place by all around, you have to do some sort of accepted submission dance at the same time to make it not like a bid for status. This seems intuitively somewhat plausible, but leaves even more questions unanswered: why does this behavior convince people around you that you are not bidding for status? Why doesn’t clarifying later work? Why are you so aggressively interpreted as bidding for status anyway, instead of making the nearest probable, non-controversial claim? 

Responses

What to do about this? One intuitive possibility is to replace words instead of just removing some. For instance we could introduce the usage:

Qualifiedly: with all due qualifications, and some more

Then instead of saying ‘he sometimes seems a bit silly, to me, but I might just be biased, and I don’t really know him that well, and I’m sure we all have different preferences anyway, and that might be relevant..’ you could just say ‘he seems qualifiedly a bit silly’.

I’m not sure if this would work in any of the above models though. It probably won’t work if the point is to pay a cost, or do a special dance. In the signaling case there is probably no way out, unless non-controversial people can find a cheaper way to set themselves apart, that is still expensive enough to keep out the controversial people.

If the important thing is just to say something that can’t be interpreted badly the first time, something like this has a better chance of working.

Turning it around, how you expect such a change of words to go says something about which theory of inefficient qualifications is correct.