tl;dr: this is a guide for running a productive disagreement workshop. Decide of modalities, choose a few techniques, study, practice, make some post-workshop materials and be cautious.
Epistemic Status: Quickly written. Literally reporting what I started doing, this is just a first iteration and it will probably evolve in the near future. This is not perfectly field-proof, specialist advice. Feel free to point to style/syntax/grammar errors in the comments.
[Update : Changed title from “conversational technique” to “handling hostile disagreements”]
There has recently been a post pondering whether LessWrong was missing an opportunity when it comes to some dialectic techniques aiming to improve reasoning and rapport. I’ve been in contact with those and diving into the related litterature for some time now -since about 2018- and saw an opportunity in trying to teach some of those techniques to people around me.
I have started to give workshops to local EAs in Paris, and have received positive feedback so far. Despite the program being in its infancy, I think it would be great if others were trying to set it up. I think it targets skills that are underdevelopped in most of the rationalist sphere, and that the people going through it are usually satisfied and happy to discover them.
Here is what I did :
1-Duration and amenities
In my opinion, a weekend where atendees come from 10am to 17pm, with a long pause from 12am to 1pm for eating, is a good formula. One technique usually requires three hours of presentation and exercices. Also, have snacks ready. During a weekend, requesting 12 hours of attention over two days from the attendees was about the maximum I could achieve without burning them out. It’s also important for people to be on time. It’s almost impossible to catch a workshop in-progress. However, you’re ultimately the one who will know why you’re giving your workshop this form and why you teach the techniques is that particular order. Your personal models about how people learn and what you can expect from them matter a lot.
I did try to make workshops for only an afternoon. More people are usually free, but they need to leave early, and it doesn’t create the long-duration, conversation-prone atmosphere that a weekend can offer. [Update : after some practice and experimentation, I think it is possible (but hard) to bring the training time down to 1 hour per technique. This however requires to shove Q&A at the end of the workshop.]
Because people will be lead to discuss personal topics and create new connexions, it’s better to do this in a place that feels welcoming and comfortable. I did it in my own house, because it checks those criteria, but if you have more money you could, I guess, book somewhere else. [Update : so far I find doing short workshops in venues people come in for an event more functional than doing them at home.]
2-Choose, study and practice techniques.
There is more to choose from than you could think: Street Epistemology, Deep Canvassing, Smart Politics techniques, Motivational Interviewing, Principled Negotiation, Crisis Negotiation, or Non-Violent Communication to name a few (some of these techniques are linked together within a knowledge-transmission network/re-discover the same phenomenon). And this is not a definitive list ! I’m just trying stuff in my living room, so feel free to search for other techniques you’ve heard of, and which seem fitting for the usual goals of rationalism. Do survey several of them, do not make yourself a Man of the One Technique. More than this, remember that a technique is merely a gesture pointing to a state of mind. [Update: I find it useful to suggest cultivating attitudes, e.g. resilience, by doing exercices designed at doing so, e.g. listening to a politician you hate and try focusing only on your bodily reactions]
Each of the techniques has its own goals and its own level of evidence. I encourage you to reflect on the following : What do you want to achieve by teaching those techniques ? You’re the one deciding what “package” you want, and why do you want that.
I strongly recommend reading the reference books for each of the techniques you want to practice (see section “Resources”), as well as watching at least one video-recorded example. The books cost some money, so you might want to ask your future participants for monetary help, or rely on the websites we all know. There are also a host of youtube videos introducing most of these methods.
However, my personal impression is that, once you’ve practiced a technique enough (I’d say about 10 hours of training, 10 hours of deployment), learning the others becomes much easier. I also feel that meditation, psychotherapeutic drugs and therapy itself helped me become more skilled in those. You’ll also find several discord severs or Facebook page where you can practice these techniques.
3-Prepare your notes and guidelines, have a whiteboard ready.
Form matters. I could not improvise transmitting those techniques. As they are transmitted, errors and mutations can come in. It’s better for me to have a reference material that I stick to.
Ethical guidelines matter too. Despite them existing for each technique, during the workshop you will be the one transmitting them, ever-so-slighty promoting them in function of your preferences. There is usually a tradeoff to strike between promoting norms that make you as reality-revealing as can be, and norms that are easy to apply and not weird. I personally insist on consent, transparency, and espitemic rationality, as I think not holding those to a very high level can make people perceive the technique as manipulative. Also, you want practicioners to keep an eye on engaging in reality-revealing puzzles rather than reality-masking ones. An important part of these guidelines, in my opinion, is deciding what personal goals are suitable for this technique.
Finally, whiteboard ! It’s absolutely essential. You will need to introduce elements progressively, answer the questions of different people, make schemes, change the schemes, all of it is very dynamic. I don’t think a powerpoint could do this satisfactorily.
4-Introduce techniques phase-by-phase through (e.g.) roleplay.
During the workshop propper, you cannot simply introduce the whole technique, then ask people to execute it. It’s better to introduce first the technique through an example video, then take questions and reactions from the atendees. You can then go on explaining the first two phases of the technique, but then an exercice is required for people to internalize them. Once you’ve done the exercices, you can add new steps. [Update : Calling bullshit on my past self. Actually, people are pretty good at learning the whole technique at once, only with a few details lacking or misunderstandings happening sporadically]
I rely on roleplay. This has its limitations -notoriously, it’s not as good as other methods- but it is one of our best alternative absent of actual data and specialized software. How does it go ? Each technique usually requires a questioner and an interlocutor. The atendees form pairs. Person A offers a belief they hold -usually something such as “animals suffer” or “God does not exist”. Person B will pretend (or not !) to believe the exact opposite, with very high confidence. Person B will be the interlocutor, and Person A will be the questioner. They have 5 to 15 minutes of conversation (depending on the phase they’re practicing), then switch.
Participants are also offered to “simulate” someone they know who holds a certain belief they share or not. If they simulate, they are the Interlocutor.
I personally request atendees not to discuss:
Claims with an evaluative or ethical component (“Elon Musk is fantastic/horrible”, “We should not eat animals”). These are harder and best to practice later on, once participants are already comfortable with discussing matter-of-facts. [Update : valid only for Street Epistemology, other techniques fare better]
Very abstract claims (“Morality is objective”, “Math is true”)
Important claims related to their career choice or potentially impactful decisions that require extremely careful deliberation (“Davidad’s plan for alignment is functional”, “Singular Learning Theory is promising”)
And, obviously, their family (“My sister is gay”, “My father is a criminal”). I insist. Talking about your family is usually doomed to crash the atmosphere. For some reason people keep doing this if not informed not to. [Update : some people handle it gracefully, so if they insist, I let them do]
At any point, the atendees can raise their hand, and an instructor will come to offer help. This usually happens a lot in the beginning to test wether a certain topic is acceptable or not.
6-Handouts and “alumni” group.
After the workshop, participants might want to look back on what they have learned. I personally made a bunch of small handouts that serve as post-workshop material in order to help them remember how the techniques go. I also created an “alumni” group on Messenger, that allows people who came to discuss their experience and the kinds of problems they might be meeting in applying the technique “in the wild”.
5-Develop finer-grained models for analyzing reasoning, argumentation, narration and conversation.
Be Curious ! All of these techniques interact with several layers of reality. Some steps are sometimes very blurry and imprecise (e.g. “Step 9: Question the reliability of your interlocutor’s method”) and might prove very hard to put in practice. You’ll need to break things down a bit more than what those techniques usually do in order to get a clearer picture. I’m not really willing to share my views on this, first because that is worth a whole sequence in itself, but also because I think we’re still in a phase where a greater diversity of perspectives is needed. We’re still trying to understand the problem, after all.
6-Resources
I hope this post will turn out helpful, feel free to contact me if you’re interested in my advice when replicating it in your own area, or online. In the meantime, you can check the ressources below. These are links of what I think is useful in order to learn each of the techniques (I’ll add recommendations suggested in the comments):
First, How Minds Change, if you haven’t already read it. [Update: You can also try this tool on Clearer Thinking]
I personally add caveats for this technique, because it doesn’t sound epistemically asymetrical to me. I usually encourage it for extending the listener’s empathy (falsify the belief “other side/population X = dumb bastard”), and if in dire need to discuss an actual belief, use it to ask a question and not falsify it (change “X is true” to “Is X true ?”, but not “X is false”).
Non-Violent Communication—the book linked. Actual retreats to practice it (I found a week of practice hard. Like, really hard. I don’t expect the book to be very effective in comparison).
Important notes :
-These workshops are not suitable for people who have a very strong personal theory of argumentation and change of mind that is driven by conflict -I guess that trauma, being marginalized or strong personality traits can contribute to this. Some people may answer with conflict when offered a technique that deliberately attempts to remove it. This is tricky and I don’t think we can handle those edge-cases for now.
-I encourage you to dispense these workshops responsibly and not to blindly advocate for them. This hasn’t been around for a long time within the rationalist community, and it will probably change and evolve as we spot some inconsistencies or harmful stuff. I encourage you to navigate this with an open and empirical mind. This also means gatekeeping workshops and refusing people who have a history of manipulation or harassment.
-If anything goes wrong, report it ! You can either send me a message or make a post if you think it’s worth it. That will help us improve it or eventually discard it as evidence accumulates that it is not worth it.
-Completely random thought : “Hey, could any of that be useful for AI Safety via Debate, somehow ?” (tentative answer: probably, but I’m not sure which idea or observation could be helpful and when).
Running your own workshop on handling hostile disagreements
tl;dr: this is a guide for running a productive disagreement workshop. Decide of modalities, choose a few techniques, study, practice, make some post-workshop materials and be cautious.
Epistemic Status: Quickly written. Literally reporting what I started doing, this is just a first iteration and it will probably evolve in the near future. This is not perfectly field-proof, specialist advice. Feel free to point to style/syntax/grammar errors in the comments.
[Update : Changed title from “conversational technique” to “handling hostile disagreements”]
There has recently been a post pondering whether LessWrong was missing an opportunity when it comes to some dialectic techniques aiming to improve reasoning and rapport. I’ve been in contact with those and diving into the related litterature for some time now -since about 2018- and saw an opportunity in trying to teach some of those techniques to people around me.
I have started to give workshops to local EAs in Paris, and have received positive feedback so far. Despite the program being in its infancy, I think it would be great if others were trying to set it up. I think it targets skills that are underdevelopped in most of the rationalist sphere, and that the people going through it are usually satisfied and happy to discover them.
Here is what I did :
1-Duration and amenities
In my opinion, a weekend where atendees come from 10am to 17pm, with a long pause from 12am to 1pm for eating, is a good formula. One technique usually requires three hours of presentation and exercices. Also, have snacks ready.
During a weekend, requesting 12 hours of attention over two days from the attendees was about the maximum I could achieve without burning them out. It’s also important for people to be on time. It’s almost impossible to catch a workshop in-progress.
However, you’re ultimately the one who will know why you’re giving your workshop this form and why you teach the techniques is that particular order. Your personal models about how people learn and what you can expect from them matter a lot.
I did try to make workshops for only an afternoon. More people are usually free, but they need to leave early, and it doesn’t create the long-duration, conversation-prone atmosphere that a weekend can offer.
[Update : after some practice and experimentation, I think it is possible (but hard) to bring the training time down to 1 hour per technique. This however requires to shove Q&A at the end of the workshop.]
Because people will be lead to discuss personal topics and create new connexions, it’s better to do this in a place that feels welcoming and comfortable. I did it in my own house, because it checks those criteria, but if you have more money you could, I guess, book somewhere else. [Update : so far I find doing short workshops in venues people come in for an event more functional than doing them at home.]
2-Choose, study and practice techniques.
There is more to choose from than you could think: Street Epistemology, Deep Canvassing, Smart Politics techniques, Motivational Interviewing, Principled Negotiation, Crisis Negotiation, or Non-Violent Communication to name a few (some of these techniques are linked together within a knowledge-transmission network/re-discover the same phenomenon).
And this is not a definitive list ! I’m just trying stuff in my living room, so feel free to search for other techniques you’ve heard of, and which seem fitting for the usual goals of rationalism. Do survey several of them, do not make yourself a Man of the One Technique. More than this, remember that a technique is merely a gesture pointing to a state of mind.
[Update: I find it useful to suggest cultivating attitudes, e.g. resilience, by doing exercices designed at doing so, e.g. listening to a politician you hate and try focusing only on your bodily reactions]
Each of the techniques has its own goals and its own level of evidence. I encourage you to reflect on the following : What do you want to achieve by teaching those techniques ? You’re the one deciding what “package” you want, and why do you want that.
I strongly recommend reading the reference books for each of the techniques you want to practice (see section “Resources”), as well as watching at least one video-recorded example. The books cost some money, so you might want to ask your future participants for monetary help, or rely on the websites we all know. There are also a host of youtube videos introducing most of these methods.
However, my personal impression is that, once you’ve practiced a technique enough (I’d say about 10 hours of training, 10 hours of deployment), learning the others becomes much easier. I also feel that meditation, psychotherapeutic drugs and therapy itself helped me become more skilled in those. You’ll also find several discord severs or Facebook page where you can practice these techniques.
3-Prepare your notes and guidelines, have a whiteboard ready.
Form matters. I could not improvise transmitting those techniques. As they are transmitted, errors and mutations can come in. It’s better for me to have a reference material that I stick to.
Ethical guidelines matter too. Despite them existing for each technique, during the workshop you will be the one transmitting them, ever-so-slighty promoting them in function of your preferences. There is usually a tradeoff to strike between promoting norms that make you as reality-revealing as can be, and norms that are easy to apply and not weird. I personally insist on consent, transparency, and espitemic rationality, as I think not holding those to a very high level can make people perceive the technique as manipulative. Also, you want practicioners to keep an eye on engaging in reality-revealing puzzles rather than reality-masking ones. An important part of these guidelines, in my opinion, is deciding what personal goals are suitable for this technique.
Finally, whiteboard ! It’s absolutely essential. You will need to introduce elements progressively, answer the questions of different people, make schemes, change the schemes, all of it is very dynamic. I don’t think a powerpoint could do this satisfactorily.
4-Introduce techniques phase-by-phase through (e.g.) roleplay.
During the workshop propper, you cannot simply introduce the whole technique, then ask people to execute it. It’s better to introduce first the technique through an example video, then take questions and reactions from the atendees. You can then go on explaining the first two phases of the technique, but then an exercice is required for people to internalize them. Once you’ve done the exercices, you can add new steps.
[Update : Calling bullshit on my past self. Actually, people are pretty good at learning the whole technique at once, only with a few details lacking or misunderstandings happening sporadically]
I rely on roleplay. This has its limitations -notoriously, it’s not as good as other methods- but it is one of our best alternative absent of actual data and specialized software. How does it go ?
Each technique usually requires a questioner and an interlocutor. The atendees form pairs. Person A offers a belief they hold -usually something such as “animals suffer” or “God does not exist”. Person B will pretend (or not !) to believe the exact opposite, with very high confidence. Person B will be the interlocutor, and Person A will be the questioner. They have 5 to 15 minutes of conversation (depending on the phase they’re practicing), then switch.
Participants are also offered to “simulate” someone they know who holds a certain belief they share or not. If they simulate, they are the Interlocutor.
I personally request atendees not to discuss:
Claims with an evaluative or ethical component (“Elon Musk is fantastic/horrible”, “We should not eat animals”). These are harder and best to practice later on, once participants are already comfortable with discussing matter-of-facts. [Update : valid only for Street Epistemology, other techniques fare better]
Very abstract claims (“Morality is objective”, “Math is true”)
Important claims related to their career choice or potentially impactful decisions that require extremely careful deliberation (“Davidad’s plan for alignment is functional”, “Singular Learning Theory is promising”)
And, obviously, their family (“My sister is gay”, “My father is a criminal”). I insist. Talking about your family is usually doomed to crash the atmosphere. For some reason people keep doing this if not informed not to. [Update : some people handle it gracefully, so if they insist, I let them do]
At any point, the atendees can raise their hand, and an instructor will come to offer help. This usually happens a lot in the beginning to test wether a certain topic is acceptable or not.
6-Handouts and “alumni” group.
After the workshop, participants might want to look back on what they have learned. I personally made a bunch of small handouts that serve as post-workshop material in order to help them remember how the techniques go. I also created an “alumni” group on Messenger, that allows people who came to discuss their experience and the kinds of problems they might be meeting in applying the technique “in the wild”.
5-Develop finer-grained models for analyzing reasoning, argumentation, narration and conversation.
Be Curious ! All of these techniques interact with several layers of reality. Some steps are sometimes very blurry and imprecise (e.g. “Step 9: Question the reliability of your interlocutor’s method”) and might prove very hard to put in practice. You’ll need to break things down a bit more than what those techniques usually do in order to get a clearer picture. I’m not really willing to share my views on this, first because that is worth a whole sequence in itself, but also because I think we’re still in a phase where a greater diversity of perspectives is needed. We’re still trying to understand the problem, after all.
6-Resources
I hope this post will turn out helpful, feel free to contact me if you’re interested in my advice when replicating it in your own area, or online. In the meantime, you can check the ressources below. These are links of what I think is useful in order to learn each of the techniques (I’ll add recommendations suggested in the comments):
First, How Minds Change, if you haven’t already read it.
[Update: You can also try this tool on Clearer Thinking]
Street Epistemology :
How to Have Impossible Conversations
This video (and the channel in general). Note that it is 5 years old and more recent ones can be found, but I haven’t watched them.
The very-soon-to-come Street Epistemology Course.
The Atheos App
This webpage
The discord server
Deep Canvassing
This Vox Article, mostly the video in it.
This video
This random webpage
Probably participating to a retreat
I personally add caveats for this technique, because it doesn’t sound epistemically asymetrical to me. I usually encourage it for extending the listener’s empathy (falsify the belief “other side/population X = dumb bastard”), and if in dire need to discuss an actual belief, use it to ask a question and not falsify it (change “X is true” to “Is X true ?”, but not “X is false”).
Smart Politics techniques
This video
Motivational Interviewing
This book
These two videos.
Probably attending a workshop.
Principled Negotiation—the book linked.
Crisis Negotiation—the article linked.
Non-Violent Communication—the book linked.
Actual retreats to practice it (I found a week of practice hard. Like, really hard. I don’t expect the book to be very effective in comparison).
Important notes :
-These workshops are not suitable for people who have a very strong personal theory of argumentation and change of mind that is driven by conflict -I guess that trauma, being marginalized or strong personality traits can contribute to this. Some people may answer with conflict when offered a technique that deliberately attempts to remove it. This is tricky and I don’t think we can handle those edge-cases for now.
-I encourage you to dispense these workshops responsibly and not to blindly advocate for them. This hasn’t been around for a long time within the rationalist community, and it will probably change and evolve as we spot some inconsistencies or harmful stuff. I encourage you to navigate this with an open and empirical mind. This also means gatekeeping workshops and refusing people who have a history of manipulation or harassment.
-If anything goes wrong, report it ! You can either send me a message or make a post if you think it’s worth it. That will help us improve it or eventually discard it as evidence accumulates that it is not worth it.
-Completely random thought : “Hey, could any of that be useful for AI Safety via Debate, somehow ?” (tentative answer: probably, but I’m not sure which idea or observation could be helpful and when).