Meteuphoric

Ben Kuhn makes some reasonable criticisms of the Effective Altruism movement. His central claim is that in the dichotomy of ‘really trying’ vs. ‘pretending to try’, EAs ‘pretend to really try’.

To be explicit, I understand these terms as follows:

‘Really trying’: directing all of your effort toward actions that you believe have the highest expected value in terms of the relevant goals

‘Pretending to try‘: choosing actions with the intention of giving observers the impression that you are trying.

‘Pretending to really try‘: choosing actions with the intention of giving observers the impression that you are trying, where the observers’ standards for identifying ‘trying’ are geared toward a ‘really trying’ model. e.g. they ask whether you are really putting in effort, and whether you are doing what should have highest expected value from your point of view.

Note the normative connotations. ‘Really trying’ is good, ‘pretending to try’ is not, and ‘pretending to really try’ is hypocritical, so better than being straight out bad, but sullied by the inconsistency.

I claim Effective Altruism should not shy away from pretending to try. It should strive to pretend to really try more convincingly, rather than striving to really try.

Why is this? Because Effective Altruism is a community, and the thing communities do well is modulating individual behavior through interactions with others in the community. Most actions a person takes as a result of being part of a community are pretty much going to be ‘pretending to try’ by construction. And such actions are worth having. If they are discouraged, the alternative will not be really trying. And pretending to try well is almost as good as really trying anyway.

Actions taken as a result of being in a community will be selected for being visible, because visible actions are the ones you will be able to pick up from others in the community. This doesn’t necessarily mean you are only pretending to try – it will just happen to look like pretending to try. But actions will also probably be visible, because by assumption you are driven to act in part by your community membership, and most ways such motivation can happen involve a possibility of others in the community being aware of your actions.

Many actions actions in communities are chosen because others are doing them, or others will approve of them, or because that’s what it seemed like a good community member would do, not because they were calculated to be best. And these dynamics help communities work and achieve their goals.

Of course people who were really trying would do better than those  who were only pretending to try. As long as they could determine what a person who was pretending would do, a person who really tries could just copy them where it is useful to have pretending-to-try type behaviors. But there is no option to make everyone into such zealous tryers that they can do as well as the effective altruism movement without the social motivations. The available options are just those of what to encourage and what to criticize. And people who are pretending to really try are hugely valuable, and should not be shamed and discouraged away.

A community of people not motivated by others seeing and appreciating their behavior, not concerned for whether they look like a real community member, and not modeling their behavior on the visible aspects of others’ behavior in the community would generally not be much of a community, and I think would do less well at pursuing their shared goals.

I don’t mean to say that ‘really trying’ is bad, or not a good goal for an individual person. But it is a hard goal for a community to usefully and truthfully have for many of its members, when so much of its power relies on people watching their neighbors and working to fit in. ‘Really trying’ is also a very hard thing to encourage others to do. If people heed your call to ‘really try’ and do the ‘really trying’ things you suggest, this will have been motivated by your criticisms, so seems more like a better quality of pretending to really try, than really trying itself. Unless your social pressure somehow pressured them to stop being motivated by social pressure.

I also fear that pretending to try, to various extents, is underestimated because we like to judge other people for their motives. A dedicated pretender, who feels socially compelled to integrate cutting edge behavioral recommendations into their pretense can be consequentially very valuable, regardless of their virtue on ethical accounts.

So I have argued that pretending to try is both inevitable and useful for communities. I think also that pretending to really try can do almost as well as really trying, as long as someone puts enough effort into identifying chinks in the mask. In the past, people could pretend to try just by giving some money to charity; but after it has been pointed out enough times that this won’t achieve their purported goals,  they have to pick up their act if they want anyone (themselves included) to believe they are trying. I think a lot of progress can come from improving the requirements one must meet to look like one is trying. This means pointing out concrete things that ‘really trying’ people would be doing, and it hopefully leads to pretending to really try, and then pretending to really really try. Honing the pretense, and making sure everyone knows what the current standards are, are important goals in a community with goals that might be forgotten if we undervalue pretending to really try.

I read Michael Huemer’s book Ethical Intuitionism at Bryan Caplan’s suggestion. Before reading it I thought that Ethical Intuitionism seemed an unlikely position and that Bryan and Michael seemed like smart guys, so I hoped I might be persuaded to significantly change my mind. I still believe both of the premises, but did not get as far as changing my mind, so here I’ll report back on some of my reasons. I actually read it a while ago, and am reconstructing some of this post from memories and scribbles in margins, so apologies if this causes any inaccurate criticism.

Ethical intuitionism is the position that there exist real, irreducible, moral properties, and that these can be known about through intuition. i.e. some things are ‘good’ or ‘right’ independent of anyone’s feelings or preferences, and you can learn about this through finding such ideas in your head. I found Huemer’s case for it well written, thought provoking, and for the most part valid. However I thought it had some problems, and was overall not compelling.

Huemer divides metaethics as into five types of theories then argues against non-cognitivism, relativism, nihilism, and naturalism. The last theory standing is intuitionism, which Humer goes on to defend from the criticisms it has been attacked with before. Note that these metaethical theories are both theories of what ethics is and theories of how we can know about it.

I

The first problem is related to intuitionism being primarily a theory of how we can know about ethics (through intuitions), while being hazier on what exactly ethics is, in terms of anything else. It’s something real that we can learn about through our intuitions. It is normative. If the theory contained any more concrete specification of the nature of ethics or how it got there, I think it would become more obviously subject to the same criticisms made of other metaethical theories. For instance, this is one of Huemer’s criticisms of subjectivism:

Fourth, consider the question, why do I approve of the things I approve of? If there is some reason why I approve of things, then it would seem that that reason, and not the mere fact of my approval, explains why they are good. If I approve of x because of some feature that makes it desirable, admirable, etc.,  in some respect, then x’s desirability (etc.) would be an evaluative fact existing prior to my approval. On the other hand, if I approve of x for no reason, or for some reason that does not show x to be desirable (etc.) in any respect, then my approval is merely arbitrary. And why would someone’s arbitrarily approving of something render that something good?

This argument does not seem specific to subjectivism. Take any purported source of morality S. If there is some reason for S justifying what it does, one could similarly say that that reason, and not the mere fact of being allowed by S, explains why they are good. And if S has no reason, then it is arbitrary. And why would arbitrarily being allowed by S render something good?

If we were clearer on the origins of goodness in the intuitionist account, doesn’t it fall prey to one side or the other of the same argument? This argument appears to rule out all sources of goodness.

Perhaps the difference with the intuitionist account here is that it is ok for it to be arbitrary, because while goodness is arbitrary, it isn’t caused by some other arbitrary thing. You can answer ‘why would arbitrarily being good render something good?’, by pointing out that this is logically necessary, whereas on the other accounts you are seeking to equate two things that are not already identical.

On the intuitionist account you are also trying to equate two non-identical things I think – I have merely labeled them ambiguously above. On the one hand, you have ‘goodness’ which is a property you receive statements about via your intuition, and on the other you have ‘goodness’ which implies you should do certain things. If these were not distinguishable, there would be no debate. But it could still be that the goodness which means you should do certain things is primary, and causes the goodness which you observe. Then the former goodness would be arbitrary, but nothing else arbitrary would ‘make’ good things good.

So perhaps the argument for intuitionism is that arbitrariness alone is ok, but it is not ok for a thing to be caused by other arbitrary things. Or perhaps only moral things shouldn’t be caused by other arbitrary things? Or only goodness? I’m not sure how these things work, but this seems very arbitrary, and I have no distinguished way to decide which things should be arbitrary and which things principled.

II

Huemer argues that we should not disregard ‘intuitions’ as vague, untrustworthy, or unscientific. He points out that most of our thoughts and beliefs are built out of intuitions at some level. Our scientific beliefs rest on reasoning that is ultimately justified by a bunch of stuff turning up in our heads – visual perceptions, logical rules, senses of explanatory satisfaction – and us feeling a strong inclination to trust it.

I think this is a good and correct point. However I don’t think it supports his position. On this conceptualization of things, non-intuitionists also reach their ethical judgements based on intuitions. They just take into account a much wider array of intuitions, and build more complicated inferential structures out of them, instead of relying solely on intuitions that directly speak of normativity.  For instance, they have intuitions about causality, and logical inference, and their perceptions, and people. And these intuitions often cause them to accept the claim that people evolved. And they have further intuitions about how disembodied moral truths might behave which make it hard for them to imagine how these would affect the evolution of humans. And they infer that such moral truths do not explain their observed mental characteristics, based on more intuitions about inference and likely states of affairs. And people do this because they find such methods more intuitive than the isolated moral intuitions.

So it seems to me that Huemer needs an argument not for intuitions, but for using very direct, isolated intuitions instead of indirect chains of them. And arguments for this more direct intuitionism seem hard to come by. At the outset, it seems to undermine itself, as it is clearly not what most people find intuitive, and so appears to need support from more complicated constructions of intuitions, if it is to become popular. However this is not a very fatal undermining. Also, indirect inferences from many intuitions woven together have been very fruitful in non-moral arenas, though whether better than direct intuitions is debatable.

In sum, Huemer makes a good case that everything is made of intuitions, but this doesn’t seem to normatively support using short chains of intuitions over the longer, more complex chains that persuade us for instance that if there was an ethical reality out there, there is no reason it would interfere with our evolution such as to instill knowledge of it in our minds.

III

One might argue that Huemer often begged the question, for instance claiming that other ethical theories are wrong because the ethical consequences are contrary to our ethical intuitions. This is perhaps an unsympathetic interpretation – usually he just says something like ‘but X is obviously not good’ – however, I’m not sure how better to interpret this. Since all thoughts can be categorized as intuitions, it would be hard not to use some sort of intuitions to support an intuitionist conclusion. However given the restatement of his position as a narrower ‘direct intuitionism’, it would be nice to at least not use just those kinds of direct intuitions as the measure.

***

The issues of how evolved creatures would come to know of moral truths, or how moral truths would have any physical effect on the world at all, seem like big problems with intuitionism to me. Huemer addresses them briefly, but doesn’t go into enough depth to satisfactorily resolve these problems I think.  I mean to write about this issue in another post.

So in sum, ethical intuitionism seems subject to similar criticisms as other metaethical theories, but avoids them by being nonspecific or begging the question. ‘What do my intuitions say?’ seems like an unfair criterion for correct ethical choices in a contest between intuitionism and other theories. The arguments for intuitions being broadly important and trustworthy do not support this thesis of intuitionism, which is about relying on isolated immediate intuitions over more indirect constructions of intuitions. A vague case can be made for moral intuitions evolving, but I doubt it works in detail, though I have also not yet detailed any argument for this.

Destruction of tropical rainforest, as prevented by Cool Earth (Photo credit: Wikipedia)

Here is the second post summarizing some of the climate change research I did at CEA last summer (the first is here). Below are links to outlines of the reasoning behind a few of the estimates there (they are the ‘structured cases’ mentioned in the GWWC post).

The organizations were investigated to very different levels of detail. This is why Sandbag for instance comes out looking quite cost effective, but is not the recommendation. I basically laid out the argument they gave, but had almost no time to investigate it. Adding details to such estimates seems to reliably worsen their apparent cost-effectiveness a lot, so it is not very surprising if something looks very cost effective at first glance.

The Cool Earth case is the most detailed, though most of the details are rough guesses at much more detailed things. The cases are designed to allow easy amendment if more details or angles on the same details are forthcoming.

As a side note, I don’t think GWWC plans on more climate change research soon, but if anyone else is interested in such research, I’d be happy to hand over some useful (and neatly organized!) bits and pieces.

Cool Earth

Solar Aid

Sandbag

Alvin Plantinga‘s ‘evolutionary argument against naturalism‘ (EAAN) goes like this:

1. If humans were created by natural selection, and also not under the guidance of a creator (‘naturalism’), then (for various reasons he gives) the probability that their beliefs are accurate is low.
2. Therefore believing in natural selection and naturalism should lead a reasonable person to abandon these beliefs (among others). i.e. belief in naturalism and natural selection is self defeating.
3. Therefore a reasonable person should not believe in naturalism and natural selection.
4. Naturalism is generally taken to imply natural selection, so a reasonable person should not believe in naturalism.

This has been attacked from many directions. I agree with others that what I have called point 1 is dubious. However even on accepting it, it seems to me the argument fails.

Let us break down the space of possibilities under consideration into:

• A: N&T: naturalism and true beliefs
• B: N&F: naturalism and false beliefs
• C: G&T: God and true beliefs

The EAAN says conditional on N, B is more likely than A, and infers from this that one cannot believe in either A or B, since both are included in N.

But there is no obvious reason to lump A and B together. Why not lump B and C together? Suppose we believe ‘natural selection has not produced true beliefs in us’. Then either natural selection has produced false beliefs, or God has produced true beliefs. If we don’t assign very high credence to the latter relative to the former, then we have a version of the EAAN that contradicts its earlier incarnation: ‘natural selection has not produced true beliefs in us’ is self-defeating. So we must believe that natural selection has produced true beliefs in us*.

What if we do assign a very high credence to C over B? It seems we can just break C up into smaller parts, and defeat them one at a time. B is more likely than C&D, where D = “I roll 1 on my n sided dice”, for some value of n. So consider the belief “B or C&D”. This is self-defeating. As it would seem “B or C&E” is, where E is “I roll 2 on my n sided dice”. And so on.

By this reasoning, if there is any possible world that is self-defeating, pretty much any other possible world can be sucked into the defeat. This depends a bit on the details about how unlikely reliable beliefs must be for belief in that situation to be self-defeating, and how the space can be broken up. But generally, this reasoning allows the self-defeatingness of any state of affairs to contaminate any other state of affairs that can be placed in disjunction with it, and that can be broken into states of affairs not much more probable than it.

It seems to me that any reasoning with this property must be faulty. So I suggest probabilistic self-defeat arguments of this form can’t work in general.

It could be that Plantinga means to make a stronger argument, for instance ‘there is no set of beliefs consistent with naturalism under which one’s beliefs have high probability’, but this seems like quite a hard argument to make. I could place a high probability on A for instance.

It could also be that Plantinga means to use further assumptions that make a distinction between grouping A and B together and grouping B and C together. One possibility is that it is important that N is a cause of T or F, but this seems both ad-hoc and possible to get around. At any rate, Plantinga doesn’t seem to articulate further assumptions in the account of his argument that I read, so his argument seems unlikely to be correct as it stands, all other criticisms aside.

—

*Note that if you wanted to turn the argument against creationism, it seems you could also just expand the space to include creators who don’t produce true beliefs, and  depending on probabilities, use this to defeat the belief in a creator, including one who does produce true beliefs.

Relatively minor technological change can move the balance of power between values that already fight within each human. Beeminder empowers a person’s explicit, considered values over their visceral urges, in much the same way that the development of better sling shots empowers one tribe over another.

In the arms race of slingshots, the other tribe may soon develop their own weaponry. In the spontaneous urges vs. explicit values conflict though, I think technology should generally tend to push in one direction. I’m not completely sure which direction that is however.

At first glance, it seems to me that explicit values will tend to have a much better weapons research program. This is because they have the ear of explicit reasoning, which is fairly central to conscious research efforts. It seems hard to intentionally optimize something without admitting at some point in the process that you want it.

When I want to better achieve my explicit goal of eating healthy and cheap food for instance, I can sit down and come up with novel ways to achieve this. Sometimes such schemes even involving trickery of the parts of myself that don’t agree with this goal, so divorced they are from this process. When I want to fulfill my urge to eat cookie dough on the other hand, I less commonly deal with this by strategizing to make cookie dough easier to eat in the future, or to trick other parts of myself into thinking eating cookie dough is a prudent plan.

However this is probably at least partly due to the cookie dough eating values being shortsighted. I’m having trouble thinking of longer term values I have that aren’t explicit on which to test this theory, or at least having trouble admitting to them. This is not very surprising; if they are not explicit, presumably I’m either unaware of them or don’t endorse them.

This model in which explicit values win out could be doubted for other reasons. Perhaps it’s pretty easy to determine unconsciously that you want to live in another suburb because someone you like lives there, and then after you have justified it by saying it will be good for your commute, then all the logistics that you need to be conscious for can still be carried out. In this case it’s easy to almost-optimize something consciously without admitting that you want it. Maybe most cases are like this.

Also note that this model seems to be in conflict with the model of human reasoning as basically involving implicit urges followed up by rationalization. And sometimes at least, my explicit reasoning does seem to find innovative ways to fulfill my spontaneous urges. For instance, it suggests that if I do some more work, then I should be able to eat some cookie dough. One might frame this as conscious reasoning merely manipulating laziness and gluttony to get a better deal for my explicit values. But then rationalization would say that. I think this is ambiguous in practice.

Robin Hanson responds to my question by saying there are not even two sets of values here to conflict, but rather one which sometimes pretends to be another. I think it’s not obvious how that is different, if pretending involves a lot of carrying out what an agent with those values would do.

An important consideration is that a lot of innovation is done by people other than those using it. Even if explicit reasoning helps a lot with innovation, other people’s explicit reasoning may side with your inchoate hankerings. So a big question is whether it’s easier to sell weaponry to implicit or explicit values. On this I’m not sure. Self-improvement products seem relatively popular, and to be sold directly to people more often than any kind of products explicitly designed to e.g. weaken willpower. However products that weaken willpower without an explicit mandate are perhaps more common. Also much R&D for helping people reduce their self-control is sponsored by other organizations, e.g. sellers of sugar in various guises, and never actually sold directly to the customer (they just get the sugar).

I’d weakly guess that explicit values will win the war. I expect future people to have better self-control, and do more what they say they want to do. However this is partly because of other distinctions that implicit and explicit values tend to go along with; e.g. farsighted vs. not. It doesn’t seem that implausible that implicit urges really wear the pants in directing innovation.