Experiment: Comment on this thread whenever you notice that you wanted to respond to a comment with something that could be encapsulated with a simple react, by adding a comment with that react.
Top level comments in this subthread should be the simplest possible expression of a reaction that conveys the right nuances. (Child comments can have fully fledged conversations)
Only add things here if a real comment that actually happened led you to wish you could do a simple react to it.
"That clarifies it to my satisfaction, thanks for explaining."
(I often see something like 👌 or ✔️being used for this on Discord/Slack, though those are more about acknowledging the explanation rather than the expression of gratitude.)
If I'm deciding whether to post a comment, and my worry is an impact on my long term karma versus how many little dopamine hits I'll get from reactions, that feels like exactly the types of questions I want to avoid in my life. If either of these things is driving my decision, rather than what would help build knowledge or be useful to myself and others, then I'd consider it time to pack it in and stop posting entirely.
Reactions will raise the toxicity and blight levels of LessWrong. Non-anonymous ones would raise it more.
I have large uncertainty about the magnitude of these effects.
But I do know that I try very, very hard to never use reacts on social media to (among other reasons) avoid there being information in my failure to react to things or any pressure to see things in order to react to them.
There's a fairly weird idea that's possible here (I have a similar sense for Shortform Feed).
LessWrong has a fundamental problem where I want the top contributors to easily be able to write short, off the cuff stuff that fleshes out their thoughts in realtime without having to stress about getting nit-picked. I want top contributors to lower their standards for engagement. I don't necessarily want newcomers to lower their standards of engagement.
A possibility for both reactions, and shortform feed, is for them to only be available to people with N karma. A lot of the blight I associate with social media has to do with large numbers of people smashing into each other without sharing context. If the number of people who receive nudges towards low-effort interactions is fairly small and high quality, it might be better than giving those nudges to everyone, or to no one.
The equation I see is something like:
default-LW-world – tribal alliances and overton fights and other things I associate with toxicity are channeled into upvotes/downvotes, and comment storms.
Reaction-World – seems like it'd spend that energy differently. I think naively implementing it would increase it's visibility, but also channel a lot of the energy into activities that don't gain you karma. I also think implementing it carefully can force people to notice and pay more attention to what they're doing.
A thing that I actually like about Facebook "Love" reacts is that... they come with my name attached, and it says something about me how freely I'm willing to use it. It's use doesn't need to be limited by explicit rules, it's limited by vague self-regulating social norms.
I don't clearly see why Reacts should be dramatically different from commenting – if you'd have been willing to write a comment (except that it was too much effort), why not use a React? I do get that the there's a high level change in the culture when some actions being easier. But this seems like it moves the world closer to in-person communicatio...
I have never used them in team slack or discord, and also haven't been tempted to do so. I mean, I can just type stuff, that's what I'm there for, and we already had emoticons, so I don't really see the point?
My experience in other circles with Slack and Discord is that the niche of emoji reactions is primarily non-interrupting room-sensing (there are also sillier uses in casual social contexts, but they don't seem relevant here). I don't feel any pressure to specifically have read something, and I haven't observed people reading anything into failure to provide a reaction. The rare exception to the latter is when there's clearly an active conversation going on that someone's already clearly been active in, which can be handled by explicitly signaling departure, which was a norm in those circumstances anyway.
Non-interrupting room-sensing in a fast-flowing channel environment has generally struck me as beneficial. Being able to quickly find the topic-flow of the current conversation is important, and reactions do not have to be scanned for topic introductions. Reactions encode leafness: you can't reply to a reaction easily, which also means giving a reaction cannot induce social pressure to reply to it. They encode weaker ties to the individual: people with the same reaction are stacked together, and it takes an extra effort to look at the list of reacting users. Differentially, reaction
My experience from seeing emoticons used on Slack/Discord is that they help combat the muted signal problem of online communication, and thus actually reduce the toxicity of discussion.
People want to feel respected, loved, appreciated, etc. When we interact physically, you can easily experience subtle forms of these feelings. For instance, even if you just hang out in the same physical space with a bunch of other people and don’t really interact with them, you often get some positive feelings regardless. Just the fact that other people are comfortable having you around, is a subtle signal that you belong and are accepted.
Similarly, if you’re physically in the same space with someone, there are a lot of subtle nonverbal things that people can do to signal interest and respect. Meeting each other’s gaze, nodding or making small encouraging noises when somebody is talking, generally giving people your attention. This kind of thing tends to happen automatically when we are in each other’s physical presence.
Online, most of these messages are gone: a thousand people might read your message, but if nobody reacts to it, then you don’t get any signal indicating that you were seen. Even gett...
Reading a message takes more cognitive resources then seeing a single emoticon. It's easy for people to click +1 to get an opinion of what multiple people think then when every person lays out "I agree with you", "I think we should adopt your proposal" etc.
Team slack is the domain that most clearly motivated me to consider reacts for LW. They also especially made me long for them in google docs.
Emoticons etc still take up a full line of text, and easily get mixed in with other discussions that are happening. Reacts allow for much higher information density, which allows the overall conversation to be more complex.
I am very much looking forward to a low effort way to give and receive more nuanced feedback! Hope you guys come up with a test version soon.
I don't think I've seen this point made in the discussion so far, so I'll note it here: Anonymous downvotes (without explanation) are frustrating, and I suspect that anonymous negative reacts would be even worse. It's one thing if someone downvotes a post I thought was great with no explanation -- trolls exist, maybe they just disagreed, whatever, nothing I can do but ignore it. If they leave an "unclear" react, I can't ignore that nearly as easily -- wait, which point was unclear? What are other people potentially missing that I meant to convey? Come back, anon!
(This doesn't overshadow the value of reacts, which I think would be positive on the whole, but I'd love to see Slashdot-style encouragement for people to share their reasoning.)
If they leave an "unclear" react, I can't ignore that nearly as easily -- wait, which point was unclear? What are other people potentially missing that I meant to convey? Come back, anon!
Maybe there should be an option that allows you to highlight a part of the comment and react to that part in particular.
Brainstorm question: Are we sure this type of feedback needs/wants to be public? I see a mode where it would be helpful to know the reason, but where having the reason by default stamped onto posts is even more demotivating than not knowing.
Not sure how this interacts with possibly being non-anonymous.
My experience is reactions are important for real time conversations with too many people at once. It allows one person to speak and several people to agree without adding another line of text and clogging up the discussion.
There is another use case of "supportive" emojis where I would react hug to "I've had a rough day" from a friend of mine.
There's all the humour uses of emoji too but that's not what we want on lw.
How are you thinking about the time-value of such react tokens? Are you trying to fix a problem with votes, or to introduce a new mechanism for a purpose orthogonal to voting?
I'd like to see more signal in the voting: slashdot-style "why are you voting this way" would naively fit that, but I don't actually like any implementation I know of, so I may be wrong in my understanding of my preference on that front.
One of the things about the current "votes" mechanism that consumes my mental energy with no value is that they...
A lot of the benefit from reacts would be the ability to distinguish between "this comment makes the thread a little worse given constraints on attention and reading time" and "die, monster, you don't belong in this world". Downvotes are aversive because they come across as a mix of those two despite being mostly the former.
It occurs to me that I have somewhat different expectations and desires for reactions on Question posts, over regular Post.
On questions, it's usually a lot more clear which comments are making concrete-progress towards answering the question, and which are not. It also seems like there's more useful feedback to give people in terms of why an Answer was useful or un-useful. (and being able to see lots of micro-feedback might help people learn to write more useful answers)
Some things that I might want to react with include:
I believe that "like" and "dislike" are good choices, especially if you want people to make a lot of votes, without spending too much time thinking about it. Anything more complex, and most people will not use it; and if that means they cannot vote, then less people will vote (and the results of voting will represent a smaller set of people, mostly the compulsive voters). Time spent voting (not per one comment, but site-wide) is a limited resource.
I think that when websites try to measure more than one dimension, the usual outcome is th...
There's an interesting parallel between Slack reactions and the wiggle fingers applause from sign language. If we have reactions on LessWrong that can express more nuance, I think it would be great to have corresponding hand gestures that can be used for offline interaction within our communities.
Should we consider a mechanism to reduce conformity bias? For example, we could allow users to blind themselves to (the nature of) existing reactions until they choose to reveal them or react themselves.
Such a mechanism may come with its own drawbacks, of course. And it's possible I'm just overthinking this. But I hadn't seen the idea discussed yet, so I thought I'd bring it up.
Facebook-style reactions typically involve a limited set of predefined reactions that users can choose from, such as like, love, or angry. discord-style reactions, on the other hand, offer a wider range of reactions, potentially with the ability for users to customize and add their own. Slashdot-style reactions focus on providing the option to clarify the reasons behind an upvote or downvote.
You express your enthusiasm for the Discord-style reactions, as they allow for more flexibility while still having commonly used default reactions. You believe that any of these styles, if implemented effectively, would be an improvement to the platform
For the past year I've wanted LessWrong to include something like Discord, Facebook or Slashdot style reactions.
Facebook Style means "there's a few key reactions that people use"
Discord Style means "there's nigh-infinite reactions and you can add more, but there still end up being a few commonly used defaults."
Slashdot Style means "after upvoting or downvoting, you have the option of clicking a button that clarifies why you upvoted or downvoted."
Of these, I'm most excited for Discord-Style. But I think any of them would be improvements (if done well)
Habryka recently wrote a shortform comment on this subject. My own thoughts come in a few different frames.
Boos/Yays vs 'approve/disapprove'
Empirically, people want to cheer for their causes, boo causes they dislike, signal their social allegiance and try to ensure the overton window moves in the direction they want. I don't think you can really fight this. But you can nudge people to disentangle this from "what gets attentional allocation on a site about rationality."
I think it's important that when you see a comment you like, and you feel the impulse to go "yeah! good point! go team!" the first impulse you have, the first button available and exciting to click, is a button that doesn't send any signals about how that comment should be sorted, and doesn't aggregate into an overall user-score you can check (that, for good or for ill, people will tend to associate with social status)
Other things vs 'approve/disapprove'
Boos/yays aren't the only thing I'm worried about. Ideally, I want LessWrong to reward good thinking over things like being funny, or exciting. (Being funny and exciting should still get rewarded, but no amount of clever injokes should add up to something greater than "wrote an actually useful, insightful point.")
"Viscerally Fun but Low Signal Buttons" should be easy to access. "Higher Signal" buttons should require more effort and thought.
With both of the above in mind, I think it's important that "Yay", or "Funny" buttons should be the first, most obvious thing to click on. They should feel satisfying to click, and you shouldn't feel motivated to click more things if that's the only reason you were upvoting.
The buttons that send more important signals should require a bit of extra effort, and force you to at least notice some cognitive dissonance if you're upvoting people just because they're on your side.
One react someone expressed interested in was a simple "acknowledged." Votes are totally anonymous, and that means if you want someone to know that you have read a thing, you have to actually comment, which is moderately high effort and takes up a lot of vertical space on the page. Whether someone has read a thing is fairly important information about how to continue a conversation.
By default, on many social-media platforms, likes are public. They were also public on the old Intelligent Agent Foundations Forum (and I think probably on Arbital, although not sure offhand).
This does two things, which I have mixed feelings about.
One is social entanglement. Visibly liking each other's comments is part of the process by which people build social trust and alliances. I think there's reason to be cautious about LessWrong directly facilitating that.
Another is clarity on who believes what, and whose judgment you trust. When you're building a serious, complex idea, it's actually important who understands what concepts, who thinks different concepts are important. There are people I do in fact trust more intellectually than others, and it's higher signal to know that one of them liked a post than some rando. It's also more informative when I know that multiple people I trust disagree.
My current best guess is that it's best for the voting on LessWrong to be anonymous, but for reactions to display usernames on hover-over. It might or might not be feasible or desirable (from a UI complexity standpoint) to let people choose whether to react publicly. But I can imagine changing my mind about this.
Receiving a downvote without explanation sucks. Some people complain about this – "can't you provide reasons for your downvotes?" Well, no. Trivial inconveniences matter. If you force people to provide information and figure out how to articulate what's wrong with something, people will probably just stop giving feedback rather than actually providing reasons.
Not only does this require figuring out how to write a comment, it opens up a line of engagement that you might have to put even more effort into defending.
[this is an empirical claim, it's perhaps worth the experiment of requiring downvotes to always require reasons, but I'm not optimistic about it].
But I think there are some fairly common reasons why a comment gets downvoted, that could at least make it lower-effort to give feedback:
It's also nice to improve the reward signal for particularly good actions:
You'll notice some issues, comparing the above feedbacks to Facebook Reacts.
Facebook reacts are "haha!" "love!" "sad!" "anger!" "wow!"
Everyone knows what those mean. Everyone knows that everyone else knows what those mean. They are very short words. They are (due to millennia of evolution, genetic and cultural) conceptually simple.
"This comment seems to be rounding things off in an oversimplified way" is a less common concept. It's more complicated. And if you simplified it slightly so that the button said "Oversimplified"... that would... actually be an oversimplified button. It's important that I'm just saying "yo this comment was oversimplified", but rather that it seemed (probably) to be making a subtle error.
I think this is really important. I think something LessWrong needs to do is nuanced critiques easier to chunk. This is pretty tricky, since, well, the whole point of nuances is that they're nuanced.
A rationalist friend once commented, in non-rationalist circles, that when they tried to say "I agree with your point but I think this particular part has a logical error", they would often have people... just completely fail to parse that. It wasn't in their schema at all.
On LessWrong, we have some shared context where we mostly all understand not to just have Arguments Be Soldiers and whatnot. Our schema includes Local Validity. But there are many important, key concepts that still take a lot more effort to express than "yay/boo" or "haha!"
And thing is... it's not like "Love" is a simple concept. When someone clicks 'Love' on one of my facebook posts, there is a fairly rich wave of senses I get (depending on my post, and depending on my relationship with the person in question). When someone posts about their pet dying and I click 'Love', there's this whole shared context about how we're both human and we know what it is to lose people and my heart goes out to them and I chest tenses slightly and there's... just a whole lot going on.
Still, I'm able to chunk that complexity into a concept called "Love", and it's easily available for me to access.
There's a potential longterm vision for LessWrong – maybe not the right vision, but possible – where part of what we're doing here is distilling concepts down so thoroughly that a single word can communicate a lot of nuance.
Language real estate is limited, and I'm not sure which concepts make the most sense to distill in such a way. There's also certainly room for this to fail, where instead of being able to more-easily-express nuanced concepts it ends up destroying nuance.
Facebook has cheapened the word "friend", and that's important. But... I also have an impression of it having made it easier for me to express love, in a way that so far seems net positive.
It feels exciting to me to imagine one day living on a world where "this changed my mind" or "this was well thought even though I disagree" feel like basic, obvious concepts that are important enough to be communicated with a single word.