Weighted Pros / Cons as a Norm
Takeaway
I think using weighted pros / cons (or more generally, arguments for / against) would be a useful norm to promote. For a summary of the reasons why, see the Example section.
Motivation
Though maybe not an explicit norm, many people in EA endorse the idea of putting probabilities to statements in order to clarify one’s credence in them. Doing so allows people to be much more precise and avoid the ambiguity of phrases like “almost certain” or “significant chance.” It’s also helpful for discussion as it can make it clearer how and to what degree people agree or disagree. It seems that many EA community members generally value “putting numbers to things.” As an extension of this, I think it would be helpful for more people to weight their pros / cons or arguments for / against when discussing a given topic. I don’t think this is currently done often and can’t recall a time when I’ve seen it done firsthand.
I started thinking about this after reading the answers on my question about AIS harm. When reading the responses, I found myself unsure how much weight the author would give some of their considerations. This made it more difficult to determine how I should update some of my thinking on the subject. Further, it made additional discussion more difficult because I wasn’t sure how much I agreed or disagreed with each response (I also unfortunately haven’t had as much time to respond as I would’ve liked). I think that not having weights makes it much more difficult to digest arguments for / against something since the reader is left in the dark as to how each consideration should stack up against the others. I also think there are benefits for the writer, which I touch on in the example below. I’m not at all trying to pick on the answers given at that post (I very much appreciated them!). From what I’ve seen, this is how most answers on most forums are given. I’m also not claiming that weights should be added in all situations, but I do think they are often helpful, especially in cases where one is explicitly listing arguments for or against something. Like quantifying our credences with probabilities, this may be another useful norm to promote. As an example, below is a weighted pros / cons list of using weighted pros / cons (or putting weights to arguments for / against, more generally).
Method
This can be done in a variety of ways, but I generally use a 1-5 scale. I typically don’t feel I need higher resolution than this. 1 is something like “not at all important,” 2 is “unimportant,” 3 is in the middle, which could mean “just as important as unimportant,” 4 is “important,” 5 is “extremely important.” Also, if one is pressed for time or the context is quicker, more informal, and doesn’t need as much precision, you could just use qualitative weights, like “important,” “very important,” “not important,” etc. I’m very open to other methods of doing this as well.
Example
Scale: 1-5 as described above, indicated in bold parentheses.
R: reader
W: writer
Note: these weights are put down lightly as I don’t have much direct experience
comparing situations when they are used to ones when they aren’t. I’ve mostly
just seen cases when they aren’t used and what the effects of that are.
Putting Weights on Arguments For / Against the Topic at Hand
Pros
From R’s perspective: clarifies their understanding of W’s position on each of the considerations. (5)
Allows for better discussion for the following reasons (5):
R can better tell if they agree / disagree with the weights put on each point, which will very likely affect the ultimate conclusion drawn from the points. If they disagree with the weights, they may very well disagree with the conclusion.
if R agrees / disagrees with one or more of the weights, R can immediately recognize this as the source of agreement / disagreement. This can otherwise be difficult to discern when faced with a large, complex set of considerations and a general feeling of agreement / disagreement.
both of the above points allow R to better understand W’s writing and put R in a better position to respond in a meaningful way.
From W’s perspective: clarifies their understanding of their position on each of the considerations. I didn’t list this as a 5 since I would guess that W has a more solid idea of what W’s weights would be than R does. (4)
Allows one to sum the weights of the “con / against” side and subtract this from the sum of the “pros / for” side, yielding a single numerical representation of how the points stack up against each other. I certainly don’t think this should be the final say on the issue, but it can be a useful input when exploring the topic at hand. (4)
Cons
Could mislead some people into taking the sum mentioned above as being ultimately decisive when it should just be used as another input (we might treat the weight of this input as corresponding to its absolute value. That is, the farther the value is from zero, the more confident it is in the direction of for / against). This doesn’t seem highly worrisome to me since I don’t think most people’s intuition tells them to blindly trust numbers without any further thought. Moreover, I think that many have probably seen 80K Hour’s career decision-making processes, which echo this advice to take the number as another input. (4)
Takes additional time to add weights. The time burden can be at least somewhat mitigated by using more broad qualitative weights if necessary, which is why I think this consideration isn’t as important. (3)
Output
(5 + 5 + 4 + 4) - (4 + 3) = 11
Note: this output number should not be taken very seriously since it’s not clear in my scale that a 1 is really half the value of a 2, for example. Instead, it could possibly be used as a very rough indication of how strong your credence is in a given direction, but it doesn’t seem like a good idea to read much into the exact number value. For better quantitative approaches, see the comments below.
Nice post! I like the general idea and agree that a norm like this could aid discussions and clarify reasoning. I have some thoughts that I hope can build on this.
I worry that the (1-5) scale might be too simple or misleading in many cases though and it doesn’t quite give us the most useful information. My first concern is that this looks like a cardinal scale (especially the way you calculate the output) but is it really the case that you should weigh arguments with score 2 twice as much as arguments with score 1 etc.? Some arguments might be much more than 5x more important than others, but that can’t be captured on the (1-5) scale.
Maybe this would work better as an ordinal ranking with 5 degrees of importance (the initial description sounds more like this). In the example, this would be sufficient to establish that the pros have more weight, but it wouldn’t always be conclusive (e.g. 5, 1 on the pro side and 4, 3 on the con side).
I think a natural cardinal alternative would be to give the Bayes’ factor for each alternative, and ideally give a prior probability at the start. Or similarly, give a prior and then update this after each argument/consideration, so you and the reader can see how much each argument/consideration affects your beliefs. I’ve seen this used before and found it helpful. And this seems to convey more important information than how important an argument/consideration is: how much we update our beliefs in response to arguments/considerations.
Great point! I understand the high-level idea behind priors and updating, but I’m not very familiar with the details of Bayes factors and other Bayesian topics. A quick look at Wikipedia didn’t feel super helpful… I’m guessing you don’t mean formally applying the equations, but instead doing it in a more approximate or practical way? I’ve heard Spencer Greenberg’s description of the “Question of Evidence” (how likely would I be to see this evidence if my hypothesis is true, compared to if it’s false?). Are there similar quick, practical framings that could be applied for the purposes described in your comment? Do you know of any good, practical resources on Bayesian topics that would be sufficient for what you described?
Good questions! It’s a shame I don’t have good answers. I remember finding Spencer Greenberg’s framing helpful too but I’m not familiar with other useful practical framings, I’m afraid.
I suggested the Bayes’ factor because it seems like a natural choice of the strength/weight of an argument but I don’t find it super easy to reason about usually.
The final suggestion I made will often be easier to do intuitively. You can just to state your prior at the start and then intuitively update it after each argument/consideration, without any maths. I think this is something that you get a bit of a feel for with practice. I would guess that this would usually be better than trying to formally apply Bayes’ rule. (You could then work out your Bayes’ factor as it’s just a function of your prior and posterior but that doesn’t seem especially useful at this point/it seems like too much effort for informal discussions.)
Is there any chance you have an example of your last suggestion in practice (stating a prior, then intuitively updating it after each consideration)? No worries if not.
Sorry for the slow reply. I don’t have a link to any examples I’m afraid but I just mean something like this:
This is just an example I wrote down quickly, not actual views. But the idea is to state explicit probabilities so that we can see how they change with each consideration.
To see you can find the Bayes’ factors, note that if P(W) is our prior probability that we should give weights, P(¬W)=1−P(W) is our prior that we shouldn’t, and P(W|A1) and P(¬W|A1)=1−P(W|A1) are the posteriors after argument 1, then the Bayes’ factor is
P(A1|W)P(A1|¬W)=P(W|A1)P(¬W|A1)P(¬W)P(W)=P(W|A1)1−P(W|A1)1−P(W)P(W)=0.650.350.40.6≈1.24Similarly, the Bayes’ factor for the second pro is 0.750.250.350.65≈1.62.
Sorry for my very slow response!
Thanks—this is helpful! Also, I want to note for anyone else looking for the kind of source I mentioned, this 80K podcast with Spencer Greenberg is actually very helpful and relevant for the things described above. They even work through some examples together.
(I had heard about the “Question of Evidence,” which I described above, from looking at a snippet of the podcast’s transcript, but hadn’t actually listened to the whole thing. Doing a full listen felt very worth it for the kind of info mentioned above.)