12 comments, sorted by Click to highlight new comments since: Today at 11:30 AM
New Comment

(Summary: A debate league's yearlong policy debate resolution is about AI; does this seem like a good outreach opportunity?)

"Resolved: The United States Federal Government should substantially reform the use of Artificial Intelligence technology."

IMO, it's not the best of wording, but that's the current team policy debate resolution in the Stoa debate league. For the next ~9 months, a few hundred high school students will be researching and debating over "the use of artificial intelligence technology." In the past, people have posted about competitive debating and its potential relationship with EA; does this at all seem like an opportunity for outreach? (To be fair, Stoa is smaller than traditional public school leagues, but the policy debate norms are way better/less toxic, making team policy one of the most popular events in Stoa)

Great spot. Presumably this means a lot of kids will be googling related terms and looking for pre-existing policy suggestions and pro/con lists.

EA (forum/community) and Kialo?

TL;DR: I’m curious why there is so little mention of Kialo as a potential tool for hashing out disagreements in the EA forum/community, whereas I think it would be at least worth experimenting with. I’m considering writing a post on this topic, but want to get initial thoughts (e.g., have people already considered it and decided it wouldn’t be effective, initial impressions/concerns, better alternatives to Kialo)

The forum and broader EA community has lots of competing ideas and even some direct disagreements. Will Bradshaw's recent comment about discussing cancel culture on the EA forum is just the latest example of this that I’ve seen. I’ve often felt that the use of a platform like Kialo would be a much more efficient way of recording these disagreements, since it helps to separate out individual points of contention and allow for deep back-and-forth, among many other reasons. However, when I search for “Kialo” in the search bar on the forum, I only find a few minor comments mentioning it (as opposed to posts) and they are all at least 2 years old. I think I once saw a LessWrong post downplaying the platform, but I was wondering if people here have developed similar impressions.

More to the point, I was curious to see if anyone had any initial thoughts on whether it would be worthwhile to write an article introducing Kialo and highlighting how it could be used to help hash out disagreements here/in the community? If so, do you have any initial objections/concerns that I should address? Do you know of any other alternatives that would be better options (keeping in mind that one of the major benefits of Kialo is its accessibility)?

How would you feel about reposting this in EAs for Political Tolerance (https://www.facebook.com/groups/159388659401670) ? I'd also be happy to repost it for you if you'd prefer.

Do you just mean this shortform or do you mean the full post once I finish it? Either way I’d say feel free to post it! I’d love to get feedback on the idea

A few months ago I wrote a post on a decision-analysis framework (the stock issues framework) that I adapted from a framework which is very popular/prominent in competitive high school policy debate (which uses the same name). I was surprised to not receive any feedback/comments (I was at least expecting some criticism, confusion, etc.), but in retrospect I realized that it was probably a rather lengthy/inefficient post. I also realized that I probably should have written a shortform post to get a sense of interest, some preliminary thoughts on the validity and novelty/neglectedness of the concept, and how/where people might misinterpret or challenge the concept (or otherwise want to see more clarity/justification). So, I’ll try to offer a simplified summary here in hopes to get some more insight on some of those things I mentioned (e.g., the potential value, novelty/neglectedness, validity, areas of confusion/skepticism).

The framework remarkably echoes the “importance, neglectedness, tractability” (INT) heuristic for cause area prioritization, except that the stock issues framework is specific to individual decisions and avoids some of the problems of the INT heuristic (e.g., the overgeneralized assumption of diminishing marginal returns). Basically, the stock issues framework holds that every advantage and disadvantage (“pro and con”) of a decision rests on four mutually exclusive and exhaustive concepts: inherency (which is reminiscent of “neglectedness,” but is more just “the descriptive state of affairs”), significance, feasibility, and solvency. (I explain them in more detail in my post.) 

Over time, I have informally thought of and jotted down some of the potential justifications for promoting this framework (e.g., checking against confirmation and other biases, providing common language and concept awareness in discourse, constructing concept categories so as to improve learning and application of lessons from similar cases). However, before I write a post about such justifications, I figured I would write this shortform to get some preliminary feedback, as I mentioned: I’d love to hear where you are skeptical, confused, interested, etc.! (Also, if you think the original post I made should/could be improved--such as by reducing caveats/parentheticals/specificities, making some explanation more clear, etc.--feel free to let me know!)

I really appreciate your constructive attitude here :) I write below some recommendations and my take on why this wasn't successful. Some of it is a bit harsh, but that's because I honestly respect you and think you'll take it well 😊

I remember coming across your post, which is in an area that I'm very interested in, but seeing that I didn't remember any details and didn't upvote, I probably just skimmed it and didn't find it worth my time to read. I've read it now, and I have some thoughts about how you could have written a post on this topic which I  would find interesting and more readable - after reading it now, I think that it has some useful content that I'd like to know.

  1. A lot of the post (and actually even most of this shortform post) is about your own views and thinking process and meta-thoughts about the post itself and it's context. This is a lot of overhead which is not needed and in fact damaging both because it is distracting and because it makes it harder to find the gold within.
  2. As you said, the post is too lengthy and inefficient. I'd guess that most readers of the forum go through posts by filtering in approximately this order: Title-> skimming first paragraphs /  look for clear bullet points or tl;dr-> skimming the post, mostly looking at headers, images, first words of paragraphs, bolded parts, bullet points -> skimming sections of interest -> dive deeper into all or what interests them. 
    1. I found myself confused from skimming the intro. I saw that you offer an alternative to ITN, but didn't understand what it is.
    2. Skimming the rest of the post, I saw the four bulleted concepts and my next thought was that I get the general idea of the post, even if I'm confused about somethings, but it's not worth my time to read through this text to understand it better.
  3. It feels that the post is aiming at persuasion rather than description. I got the feeling that I was being sold some new shiny framework, and that most of the effort in the post goes there instead of just explaining what it's all about. I really do think that you overpromise here, and by doing that I could easily discard the whole idea as not worthwhile even if it has some merit.
  4. Relatedly, I found the attitude in the post somewhat vain and dismissive towards existing ideas and the readers. As I write this, I look back and didn't find any clear examples of that so perhaps I'm misjudging the post here. Perhaps it's because you make it seem like it's your idea.
  5. Key ideas of the framework are not explained properly. I don't understand how exactly one uses this framework. Can you put a "number" or evaluation on inherency? How exactly do accounts of diminishing returns enter this framework? What do we do about some overlap between different parts? You write that you hope for people to comment and ask questions, but I think that this is too much to ask - it takes a while to clarify to myself what I don't understand, and it's a lot of overhead anyway.

What I'd really hope you will do is to write a short post (not a shortform) which only explains this framework and some of its features, without unneeded meta-discussion.  I've tried skimming the Wikipedia page, but it's in a different enough context and language that it's difficult for me to understand without a lot of effort.

Thanks for the insight/feedback! I definitely see what you are saying on a lot of points. I’ll be working on an improved post soon that incorporates your feedback.

Working title: Collaborative Discussion Spaces and "Epistemic Jam Sessions" for Community Building Claims/Ideas?

Tl;dr: I created an example discussion space on Kialo for claims/ideas about EA community building, with the idea being that community builders could collaborate via such structured discussions. Does this seem like something that could be valuable? Is it worth making this shortform into a full post?

 

I’m a big fan of structured discussions, and while reading this post early last month I wondered: would it be helpful if there were some kind of virtual space for sharing claims and ideas—and arguments for/against those claims and ideas—about community organizing? 

For example, 

  • “Intro groups/fellowships at universities should not have "required readings"
  • “University organizers should prioritize creating fellowships that require applications over open-to-anyone reading groups and other ad-hoc events.”

I went ahead and created a toy example of what such a structured discussion could look like on Kialo, with some example claims and arguments for/against those claims.

Building on this, I also wondered if it might be good to designate some 1–3 day period each month as a focal/Schelling point for community organizer participation, perhaps also with some non-binding/optional goals laid out in advance (e.g., “we want to get a better sense of how to improve outreach/success at lower-prestige universities,” “we want to hear about your experiences/advice regarding outreach to STEM groups”). Perhaps you could call these “epistemic jam sessions” (I’m totally open to accepting better name ideas). Regardless, these discussions could be open to contributions at any time.

 

Ultimately, I’d love to hear your thoughts on:

  1. Whether anything else like this already exists;
  2. Whether something like this seems valuable;
  3. Any recommendations for alternative characteristics/designs;
  4. Whether I should make this into a full post.

I'm considering doing another pilot "epistemic map", but I'm trying to decide what topic I should do it on, and thus soliciting suggestions. 

For more on epistemic mapping, you can see here for a presentation I recently gave on the topic (just ignore the technical issues in the beginning)

Whereas my last pilot/test map focused on the relationship between poverty and terrorism (and the associated literature), I want to do this one on something EA-relevant. FWIW, I think that epistemic mapping is probably most valuable for topics that are important, dynamic (e.g., assumptions or technological capabilities may change over time), unsettled/divisive, and/or have a non-small literature base (among a few other considerations).

Some of my ideas thus far have been the controversial Democratising Risk paper (or something else X-risk related), the Worm Wars debate, biosecurity/pandemic risks, or maybe something about AI. But I'd love to hear any other suggestions (or feedback on those ideas I listed)!

[Summary: Most people would probably agree that science benefited greatly from the shift to structured, rigorous empirical analyses over the past century, but some fields still struggle to make progress.  I’m curious whether people think that we could/should seek to introduce more structure/sophistication to the way researchers make and engage with theoretical analyses, such as something like "epistemic mapping"]

I just discovered this post, and I was struck by how it echoed some of my independent thoughts and impressions, especially the quote: "But it should temper our enthusiasm about how many insights we can glean by getting some data and doing something sciency to it."

(What follows is shortform-level caveating and overcomplicating, which is to say, less than I normally would provide, and more about conveying the overall idea/impression)

I've had some (perhaps hedgehoggy) "big ideas" about the potential value of what I call "epistemic mapping" for advancing scientific study/inquiry/debate in a variety of fields. One of them relates to the quote above: the "empirical-scientific revolution" of the past ~100-200 years (e.g., the shift to measuring medical treatment effectiveness through inpatient/outpatient data rather than professionals’ impressions) seems to have been crucial in the advancement of a variety of fields. 

However, there are still many fields where such empirical/data-heavy methods appear insufficient and where it seems like progress languishes: my impression has been that this especially includes many of the social sciences (e.g., conflict studies, political science, sociology). There are no doubt many  possible explanations, but over time I've increasingly wondered whether a major set of problems is loosely that the overall complexity of the systems (e.g., human decision making process vs. gravitational constants) + the difficulty of collecting sufficient data for empirical analyses +  (a few other factors) leads to a situation of high information lossage between researchers/studies and/or people are incentivized to oversimplify things (e.g., following the elsewhere-effective pattern of regression analyses and p<0.05 = paper). I do not know, but if the answer is yes, that leads to a major question:

How could/should we attempt to solve or mitigate this problem? One of the (hedgehoggy?) questions that keeps bugging me: We have made enormous advances in the past few hundred years when it comes to empirical analyses; in comparison, it seems that we have only fractionally improved the way we do our theoretical analysis... could/should we be doing better? [Very interested to get people's thoughts about that overall characterization, which even I'll admit I'm uncertain about] 

So, I'm curious if people share similar sentiment about our ability/need to improve our methods of theoretical analysis, including how people engage with the broader literature aside from the traditional (and, IMO, inefficient) paragraph-based literature reviews. If people do share similar sentiment, what do you think about that concept of epistemic mapping as a potential way of advancing some sciences forward? Could it be the key to efficient future progress in some fields? My base rates for such a claim are really low, and I recognize that I'm biased, but I feel like it's worth posing the question if only to see if it advances the conversation. 

(I might make this into an official post if people display enough interest)