I don’t know why in-person group is important, but it seems to be—all the people who have replied to me so far saying they get useful rational practice out of the LW community said the growth came through attending local meetups (example). We can easily invent some evolutionary psychology story for this, but it doesn’t matter why, at this point it’s enough to just know.
I think a neat thing to try would be for people to reply to this comment with suggestions and links for things to try at the SoCal meetup even if you’re not coming, suggestions appreciated.
Put suggested meetup activities in their own comments like a quotes thread (one suggestion per comment, minimize kibitzing, etc), and vote up or down the ones you think would be particularly high (or low) value.
Before going to the meetup, I will take responsibility for printing out some of the suggestions that have more votes (with ancillary text from links as appropriate) and bring them to the meetup in the hopes that we can try implementing one or more suggestions and then report back with a “background/method/results” format in followup articles. Part of the test would be object level (does a particular suggestion have positive effects) and part of the test would be meta level (can we honestly use LW’s existing infrastructure and community to improve community practices).
I will make a couple suggestions by way of example :-)
It is probably hardest to think rationally about high-impact, emotional subjects, and also high value to think rationally about such subjects.
Accordingly:
Have everyone who is willing to share name a major mistake they have made, and what role self-deception (including targeted failure-to-think, e.g. because the subject was painful to think about) may have played in the mistake. Then, brainstorm heuristics as a group for noticing when such failure modes are especially likely, and for reducing their likelihood. Post the results to LessWrong.
Give out pen and paper, and separate out for five minutes to think and jot notes. Everyone chooses a belief that substantially affects what they do. Ask: “Why do I believe what I believe?” Ask: “Given how I actually came to believe this, should I expect my belief to be accurate?”
The ideas below would probably work best for focused discussion on certain specific issues (e.g. reasons for hard takeoff, threat from nanotechnological disasters), but I think what I say below is relevant to rationalist training.
I personally often find that simple freeflowing discussion can be less than optimally effective. I often wish there was a more organized way to focus discussion to get maximum skill development. I’m afraid I’m not aware of research into refinements of general discussion designed to improve knowledge and skill acquisition.
Here are some thoughts I have.
I have found Toastmasters-style impromptu speaking practice to be an efficient and effective way to develop my ability to talk about topics. Impromptu Speaking I can imagine many ways in which such techniques could be employed for rationalist training.
Arguably some ways of doing this could put too much emphasis on rhetorical skills, but I don’t think this need be true. Impromptu speaking exercises are a quick and efficient way to allow people to explore their understanding of a topic in front of an audience. With judicious use of immediate feedback, impromptu speaking could promote rationalist skill acquisition.
One advantage is that impromptu speaking makes efficient use of time. People have to think on their feet and present material fast. (For hard topics, speakers might get one or two minutes to prepare the key points of their presentation.) Public speaking practice improves understanding; even when you understand something fairly well, if you don’t know it well enough to explain to someone else, you don’t really understand it. (The medical school three-step program for learning how to perform a procedure seems appropriate here: See one. Do one. Teach one.)
For impromptu speaking, it is important that speakers only learn what topic they will be talking about after they volunteer; otherwise audience members will be rehearse what they will say when it is their turn to go up. I’ve found that one can cover quite a bit of what one understands in even two or three minutes. If done solely for the purpose of speaker skill acquisition, a designated mentor might call for the speaker to move on as soon as she demonstrates mastery of some area of concern. One key feature found in graduate school oral exams is that as soon as a student appears to demonstrate mastery in a given subject area the testers move the discussion to another subject where the student is potentially weak. It seems the same principle could work here as well.
Alternatively, If done supportingly and judiciously, audience members could interject when advanced participants are speaking and force the speaker to consider some critical point. (If this liberty is abused, each audience member might be restricted to one interjection.)
Other notes:
a) There might also be a designated Socratic questioner who could potentially interpret the speaker at any given point and ask for further refinement or support of some argument. Even just watching such an interaction could be an efficient way to learn, more so than just listening to someone lecture.
b) There are no doubt other ways the feedback could best be tailored to the knowledge-level of the speaker.
c) Perhaps a list of focused readings might be put together ahead of time.
d) There might also be a running EtherPad-type commentary that would allow the speaker to be able to go back and read people’s reactions and commentary on her content. If, at the end, the group desired to have a general discussion having the EtherPad could be very productive because they could start from this base of learning. (If someone knew what the best way to sync the audio with the text that would be helpful too. I know there is a feature like that for Microsoft OneNote, but that wouldn’t allow for network-effect advantages.)
Another idea. What about this? People gain insights into problems with their own views when people with superior insight point out flaws, assumptions, fallacies, biases, etc. in a speaker’s arguments. But if there were simply an unfocused discussion among 10 people the amount of useful feedback might be limited. As a college professor, I’ve found that pair work can be effective to get students practice in articulating their views. Ideally, of course, everyone could pair up with someone who sees things they are missing. Here’s an idea: people first estimate their degree of knowledge about a topic. If there are ten people, someone volunteers for 10 (the highest), then people successively volunteer for 9, 8, 7, 6, and so on. After a topic is presented (and feedback is given to an impromptu speaker), people could sit in a circle in an ordered hierarchy of mentors for that topic. A facilitator could encourage all the organizational issues to take place rapidly. First, the evens turn to their right and discuss for two minutes. In this sempai-kohai relationship, first the kohai (junior member) gives his understanding of the issue, then, for each point, if the sempai has something constructive to offer, he voices it; if not, they move on to the next point. After a period of two or three minutes, the evens then turn to the left. That way, everyone would have the opportunity for focused one-on-one interaction with someone who is potentially more (and less) knowledgeable than themselves. (#10 wouldn’t of course. Perhaps talking #1 could help him to have dramatic gains in understanding though.)
I can imagine different ways the above ranked ring arrangement could be used in combination with the impromptu speaking method. Perhaps, people would only give impromptu talks on one of the topics that they are generally confident about. Then the pair work discussion would commence (potentially after some form of feedback to the impromptu speaker).
Inspired by the recent PUA thread, and by behavioural experiments from CBT:
Take turns approaching random strangers (at a bar, bus stop, etc) and try to build rapport with them for a minute or so. Everyone else observes and provides feedback on their approach and body language.
We could try paranoid debating, as jimmy suggested to me during an IM conversation. One thing to note is that it would be good if rules and suggestions for the game were actually spelled out in the wiki in more detail… the article there is sort of a stub, and could use advice on optimal group size, how long a game runs, a scoring rule, and so on.
Someone could bring a laptop, loaded up with their favorite dual N-back software (like this one?), and have it available for demo purposes. The short term goal would be to inspire some people to be exposed and try it, the longer term goal would be to get people to keep using it long enough to get effects, still farther out we’d hope that the positive effects had a detectable and positive effect on people’s lives. The first two outcomes seem relatively easy to measure and the third much harder.
Again, separate for five minutes and jot notes on the following question:
Exercise: How does the idea of choosing your beliefs “from evidence” make you feel? Are there fears that come up? About what specific beliefs does the idea of choosing your beliefs “based only on evidence” prompt some fear?
Special cases worth attention:
beliefs about yourself
beliefs about friends, family, and other allies (whose feelings you might hurt, or who you might alienate, if you seem to hold negative beliefs about them)
beliefs about what goals are worth hitting
Then, reconvene and discuss (honestly, without rigging the discussion) what the consequences of accurate beliefs on the noted subjects are likely to be.
In my experience there tend to be several quiet members of a group, who don’t speak much in the group setting but will talk in one-on-one settings. It could be a fun exercise to break up into pairs, and have each pair chat individually for a period of time, before rejoining the group conversation.
This would be useful just to build stronger connections between individual members of the group, but it is especially useful if you’re trying to have a focused discussion. Usually a few voices end up dominating, and if that occurs from the very beginning only a few ideas get discussed.
Disclaimer: the experiment I’ve linked to is highly susceptible to priming, so I would advise anyone who will be going to the meetup not to read it (with the exception of JenniferRM, obviously)
In Self-Improvement or Shiny Distraction: Why Less Wrong is anti-Instrumental Rationality patrissimo wrote:
I think a neat thing to try would be for people to reply to this comment with suggestions and links for things to try at the SoCal meetup even if you’re not coming, suggestions appreciated.
Go ahead, you know you want to other-optimize!
Put suggested meetup activities in their own comments like a quotes thread (one suggestion per comment, minimize kibitzing, etc), and vote up or down the ones you think would be particularly high (or low) value.
Before going to the meetup, I will take responsibility for printing out some of the suggestions that have more votes (with ancillary text from links as appropriate) and bring them to the meetup in the hopes that we can try implementing one or more suggestions and then report back with a “background/method/results” format in followup articles. Part of the test would be object level (does a particular suggestion have positive effects) and part of the test would be meta level (can we honestly use LW’s existing infrastructure and community to improve community practices).
I will make a couple suggestions by way of example :-)
It is probably hardest to think rationally about high-impact, emotional subjects, and also high value to think rationally about such subjects.
Accordingly:
Have everyone who is willing to share name a major mistake they have made, and what role self-deception (including targeted failure-to-think, e.g. because the subject was painful to think about) may have played in the mistake. Then, brainstorm heuristics as a group for noticing when such failure modes are especially likely, and for reducing their likelihood. Post the results to LessWrong.
Give out pen and paper, and separate out for five minutes to think and jot notes. Everyone chooses a belief that substantially affects what they do. Ask: “Why do I believe what I believe?” Ask: “Given how I actually came to believe this, should I expect my belief to be accurate?”
After jotting down answers, reconvene and share.
The ideas below would probably work best for focused discussion on certain specific issues (e.g. reasons for hard takeoff, threat from nanotechnological disasters), but I think what I say below is relevant to rationalist training.
I personally often find that simple freeflowing discussion can be less than optimally effective. I often wish there was a more organized way to focus discussion to get maximum skill development. I’m afraid I’m not aware of research into refinements of general discussion designed to improve knowledge and skill acquisition.
Here are some thoughts I have.
I have found Toastmasters-style impromptu speaking practice to be an efficient and effective way to develop my ability to talk about topics. Impromptu Speaking I can imagine many ways in which such techniques could be employed for rationalist training.
Arguably some ways of doing this could put too much emphasis on rhetorical skills, but I don’t think this need be true. Impromptu speaking exercises are a quick and efficient way to allow people to explore their understanding of a topic in front of an audience. With judicious use of immediate feedback, impromptu speaking could promote rationalist skill acquisition.
One advantage is that impromptu speaking makes efficient use of time. People have to think on their feet and present material fast. (For hard topics, speakers might get one or two minutes to prepare the key points of their presentation.) Public speaking practice improves understanding; even when you understand something fairly well, if you don’t know it well enough to explain to someone else, you don’t really understand it. (The medical school three-step program for learning how to perform a procedure seems appropriate here: See one. Do one. Teach one.)
For impromptu speaking, it is important that speakers only learn what topic they will be talking about after they volunteer; otherwise audience members will be rehearse what they will say when it is their turn to go up. I’ve found that one can cover quite a bit of what one understands in even two or three minutes. If done solely for the purpose of speaker skill acquisition, a designated mentor might call for the speaker to move on as soon as she demonstrates mastery of some area of concern. One key feature found in graduate school oral exams is that as soon as a student appears to demonstrate mastery in a given subject area the testers move the discussion to another subject where the student is potentially weak. It seems the same principle could work here as well.
Alternatively, If done supportingly and judiciously, audience members could interject when advanced participants are speaking and force the speaker to consider some critical point. (If this liberty is abused, each audience member might be restricted to one interjection.)
Other notes: a) There might also be a designated Socratic questioner who could potentially interpret the speaker at any given point and ask for further refinement or support of some argument. Even just watching such an interaction could be an efficient way to learn, more so than just listening to someone lecture.
b) There are no doubt other ways the feedback could best be tailored to the knowledge-level of the speaker.
c) Perhaps a list of focused readings might be put together ahead of time.
d) There might also be a running EtherPad-type commentary that would allow the speaker to be able to go back and read people’s reactions and commentary on her content. If, at the end, the group desired to have a general discussion having the EtherPad could be very productive because they could start from this base of learning. (If someone knew what the best way to sync the audio with the text that would be helpful too. I know there is a feature like that for Microsoft OneNote, but that wouldn’t allow for network-effect advantages.)
Another idea. What about this? People gain insights into problems with their own views when people with superior insight point out flaws, assumptions, fallacies, biases, etc. in a speaker’s arguments. But if there were simply an unfocused discussion among 10 people the amount of useful feedback might be limited. As a college professor, I’ve found that pair work can be effective to get students practice in articulating their views. Ideally, of course, everyone could pair up with someone who sees things they are missing. Here’s an idea: people first estimate their degree of knowledge about a topic. If there are ten people, someone volunteers for 10 (the highest), then people successively volunteer for 9, 8, 7, 6, and so on. After a topic is presented (and feedback is given to an impromptu speaker), people could sit in a circle in an ordered hierarchy of mentors for that topic. A facilitator could encourage all the organizational issues to take place rapidly. First, the evens turn to their right and discuss for two minutes. In this sempai-kohai relationship, first the kohai (junior member) gives his understanding of the issue, then, for each point, if the sempai has something constructive to offer, he voices it; if not, they move on to the next point. After a period of two or three minutes, the evens then turn to the left. That way, everyone would have the opportunity for focused one-on-one interaction with someone who is potentially more (and less) knowledgeable than themselves. (#10 wouldn’t of course. Perhaps talking #1 could help him to have dramatic gains in understanding though.)
I can imagine different ways the above ranked ring arrangement could be used in combination with the impromptu speaking method. Perhaps, people would only give impromptu talks on one of the topics that they are generally confident about. Then the pair work discussion would commence (potentially after some form of feedback to the impromptu speaker).
Inspired by the recent PUA thread, and by behavioural experiments from CBT:
Take turns approaching random strangers (at a bar, bus stop, etc) and try to build rapport with them for a minute or so. Everyone else observes and provides feedback on their approach and body language.
We could try paranoid debating, as jimmy suggested to me during an IM conversation. One thing to note is that it would be good if rules and suggestions for the game were actually spelled out in the wiki in more detail… the article there is sort of a stub, and could use advice on optimal group size, how long a game runs, a scoring rule, and so on.
Someone could bring a laptop, loaded up with their favorite dual N-back software (like this one?), and have it available for demo purposes. The short term goal would be to inspire some people to be exposed and try it, the longer term goal would be to get people to keep using it long enough to get effects, still farther out we’d hope that the positive effects had a detectable and positive effect on people’s lives. The first two outcomes seem relatively easy to measure and the third much harder.
Again, separate for five minutes and jot notes on the following question:
Exercise: How does the idea of choosing your beliefs “from evidence” make you feel? Are there fears that come up? About what specific beliefs does the idea of choosing your beliefs “based only on evidence” prompt some fear?
Special cases worth attention:
beliefs about yourself
beliefs about friends, family, and other allies (whose feelings you might hurt, or who you might alienate, if you seem to hold negative beliefs about them)
beliefs about what goals are worth hitting
Then, reconvene and discuss (honestly, without rigging the discussion) what the consequences of accurate beliefs on the noted subjects are likely to be.
Go around in a circle and name something you protect, and what subjects you need accurate beliefs about in order to protect it effectively.
In my experience there tend to be several quiet members of a group, who don’t speak much in the group setting but will talk in one-on-one settings. It could be a fun exercise to break up into pairs, and have each pair chat individually for a period of time, before rejoining the group conversation.
This would be useful just to build stronger connections between individual members of the group, but it is especially useful if you’re trying to have a focused discussion. Usually a few voices end up dominating, and if that occurs from the very beginning only a few ideas get discussed.
Disclaimer: the experiment I’ve linked to is highly susceptible to priming, so I would advise anyone who will be going to the meetup not to read it (with the exception of JenniferRM, obviously)
Try this breathing experiment out, and document the results for posterity.