Rationally Ending Discussions

Dis­cus­sions should end grace­fully. There should be some clar­ity about why they’re end­ing and why that’s ra­tio­nally ac­cept­able. If some­one wants to con­tinue the dis­cus­sion, they should have the op­por­tu­nity to see why this end­ing is rea­son­able. (Or, failing that, they should have some ev­i­dence available to en­able them to ar­gue their case that it was un­rea­son­able. They shouldn’t be left with lit­tle or no data about what hap­pened.)

If you dis­cuss poorly, it’s im­por­tant that you can learn from that and do bet­ter next time. If you want it, you should have ac­cess to some crit­i­cal feed­back, some ex­pla­na­tion of the other per­son’s per­spec­tive, or some­thing to en­able progress. If you’re run­ning into prob­lems in­ter­act­ing with peo­ple, but no one will tell you what the prob­lem is, that’s bad for ra­tio­nal progress.

The more eas­ily dis­cus­sions can end, the more eas­ily they can start. We don’t want dis­cussing to be a big bur­den or com­mit­ment (it’s OK if a few dis­cus­sions re­quire high effort, when peo­ple have rea­son to vol­un­tar­ily choose that, but that shouldn’t be the de­fault).

Dis­cus­sions can end by mu­tual agree­ment. If no one ob­jects to the end­ing, that’s fine. That’s one way to end dis­cus­sions. Pretty much ev­ery­one agrees on this. The difficulty is that some­times peo­ple don’t agree. Some­one wants to stop dis­cussing but some­one else wants to pur­sue the mat­ter it fur­ther.

I don’t think agree­ing to dis­agree, by it­self, is a good rea­son to end a dis­cus­sion by mu­tual agree­ment. Disagree­ments can be re­solved! There should be an ex­tra fac­tor like agree­ing that it’s rea­son­able to pri­ori­tize other stuff (which we may leave im­plied with­out ex­plic­itly men­tion­ing). There are many prob­lems to work on, and time/​en­ergy/​etc are scare re­sources, so it’s fine to drop a dis­agree­ment (for now, in­definitely) if peo­ple think they can get more value work­ing on other stuff.

End­ing dis­cus­sions when a sin­gle per­son wants to, for any rea­son, with no ex­pla­na­tion, is prob­le­matic. For ex­am­ple, peo­ple will end dis­cus­sions when they (the ideas they are bi­ased in fa­vor of) start los­ing the ar­gu­ment.

But we don’t want to trap peo­ple in dis­cus­sions. Some­times one guy has a lot of en­ergy to dis­cuss some­thing for­ever but you want to stop. There are lots of le­gi­t­i­mate rea­sons to end a dis­cus­sion.

You can try to ex­plain your rea­sons for end­ing a dis­cus­sion, but no mat­ter what you say, the other guy might dis­agree. This is a real con­cern be­cause you don’t want to dis­cuss ex­tra with the most un­rea­son­able peo­ple who give low qual­ity coun­ter­ar­gu­ments to all your rea­sons for stop­ping dis­cus­sion. Mean­while the most rea­son­able peo­ple tend vol­un­tar­ily to let you out of dis­cus­sions, so you dis­cuss with them the least!?

There has to be a way to unilat­er­ally end dis­cus­sions. You end it with­out agree­ment. But it should have some pro­tec­tions against abuse. We don’t want it to be ac­cept­able to ar­bi­trar­ily end any dis­cus­sion at any time for no rea­son or awful rea­sons. This is for your own sake, too, not just for the benefit of oth­ers. If I want to end a dis­cus­sion and the other guy dis­agrees, I ought to con­sider that maybe I’m bi­ased. I might be evad­ing the is­sue, avoid­ing be­ing cor­rected, etc. I shouldn’t just fully trust my own judg­ment. I should want some dis­cus­sion poli­cies that make it harder for me to be and stay ir­ra­tional, bi­ased, un­rea­son­able, etc.

Note: Of course any­one can stop talk­ing at any time for no rea­son. That’s a mat­ter of free­dom. No one should be held hostage. The is­sue is the con­se­quences for your rep­u­ta­tion. What is con­sid­ered rea­son­able or ra­tio­nal? Some dis­cus­sion end­ing be­hav­ior ought to be seen nega­tively by the com­mu­nity. Some good ways of end­ing dis­cus­sions out to be en­couraged, wide­spread and nor­mal. Some bad ways of end­ing dis­cus­sions should be dis­in­cen­tivized.

Note: Even if a dis­cus­sion ac­tively con­tinues, you could par­ti­ci­pate in it on less than a daily ba­sis. Dis­cus­sions don’t have to be fast. Some norms are needed for what is stal­ling a dis­cus­sion out (e.g. one re­ply per decade would be a way to pre­tend you didn’t end the dis­cus­sion when ba­si­cally you did). In my ex­pe­rience, peo­ple are usu­ally pretty rea­son­able about dis­cus­sion pac­ing, with a few no­table ex­cep­tions. (The main prob­lem I see is peo­ple who dis­cuss a bunch ini­tially and then never come back to it as soon as they go do some­thing else or as soon as they sleep once.)

So we want a way to end a dis­cus­sion, even if other peo­ple dis­agree with end­ing the dis­cus­sion, but which has some pro­tec­tion against abuse (bad faith), er­ror, bias, ir­ra­tional­ity, etc. It should ideally provide some trans­parency: some ev­i­dence about why the dis­cus­sion is end­ing that can be judged neu­trally, pos­i­tively or nega­tively by the au­di­ence. And it should provide some sort of feed­back or learn­ing op­por­tu­nity for the guy who didn’t want to stop here.

So here’s the first draft of a policy: When you want to end a dis­cus­sion, you are ex­pected to right one fi­nal mes­sage which ex­plains why you’re end­ing the dis­cus­sion. You’re also ex­pected to read one more mes­sage from the other guy, so he has one chance to point out that you’re mak­ing a mis­take and he could po­ten­tially tell you why you should change your mind about end­ing the dis­cus­sion.

What’s good about this policy? It helps limit abuse. It’s harder to end a dis­cus­sion for a bad rea­son if you have to ex­plain your­self. The other guy gets some info about what hap­pened. The other guy has a chance at a re­but­tal so you could po­ten­tially be cor­rected. And it’s not very time con­sum­ing. It puts a small, strict limit on how much more dis­cus­sion hap­pens af­ter you de­cide you’d like to wrap it up.

This is a pretty min­i­mal policy. I think it could be a good de­fault ex­pec­ta­tion that LW could use for any dis­cus­sion where peo­ple have each writ­ten 3+ mes­sages (or maybe 5+ to re­duce the bur­den? The num­ber could be tuned if this was tried out for a while). That way it won’t add an ex­tra bur­den to re­ally small dis­cus­sions. Tiny dis­cus­sions would be dis­cour­aged by any over­head at all. Tiny dis­cus­sions are also lower in­vest­ment so end­ing them is a smaller deal. Peo­ple haven’t formed a rea­son­able ex­pec­ta­tion of reach­ing a con­clu­sion, get­ting their ques­tions an­swered, or any­thing else. They’re just shar­ing some thoughts on an ad hoc, no-obli­ga­tion ba­sis. That’s fine. But for more sub­stan­tive dis­cus­sions, I think adding a lit­tle bit of an end­ing cost is rea­son­able.

The min­i­mal policy has some down­sides. If we had a policy that takes more effort, we could fix some prob­lems and get some benefits. So I think for dis­cus­sions that go longer (e.g. 10+ mes­sages each) or when peo­ple mu­tu­ally agree to make it a sub­stan­tive dis­cus­sion, then a longer but bet­ter ap­proach could be used for unilat­er­ally end­ing a dis­cus­sion.

What are prob­lems with the sin­gle part­ing mes­sage? It could be un­clear. It could ig­nore a key is­sue. It could mis­rep­re­sent the other guy’s po­si­tions or mis­rep­re­sent what hap­pened in the dis­cus­sion. It could be poorly thought through and show ma­jor signs of bias.

What are the prob­lems with a sin­gle re­but­tal to the part­ing mes­sage that won’t be replied to? If it asks any clar­ify­ing ques­tions, they won’t be an­swered. Any great points could be ig­nored with­out ex­pla­na­tion.

So as a next step up, we could have a two-part dis­cus­sion end­ing. In­stead of one more back and forth (part­ing mes­sage + re­but­tal), we could have two more back and forths. Ini­tial part­ing mes­sage, re­but­tal and ques­tions, fi­nal part­ing mes­sage, and then fi­nal re­but­tal.

BTW, the re­but­tals are semi-op­tional. You can just de­cide to agree with the guy’s part­ing mes­sage if you want (maybe it makes sense to you once he ex­plains his po­si­tion). Or in­stead of a re­but­tal or agree­ment, your other op­tion is to write your own part­ing mes­sage. But you shouldn’t dis­agree with their part­ing mes­sage and then silently end the dis­cus­sion with zero ex­pla­na­tion of what’s go­ing on.

Note: Part­ing mes­sages don’t have to be very long. A few clearly writ­ten sen­tences can cover the key points (e.g. your opinion of the dis­cus­sion, your fi­nal com­ments on some open is­sues, your rea­sons for end­ing). A bit longer is needed if you write fluff. And gen­er­ally you should write a bit more if the dis­cus­sion was longer.

With a two back-and-forth dis­cus­sion end­ing, it’s still pos­si­ble to dodge ques­tions, avoid key is­sues, etc. It can take quite a few iter­a­tions to get some stuff cleared up, and that’s if peo­ple are be­ing rea­son­able. Un­rea­son­able peo­ple can sab­o­tage dis­cus­sions in­definitely.

So what about a three back-and-forth dis­cus­sion end­ing? Or four or five? Noth­ing will be perfect or give a guaran­tee.

Let’s con­sider other ap­proaches. What about a 5% end­ing? How­ever many words you wrote in the dis­cus­sion, you should write 5% of that num­ber in the dis­cus­sion end­ing phase. That seems kinda rea­son­able. That means for ev­ery 20 words of dis­cus­sion you write, you’re po­ten­tially obli­gat­ing your­self to one word later. This might need to be capped for very long dis­cus­sions. This means your effort to end the dis­cus­sion grace­fully is pro­por­tional to the effort you put into the dis­cus­sion.

This ap­proach still suffers from be­ing a fairly ar­bi­trary cut­off. You just de­cide to end the dis­cus­sion, say a few things that hope­fully do a good job of ex­pos­ing your rea­son­ing to crit­i­cism and giv­ing the other guy the chance to learn that he’s wrong, and say a few things to wrap up the open is­sues (like briefly an­swer­ing a few key ques­tions you hadn’t got­ten to, so the dis­cus­sion is left in a more com­plete form and your case is left ad­e­quately com­plete that some­one could learn from you). I think that’s way bet­ter than noth­ing but still has sig­nifi­cant po­ten­tial to go wrong.

One use­ful tech­nique is agree­ing to es­ca­late the com­mit­ment to the dis­cus­sion. You can say “I will dis­cuss X but only if you’ll agree to a 3 back and forth end­ing if you want to end the dis­cus­sion when I don’t (which I’ll also agree to if I want to end it unilat­er­ally).” It some­times makes sense to not want to in­vest in a dis­cus­sion then have it end abruptly in a way that’s un­satis­fy­ing and in­con­clu­sive from your per­spec­tive.

It makes sense to want a dis­cus­sion to ei­ther be pro­duc­tive or else the other guy makes a clear claim – ex­plained enough that you could learn from it – about what you’re do­ing wrong, so you have the op­por­tu­nity to im­prove (or maybe to crit­i­cize his er­ror and judge him, rather than be­ing left with a lack of data). Some­one could also ex­plain why the dis­cus­sion isn’t work­ing in a no-fault way, e.g. you and he have some in­com­pat­i­ble traits.

Say­ing “I’m busy” is broadly a bad ex­cuse to end dis­cus­sions. You were busy when you started, too, right? What changed? Some­times peo­ple ac­tu­ally get sig­nifi­cantly bus­ier in an un­fore­see­able way in the mid­dle of a dis­cus­sion, but that shouldn’t be very com­mon. Usu­ally “I’m busy” is code for “I think your mes­sages are low qual­ity and in­ad­e­quately valuable”. That claim isn’t very satis­fy­ing for the other guy with­out at least one ex­am­ple quote and some crit­i­cal anal­y­sis of what is bad about the quote. Often peo­ple speak in gen­eral terms about low qual­ity dis­cus­sion with­out any quoted ex­am­ples, which also tends to be un­satis­fac­tory, be­cause the per­son be­ing crit­i­cized is like “Uhh, I don’t think I did that thing you’re ac­cus­ing me of. I can’t learn from these vague claims. You aren’t show­ing me any ex­am­ples. Maybe you mi­s­un­der­stood me or some­thing.”

It can be rea­son­able to say “I thought I’d be in­ter­ested in this topic but it turns out I’m not that in­ter­ested.” You shouldn’t say this of­ten but oc­ca­sion­ally is OK. Shit hap­pens. You can end a dis­cus­sion due to your own mis­take. When you do, you shouldn’t hide it. Let the other guy and the au­di­ence know that you aren’t blam­ing him. And maybe by shar­ing the prob­lem you’ll be able to get some ad­vice about how to do bet­ter next time. Or if you share the prob­lem, maybe af­ter a bunch of dis­cus­sions you’ll be able to re­view why they ended and find some pat­terns and then re­al­ize you have a re­cur­ring prob­lem you should work on.

Im­passe Chains

Be­sides try­ing to end a dis­cus­sion grace­fully with a part­ing phase where a few things get ex­plained, I have a differ­ent pro­posal: im­passe chains.

An im­passe is a state­ment about why the dis­cus­sion isn’t work­ing. We’re stuck be­cause of this im­passe. It’s ex­plain­ing some prob­lem in the dis­cus­sion which is im­por­tant enough to end the dis­cus­sion (rather than be­ing minor and ig­nor­able). What if no one prob­lem is ru­inous but sev­eral are adding up to a ma­jor prob­lem? Then the im­passe is the con­junc­tion of the smaller prob­lems: group them to­gether and ex­plain why the group is an im­passe.

Stat­ing an im­passe pro­vides trans­parency and gives the other guy some op­por­tu­nity to po­ten­tially ad­dress or learn from the dis­cus­sion prob­lem.

Im­passes are meant to, hope­fully, be solved. You should try to say what’s go­ing wrong that, if it was changed to your satis­fac­tion, you’d ac­tu­ally want to con­tinue.

The other guy can then sug­gest a solu­tion to the im­passe or agree to stop. A solu­tion can be a di­rect solu­tion (fix the prob­lem) or an in­di­rect solu­tion (a workaround or a bet­ter way to think about the is­sue, e.g. a rea­son the prob­lem is mis­con­ceived and isn’t re­ally a prob­lem). You should also try to think about solu­tions to im­passes your­self.

Some­times the guy will rec­og­nize the im­passe ex­ists. Other times it’ll seem strange to him. He wasn’t see­ing the dis­cus­sion that way. So there’s some op­por­tu­nity for clar­ifi­ca­tion. Lots of times that some­one wants to end a dis­cus­sion, it fol­lows some sort of fix­able mi­s­un­der­stand­ing.

So far an im­passe is just a way to think about a part­ing mes­sage, and you can hope­fully see why con­tin­u­ing at least one more mes­sage past the ini­tial im­passe claim makes sense. So you may think this im­passe ap­proach just sug­gests hav­ing 2-5 mes­sages (per per­son) to end dis­cus­sions. And that’s de­cent – a lot of dis­cus­sions do way worse – but I also have a differ­ent sug­ges­tion.

The sug­gest is to chain im­passes to­gether.

So step 0, we dis­cuss.

Step 1, I say an im­passe and we try to solve it. This is an im­passe re­gard­ing step 0.

Step 2, I de­cide the prob­lem solv­ing isn’t work­ing in some way (oth­er­wise I’d be happy to con­tinue). So I state an im­passe with the prob­lem solv­ing. This is an im­passe re­gard­ing step 1.

Step 3, we try to solve the sec­ond im­passe. Either this prob­lem solv­ing dis­cus­sion satis­fies me or I see some­thing wrong with it. If it’s not work­ing, I say an im­passe with this dis­cus­sion. This is the third im­passe.

Each time an im­passe is stated, the pre­vi­ous dis­cus­sion is set aside and the im­passe be­comes the new topic of dis­cus­sion. (Though a few clos­ing com­ments on the pre­vi­ous dis­cus­sion are fine and may be a good idea.) The im­passe is ei­ther solved (and then you can re­turn to the prior dis­cus­sion) or leads to a new im­passe. This can re­peat in­definitely. You can have an im­passe about the dis­cus­sion of an im­passe about the dis­cus­sion of an im­passe in the origi­nal dis­cus­sion.

The im­passes are chained to­gether. Each one is linked to the pre­vi­ous one. This is differ­ent than mul­ti­ple sep­a­rate im­passes with the origi­nal dis­cus­sion. Here, we’re deal­ing with one im­passe for the origi­nal dis­cus­sion and then the other im­passes in the chain are all at meta lev­els.

Note: If you see mul­ti­ple im­passes with the origi­nal dis­cus­sion, of­ten that means you tried to ig­nore one. In­stead of bring­ing up the first one and try­ing to do prob­lem solv­ing, you let prob­lems ac­cu­mu­late. It’s pos­si­ble for mul­ti­ple im­passes to come up at the same time but it isn’t very com­mon. In any case, you can deal with the im­passes one at a time. Pick one to fo­cus on. If it gets re­solved, move on to the next one.

It doesn’t make sense to ask some­one to dis­cuss X fur­ther when he sees an im­passe with dis­cus­sion X (mean­ing a rea­son that dis­cus­sion isn’t work­ing). You’ll have to ad­dress that prob­lem in some way or agree to stop. Dis­cus­sion the prob­lem it­self is a differ­ent dis­cus­sion than dis­cussing X, so it should be pos­si­ble to try it.

The more im­passes chain, usu­ally the clearer the situ­a­tion gets. Each level tends to be sim­pler than the pre­vi­ous level. There are fewer is­sues in play. It be­comes more ob­vi­ous what to do. This helps but isn’t nearly enough to make im­passe chains get ad­dressed (ei­ther solve ev­ery im­passe or agree to stop) in a rea­son­able amount of time.

Im­passe chains of­ten get repet­i­tive. Prob­lems re­oc­cur. Sup­pose I think you keep say­ing non se­quiturs. We can’t dis­cuss the origi­nal topic be­cause of that. So then we try to dis­cuss that im­passe. What hap­pens? More non se­quiturs (at least from my point of view)!

Some dis­cus­sion prob­lems won’t af­fect meta lev­els but some are more generic and will. You can try to say “OK given our dis­agree­ments about X, Y and Z, in­clud­ing method­ol­ogy dis­agree­ments, what can we still do to con­tinue prob­lem solv­ing which is neu­tral – which makes sense re­gard­less of what’s cor­rect about those open is­sues, and makes sense from both of our points of view?” Often what hap­pens is you can’t think of any­thing. It’s hard. Oh well. Then you can mu­tu­ally agree to end the dis­cus­sion since nei­ther of you knows a good way to con­tinue.

When im­passe chains get long, you tend to ei­ther have a lot of is­sues that are be­ing set aside (given A, B, C, D, E, F, G, H are un­re­solved, what can we do?) or you have a lot of rep­e­ti­tion (ev­ery meta level is the same prob­lem, e.g. my be­lief that you’re replies are non se­quiturs). Rep­e­ti­tion is it­self a rea­son to end the dis­cus­sion. It’s just re­peat­ing and nei­ther of us knows a way to fix it, so we can agree to stop.

This kind of end­ing is satis­fy­ing in a way. It gives trans­parency about why the dis­cus­sion ended. It means (worst case sce­nario) I’ve got­ten to make my case about what you’re do­ing wrong, and you’ve failed to an­swer it sub­stan­tively, so now I’m rea­son­ably con­tent (not my fa­vorite out­come, but what more would I hope to gain from con­tin­u­ing?).

So a policy can be e.g. to dis­cuss un­til an im­passe chain with the same prob­lem 3 times in a row. Or to dis­cuss un­til an im­passe chain of at least length 5. Gen­er­ally 5 is plenty. Reach­ing the 5th metal level can be quite fast and clar­ify­ing.

Im­passes can be de­clared in bad faith. Peo­ple can dis­agree about what is an im­passe? Then what? Dis­cus­sions have to pro­ceed in ways that make sense to all par­ties. If some­one thinks there is an im­passe, then there is one, even if the im­passe con­sists of his mis­con­cep­tion. And yes bad faith is pos­si­ble. What can be done about that? Trans­parency. Ex­pos­ing it to daylight. Hav­ing sys­tems like this that make bad faith more visi­ble and eas­ier to crit­i­cize. Hav­ing pro­ce­dures that cre­ate more ev­i­dence of bad faith.

In gen­eral, by an im­passe chain of length 5, if one per­son is be­ing ra­tio­nal and the other isn’t, it gets re­ally ob­vi­ous who is who. This gives the ra­tio­nal guy rea­son to be satis­fied with end­ing the dis­cus­sion (he knows what hap­pened and why, and he had some chances to try to solve it) and it gives ev­i­dence about both par­ties. If both peo­ple are fairly equal in skill or ra­tio­nal­ity, or both are pretty bad (even if un­equally bad), then you can much more eas­ily have muddy wa­ters af­ter an im­passe chain of length 5. Oh well. Creat­ing clearer im­passe chains is a skill you can work on. You can learn from your at­tempts to cre­ate some clar­ity and what didn’t work and why, and try to do bet­ter next time. And you can try to learn from the other guy’s at­tempts to.

The im­passe chain sys­tem is un­nec­es­sary for ev­ery dis­cus­sion. It’s a bit heavy­weight and high trans­ac­tion cost to use all the time. But it’s pretty limited. If you agree to a 5 im­passe chain, you’re always 5 mes­sages away from get­ting out of the dis­cus­sion. The only rea­son it’d take more is if the other guy said rea­son­able stuff. But if he says un­rea­son­able stuff, you’re done in 5 mes­sages, and some of those mes­sages will of­ten be just one para­graph or even one sen­tence long.

This ap­proach is good when peo­ple want to claim they are open to de­bate and that their views have stood up to de­bate. That leads to dis­putes over what it means to be open to de­bate, etc. I pro­pose be­ing will­ing to en­ter into dis­cus­sions ter­mi­nated by a length 5 im­passe chain (or mu­tual agree­ment) as a rea­son­able crite­rion for a (self-de­clared) se­ri­ous in­tel­lec­tual to say he’s ac­tu­ally open to sub­stan­tive de­bate about some­thing and is ac­tu­ally ad­dress­ing crit­ics.

And the im­passe chain ap­proach can be re­quested when you want to have a dis­cus­sion if and only if there will be a sub­stan­tial end­ing to pro­tect your effort in­vest­ment. If you want to avoid a case of putting in a bunch of effort now and then the guy just leaves, you can ask for an im­passe chain pre­com­mit­ment or 5 part­ing mes­sage pre­com­mit­ment or other way to help pro­tect (and there­fore en­able) your en­ergy in­vest­ment into the dis­cus­sion.

Con­clud­ing Thoughts

What’s the typ­i­cal failure case look like? Joe is try­ing to have a ra­tio­nal dis­cus­sion and Bob says “eh, your mes­sages are lame; bye” and won’t an­swer ques­tions or ad­dress ar­gu­ments. Or, worse, Bob ex­plains even less than that. If Bob would ex­plain that much, at least peo­ple could see “OK Bob ac­cused Joe of lame mes­sages and gave zero ar­gu­ments. There­fore Bob is lame.” Im­passe chains or even just part­ing mes­sages help en­able prob­lem solv­ing as well as clar­ify­ing what hap­pened in bad out­comes.

Often a dis­cus­sion looks like this: Joe writes a blog post. Bob says some crit­i­cism. Joe sees many flaws in the crit­i­cism. Joe ex­plains the flaws. Bob stops talk­ing. This isn’t satis­fy­ing for Joe. He never got feed­back on his post from post-mis­con­cep­tion Bob. And Bob prob­a­bly didn’t change his mind. And Bob didn’t even say what the out­come was. Or if Bob did change his mind about some­thing, it’s of­ten a par­tial change fol­lowed im­me­di­ately by like a “you win; bye”. Peo­ple rou­tinely use con­ced­ing as a way to end dis­cus­sions with no fol­lowups: no post mortem (learn­ing about why the er­ror hap­pened and how to fix the un­der­ly­ing cause), no work­ing out the con­se­quences of the right ideas, etc. The cor­rec­tion doesn’t go any­where.

That’s sad for Joe. He didn’t want to cor­rect Bob just for fun. He wanted to cor­rect Bob so it’d lead to some­thing more di­rectly benefi­cial to Joe. E.g. Joe’s cor­rec­tion could be crit­i­cized and that’d have value for Joe (he learns he was ac­tu­ally wrong in some way). Or Joe cor­rects Bob and then it leads to fur­ther dis­cus­sion that’s valuable to Joe. If cor­rect­ing peo­ple is pure char­ity – and you usu­ally get ghosted with­out them ad­mit­ting they were cor­rected – then peo­ple will even try to do it way less. There should be re­wards like some sort of re­s­olu­tion to the is­sues and con­tinu­a­tion. Dis­cuss pro­duc­tively and keep go­ing (and maybe Joe learns some­thing later or, failing that, at least gets a good stu­dent who learns a bunch of things and may be able to sug­gest good ideas in the fu­ture), or say the im­passe for why it’s not pro­duc­tive.

Often Bob thinks Joe is do­ing some­thing wrong in the dis­cus­sion but won’t ex­plain it enough for Joe to have a rea­son­able op­por­tu­nity to learn bet­ter. Note that cites are fine. If it’s already ex­plained some­where, link it. Just take re­spon­si­bil­ity for the ma­te­rial you re­fer peo­ple to: you’re us­ing it as our proxy, to speak for you, so you ought to have similar will­ing­ness to ad­dress ques­tions and crit­i­cisms as if you’d writ­ten it your­self (but if it’s pop­u­lar stuff with lots of ex­ist­ing dis­cus­sion, then you can ad­dress FAQs by refer­ring the guy to the FAQ, can di­rect him to the fo­rum for that com­mu­nity to get ques­tions an­swered and only an­swer them your­self if the fo­rum won’t an­swer, and you can link other blog posts, books, pa­pers, etc. to ad­dress fol­lowup is­sues if those ex­ist, etc.)

Im­passe chains help ad­dress these prob­lems and help make it harder to end dis­cus­sions due to your own er­ror and bias. And they provide op­por­tu­ni­ties to solve dis­cus­sion prob­lems in­stead of just giv­ing up at the first prob­lem, or in the al­ter­na­tive at least more trans­parency about the prob­lems can be achieved.

See also

My prior ar­ti­cle on Im­passe Chains.

My ar­ti­cles on Paths For­ward (about dis­cussing in such a way that if you’re wrong and any­one knows it and is will­ing to tell you, you never block that off with no way for your er­ror to be cor­rected), in­clud­ing the ar­ti­cle Us­ing In­tel­lec­tual Pro­cesses to Com­bat Bias.

My de­bate policy.