That’s true. But I wonder why you’re “worried” or “feel deeply uncomfortable” about “someone who is not aware of what they’re doing, but unthinkingly goes down the ladder”. To me, that seems like a pretty harmless and uncommon mistake. So I’d still appreciate if you could give one example of how that’s plausibly harmful.
Also, asking for examples of other times when you had headaches probably brings out more specifics and does meaningfully go down the ladder of abstraction.
The failure mode happens quite often to me. People go down the ladder of abstraction often miss the point of staying at the higher level, derailing the conversation. The TRIZ prism is instructive here in why this is bad, and this failure mode of asking for specific solutions when explorong meta solutions happens quite a bit. The other example of a person asking for definitions is another example already given.
I’ll note that this is a “Rock and hard place” situation, but so far O haven’t seen you acknowledge that there is in fact a hard place.
The other example of a person asking for definitions is another example already given.
I explained how your example of the fellow asking for definitions actually wasn’t an example of someone going down the ladder of abstraction. So from my perspective, you still haven’t given a single example of this thing that supposedly “happens quite often”.
I want to take a step back and go up the ladder of abstraction, as I feel we’re a bit too “in the weeds” at the moment.
Here’s my ITT of you, let me know if this feels accurate:
Most conversations have a chronic problem of “too much abstraction”. Often times, if people would just be specific about they’re claims, they would realize that they’re just too vague, or they don’t have a real plan, or there’s no real disagreement, or their grievances are false.
If we could start being aware when we’re too abstract in our thinking, and develop the habit of constantly being more specific, we could improve dialogue. Furthermore, this problem is SOO widespread, that even if people didn’t understand that they were doing, just developing the habit of being more specific without any foundation would be an improvement on the status quo.
It’s pretty common for someone to derail a discussion by going down a rabbit hole such as asking everyone to precisely define all terms, or asking for a lot of detail and follow-up questions about some small part of the larger discussion.
Oftentimes, a good move in a discussion is to take it up the ladder of abstraction and refer back to the high-level goals of the discussion.
Great! The one part that’s missing is another big goal of abstraction that Ive mentioned, which is to allow abstract framings of a problem that suggest solutions.
Cool. So my crux is: Going down a rabbit hole is its own failure mode, it’s not the same thing as going down the ladder of abstraction. If people listen to my advice to be more specific, I’m not worried about unleashing more rabbit-holers.
That’s why I was wondering if you have other examples where specificity is harmful besides the known failure mode of people going down a rabbit hole.
I found a few examples on LW, but don’t really want to call out anyone specifically. Here’s another example:
At the EA hotel, talking about specific problems/demographics of the EA community, we were talking in general terms, trying to describe a particular demographic. At some point someone was like “hold on what are we discussing here, can we explain specifically what we would expect to see?”
This was actually a good move, but it was too early. Specifically, we were still fruitfully changing and exploring our model, finding the most useful way to actually think about this particular dynamic in the EA community. By forcing us to go specific at this time, we ended up “locked in” to the specific frame we were at when we went more specific, and it took about 20 minutes of conversational maneuvering to get back to the “exploration” phase. At that point, we eventually did settle on a better frame that seemed useful to everyone, and let the conversation naturally lead to examples and specificity.
I recognize that both of the above examples are a bit general, I’m having trouble finding specific examples that aren’t a bit controversial or would point too much towards blaming a specific person if they saw me write.
So the push for specificity helped clarify people’s thinking, but the discussion got derailed because no one said “let’s consider another possibility for the claim we want to make”.
I don’t think that means anyone was failing by being too specific. There seems to be a separate kind of failure mode in the domain of exloratory-discussion steering efficiency.
But I’m happy with the quality of the examples you’re providing to facilitate our discussion.
In One’s Own Thinking: When you have any kind of thought or belief, try making it specific when you get a chance. You’ll probably get some value out of the exercise. You neglect the exercise at your peril.
In Discourse: If you’re putting forth a claim, then it’s worth trying to provide specifics for it. If you’re not yet making a claim, but more like exploring ideas, then specifics are not yet mandatory.
I think there’s a similar danger in trying to be too specific in ones’ own thinking. I can’t quite articulate it yet, but the idea of “holding a question” in this article feels internally to me like a very different stance then one where I’m requesting specificity from myself, and I find it highly valuable.
There’s something about the move of specificity that doesn’t allow for “space”, which occurs both internally and in conversation.
Yes, and them right after that you agreed that in other circumstances, asking for the definition is the right move for going down the ladder of abstraction… So, this was an example for someone applying specificity unthinkingly instead of understanding the ladder of abstraction and when it’s useful.
Another example is at a conference, discussing different types of forecasting, being derailed by someone asking how the specific algorithms we were talking about would be used, not realizing that we hadn’t explored the abstract solution space of forecasting enough yet to answer that, and that it would derail the conversation.
I can keep giving examples like this btw, and we can go through them one by one, or we can use abstraction and talk about them as a group phenomena.
Another example is at a conference, discussing different types of forecasting, being derailed by someone asking how the specific algorithms we were talking about would be used, not realizing that we hadn’t explored the abstract solution space of forecasting enough yet to answer that, and that it would derail the conversation.
I’d be interested to understand the discussion in more detail and whether over-zealous specificity is really a cause of harm here. Because it seems to me like spending a minute to get a sense of how specifically a certain algorithm might be used could add value to the discussion.
I can keep giving examples like this btw, and we can go through them one by one, or we can use abstraction and talk about them as a group phenomena.
I consider the above example to be the only valid example you’ve given, and it’s still an unclear one to me, so I think the best next step is either to clarify and improve on this one example or to provide another better example.
That’s true. But I wonder why you’re “worried” or “feel deeply uncomfortable” about “someone who is not aware of what they’re doing, but unthinkingly goes down the ladder”. To me, that seems like a pretty harmless and uncommon mistake. So I’d still appreciate if you could give one example of how that’s plausibly harmful.
Also, asking for examples of other times when you had headaches probably brings out more specifics and does meaningfully go down the ladder of abstraction.
The failure mode happens quite often to me. People go down the ladder of abstraction often miss the point of staying at the higher level, derailing the conversation. The TRIZ prism is instructive here in why this is bad, and this failure mode of asking for specific solutions when explorong meta solutions happens quite a bit. The other example of a person asking for definitions is another example already given.
I’ll note that this is a “Rock and hard place” situation, but so far O haven’t seen you acknowledge that there is in fact a hard place.
What’s one specific example???
I explained how your example of the fellow asking for definitions actually wasn’t an example of someone going down the ladder of abstraction. So from my perspective, you still haven’t given a single example of this thing that supposedly “happens quite often”.
I want to take a step back and go up the ladder of abstraction, as I feel we’re a bit too “in the weeds” at the moment.
Here’s my ITT of you, let me know if this feels accurate:
Most conversations have a chronic problem of “too much abstraction”. Often times, if people would just be specific about they’re claims, they would realize that they’re just too vague, or they don’t have a real plan, or there’s no real disagreement, or their grievances are false.
If we could start being aware when we’re too abstract in our thinking, and develop the habit of constantly being more specific, we could improve dialogue. Furthermore, this problem is SOO widespread, that even if people didn’t understand that they were doing, just developing the habit of being more specific without any foundation would be an improvement on the status quo.
------
Does that land for you?
Yep!
Would you be willing to ITT me? If not, I can try to write up a similar high level summary of my position.
Sure.
How’s that?
Great! The one part that’s missing is another big goal of abstraction that Ive mentioned, which is to allow abstract framings of a problem that suggest solutions.
Cool. So my crux is: Going down a rabbit hole is its own failure mode, it’s not the same thing as going down the ladder of abstraction. If people listen to my advice to be more specific, I’m not worried about unleashing more rabbit-holers.
That’s why I was wondering if you have other examples where specificity is harmful besides the known failure mode of people going down a rabbit hole.
I found a few examples on LW, but don’t really want to call out anyone specifically. Here’s another example:
At the EA hotel, talking about specific problems/demographics of the EA community, we were talking in general terms, trying to describe a particular demographic. At some point someone was like “hold on what are we discussing here, can we explain specifically what we would expect to see?”
This was actually a good move, but it was too early. Specifically, we were still fruitfully changing and exploring our model, finding the most useful way to actually think about this particular dynamic in the EA community. By forcing us to go specific at this time, we ended up “locked in” to the specific frame we were at when we went more specific, and it took about 20 minutes of conversational maneuvering to get back to the “exploration” phase. At that point, we eventually did settle on a better frame that seemed useful to everyone, and let the conversation naturally lead to examples and specificity.
I recognize that both of the above examples are a bit general, I’m having trouble finding specific examples that aren’t a bit controversial or would point too much towards blaming a specific person if they saw me write.
So the push for specificity helped clarify people’s thinking, but the discussion got derailed because no one said “let’s consider another possibility for the claim we want to make”.
I don’t think that means anyone was failing by being too specific. There seems to be a separate kind of failure mode in the domain of exloratory-discussion steering efficiency.
But I’m happy with the quality of the examples you’re providing to facilitate our discussion.
Sort of, it helped solidify a not yet solidified frame, which was a waste of time, because the frame was rapidly changing.
I said this, but the person I was talking to had a strong aesthetic need for specificity and wouldn’t let it go.\
I think there was failingin asking for specificity at the wrong time.
Maybe we can agree to say this:
In One’s Own Thinking: When you have any kind of thought or belief, try making it specific when you get a chance. You’ll probably get some value out of the exercise. You neglect the exercise at your peril.
In Discourse: If you’re putting forth a claim, then it’s worth trying to provide specifics for it. If you’re not yet making a claim, but more like exploring ideas, then specifics are not yet mandatory.
I think there’s a similar danger in trying to be too specific in ones’ own thinking. I can’t quite articulate it yet, but the idea of “holding a question” in this article feels internally to me like a very different stance then one where I’m requesting specificity from myself, and I find it highly valuable.
There’s something about the move of specificity that doesn’t allow for “space”, which occurs both internally and in conversation.
Just wanted to say I appreciated this exchange in both directions.
Yes, and them right after that you agreed that in other circumstances, asking for the definition is the right move for going down the ladder of abstraction… So, this was an example for someone applying specificity unthinkingly instead of understanding the ladder of abstraction and when it’s useful.
Another example is at a conference, discussing different types of forecasting, being derailed by someone asking how the specific algorithms we were talking about would be used, not realizing that we hadn’t explored the abstract solution space of forecasting enough yet to answer that, and that it would derail the conversation.
I can keep giving examples like this btw, and we can go through them one by one, or we can use abstraction and talk about them as a group phenomena.
I’d be interested to understand the discussion in more detail and whether over-zealous specificity is really a cause of harm here. Because it seems to me like spending a minute to get a sense of how specifically a certain algorithm might be used could add value to the discussion.
I consider the above example to be the only valid example you’ve given, and it’s still an unclear one to me, so I think the best next step is either to clarify and improve on this one example or to provide another better example.