I quit the Effective Altruism forum due to a new rule requiring all new posts and comments be basically put in the public domain without copyright, so anyone could e.g. sell a book of my posts without my consent (they’d just have to give attribution). More info. I had a bunch of draft posts, so I’m posting some of them here with minimal editing. In general, I’m not going to submit them as link posts at EA myself. If you think they should be shared with EA as link posts, please do it yourself. I’m happy for other people to share links to my work at EA or on social media. Please share stuff in whatever ways you think are good to do.
Suppose EA is making an important error. John knows a correction and would like to help. What can John do?
Whatever the answer is, this is something EA should put thought into. They should advertise/communicate the best process for John to use, make it easy to understand and use, and intentionally design it with some beneficial features. EA should also consider having several processes so there are backups in case one fails.
Failure is a realistic possibility here. John might try to share a correction but be ignored. People might think John is wrong even though he’s right. People might think John’s comment is unimportant even though it’s actually important. There are lots of ways for people to reject or ignore a good idea. Suppose that happens. Now EA has made two mistakes which John knows are mistakes and would like to correct. There’s the first mistake, whatever it was, and now also this second mistake of not being receptive to the correction of the first mistake.
How can John get the second mistake corrected? There should be some kind of escalation process for when the initial mistake correction process fails. There is a risk that this escalation process would be abused. What if John thinks he’s right but actually he’s wrong? If the escalation process is costly in time and effort for EA people, and is used frequently, that would be bad. So the process should exist but should be designed in some kind of conservative way that limits the effort it will cost EA to deal with incorrect corrections. Similarly, the initial process for correcting EA also needs to be designed to limit the burden it places on EA. Limiting the burden increases the failure rate, making a secondary (and perhaps tertiary) error correction option more important to have.
When John believes he has an important correction for EA, and he shares it, and EA initially disagrees, that is a symmetric situation. Each side thinks the other is wrong. (That EA is multiple people, and John also might actually be multiple people, makes things more complex, but without changing some of the key principles.) The rational thing to do with this kind of symmetric dispute is not to say “I think I’m right” and ignore the other side. If you can’t resolve the dispute – if your knowledge is inadequate to conclude that you’re right – then you should be neutral and act accordingly. Or you might think you have crushing arguments which are objectively adequate to resolve the dispute in your favor, and you might even post them publicly, and think John is responding in obviously unreasonable ways. In that case, you might manage to objectively establish some kind of asymmetry. How to do objectively establish asymmetries in intellectual disagreements is a hard, important question in epistemology which I don’t think has received appropriate research attention (note: it’s also relevant when there’s a disagreement between two ideas within one person).
Anyway, what can John do? He can write down some criticism and post it on the EA forum. EA has a free, public forum. That is better than many other organizations which don’t facilitate publicly sharing criticism. Many organizations either have no forum or delete critical discussions while making no real attempt at rationality (e.g. Blizzard has forums related to its games, but they aren’t very rational, don’t really try to be, and delete tons of complaints). Does EA ever delete dissent or ban dissenters? As someone who hasn’t already spent many years paying close attention, I don’t know and I don’t know how to find out in a way that I would trust. Many forums claim not to delete dissent but actually do; it’s a common thing to lie about. Making a highly credible claim not to delete or punish dissent is important or else John might not bother trying to share his criticism.
So John can post a criticism on a forum, and then people may or may not read it and may or may not reply. Will anyone with some kind of leadership role at EA read it? Maybe not. This is bad. The naive alternative “guarantee plenty of attention from important people to all criticism” would be even worse. But there are many other possible policy options which are better.
To design a better system, we should consider what might go wrong. How could John’s great, valuable criticism receive a negative reaction on an open forum which is active enough that John gets at least a little attention? And how might things go well? If the initial attention John gets is positive, that will draw some additional attention. If that is positive too, then it will draw more attention. If 100% of the attention John gets results in positive responses, his post will be shared and spread until a large portion of the community sees it including people with power and influence, who will also view the criticism positively (by premise) and so they’ll listen and act. A 75% positive response rate would probably also be good enough to get a similar outcome.
So how might John’s criticism, which we’re hypothetically supposing is true and important, get a more negative reception so that it can’t snowball to get more attention and influence important decision makers?
John might have low social status, and people might judge more based on status than idea quality.
John’s criticism might offend people.
John’s criticism might threaten people in some way, e.g. implying that some of them shouldn’t have the income and prestige (or merely self-esteem) that they currently enjoy.
John’s criticism might be hard to understand. People might get confused. People might lack some prerequisite knowledge and skills needed to engage with it well.
John’s criticism might be very long and hard to get value from just the beginning. People might skim but not see the value that they would see if they read the whole thing in a thoughtful, attentive way. Making it long might be an error by John, but it also might be really hard to shorten and still have a good cost/benefit ratio (it’s valuable enough to justify the length).
John’s criticism might rely on premises that people disagree with. In other words, EA might be wrong about more than one thing. An interconnected set of mistakes can be much harder to explain than a single mistake even if the critic understands the entire set of mistakes. People might reject criticism of X due to their own mistake Y, and criticism of Y due to their own mistake X. A similar thing can happening involving many more ideas in a much more complicated structure so that it’s harder for John to point out what’s going on (even if he knows).
What can be done about all these difficulties? My suggestion, in short, is to develop a rational debate methodology and to hold debates aimed at reaching conclusions about disagreements. The methodology must include features for reducing the role of bias, social status, dishonesty, etc. In particular, it must prevent people from arbitrarily stopping any debates whenever they feel like it (which tends to include shortly before losing, which prevents the debate from being conclusive). The debate methodology must also have features for reducing the cost of debate, and ending low value debates, especially since it won’t allow arbitrarily quitting at any moment. A debate methodology is not a perfect, complete solution to all the problems John may face but it has various merits.
People often assume that rational, conclusive debate is too much work so the cost/benefit ratio on it is poor. This is typically a general opinion they have rather than an evaluation of any specific debate methodology. I think they should reserve judgment until after they review some written debate methodologies. They should look at some actual methods and see how much work they are, and what benefits they offer, before reaching a conclusion about their cost/benefit ratio. If the cost/benefit ratios are poor, people would try to make adjustments to reduce costs and increase benefits before giving up on rational debate.
Can people have rational debate without following any written methodology? Sure that’s possible. But if that worked well for some people and resulted in good cost/benefit ratios, wouldn’t it make sense to take whatever those successful debate participants are doing and write it down as a method? Even if the method had vague parts that’d be better than nothing.
Although under-explored, debate methodologies are not a new idea. E.g. Russell L. Ackoff published one in a book in 1978 (pp. 44-47). That’s unfortunately the only very substantive, promising one I’ve found besides developing one of my own. I bet there are more to be found somewhere in existing literature though. The main reason I thought Ackoff’s was a valuable proposal were that 1) it was based on following specific steps (in other words, you could make a flowchart out of it); and 2) it aimed at completeness, including using recursion to enable it to always succeed instead of getting stuck. Partial methods are common and easy to find, e.g. “don’t straw man” is a partial debate method, but it’s just suitable for being one little part of an overall method (and it lacks specific methods of detecting straw men, handling them when someone thinks one was done, etc. – it’s more of an aspiration than specific actions to achieve that aspiration).
A downside of Ackoff’s method is that it lacks stopping conditions besides success, so it could take an unlimited amount of effort. I think unilateral stopping conditions are one of the key issues for a good debate method: they need to exist (to prevent abuse by unreasonable debate partners who don’t agree to end the debate) but be designed to prevent abuse (by e.g. people quitting debates when they’re losing and quitting in a way designed to obscure what happened). I developed impasse chains as a debate stopping condition which takes a fairly small, bounded amount of effort to end debates unilaterally but adds significant transparency about how and why the debate is ending. Impasse chains only work when the further debate is providing low value, but that’s the only problematic case – otherwise you can either continue or say you want to stop and give a reason (which the other person will consent to, or if they don’t and you think they’re being unreasonable, now you’ve got an impasse to raise). Impasse chains are in the ballpark of “to end a debate, you must either mutually agree or else go through some required post-mortem steps” plus they enable multiple chances at problem solving to fix whatever is broken about the debate. This strikes me as one of the most obvious genres of debate stopping conditions to try, yet I think my proposal is novel. I think that says something really important about the world and its hostility to rational debate methodology. (I don’t think it’s mere disinterest or ignorance; if it were, the moment I suggested rational debate methods and said why they were important a lot of people would become excited and want to pursue the matter; but that hasn’t happened.)
Another important and related issue is how can you write, or design and organize a community or movement, so it’s easier for people to learn and debate with your ideas? And also easier to avoid low value or repetitive discussion. An example design is an FAQ to help reduce repetition. A less typical design would be creating (and sharing and keeping updated) a debate tree document organizing and summarizing the key arguments in the entire field you care about.
Messages