Designing feedback systems that encourage deliberation

Sorry for sounding too academic. Due to the lack of real life feedback systems like facial expressions, body language, and tone, online interactions are flatter than in-person human experiences. In large groups, nuance tends to get lost even more in favor of conformity. Once a post/reply gets its first few likes, other people feel safer liking it. If likes are neither anonymous nor retractable (after a grace period), there’s reason to believe the social pressure only gets worse.

This snowballing effect disproportionately rewards posts that are socially easy to approve of. This kind of positivity can come at the expense of nuance which is harder to reward because it is often emotionally neutral. I think it’s important that feedback systems are designed to be informative, rather than overly formative. Moreover, since readers will skip reading replies to save time, or give more weight to replies based on likes, it’s important they stay grounded in truth and not only emotionally compelling.

Some ideas I came up with are:

  1. Hide likes for 24-48 hours on new posts/replies. This is often the phase it gets the most views, and people react the quickest. Hiding likes is probably good for mental health and encourages reflection in personal judgements. It’s still good to not obfuscate them permanently so people can look back on the consensus retrospectively though.

  2. Hearts and “likes” may be symbolically too strong/personal and lead to polarization. Emotionally-inert terminology like “respect” could make people more willing to acknowledge counter-attitudinal comments. Furthermore, extra metrics on top of likes such as “thanks” “helpful” “informative” “skeptical” could also be practical. I suspect that when there are multiple metrics, and people see that metrics disagree, the public record will capture more nuance, so the system will stop optimizing for a single metric only. The symbols could be like a handshake, lifebuoy, lightbulb, hammer&wrench, nodding face, bowing figure.

  3. “Inline caveats”. Slightly avante-garde, maybe you could like a post/reply, but highlight a sentence, and press “skeptical” on that only, to indicate overall positive on the post, but a more unique reaction to a specific part, so the feedback is even more granular/nuanced and less overarching.

  4. When posts go off topic, self-moderated tangential debates should be able to coexist with replies to the original post, but be collapsed out of view and conducted in a healthier format. Maybe in the form of a “Start Discussion” or “Reply as New/Linked Topic” button. These should also be able to be collapsed heavily out of the timeline on a post, or redirect to a new post, so people who want to read on-topic replies can just pretend those 4 guys arguing are not there.

Here’s what I like about the Ziggit formula that I don’t want changed:

  1. Chronological order by default within a thread, as opposed to sorting by most liked
  2. Likes exist at all. I want there to be a feedback system, it’s useful for readers who are still forming an opinion on new topics
  3. Likes are not anonymous and can be tracked, accountability on positive feedback builds friendships

If you read this far, please let me know which ideas you like, which you hate, whether the platform is fine as is, or if there’s any suggestions you’d add.

14 Likes

Just one thing - replace hearts with more suitable emoji

How to …

As far as I understand it’s allowed for forum admins

8 Likes

Note that this already sort of works and is common practice in that when you ping the mods here on the forum, they can split off-topic/tangential debates to their own threads. Discourse (the software this forum runs on) supports this nicely and provides bi-directional links (both from the new thread to the original one and vice versa).


Besides that, I think these points are worth considering. It all also depends on what the Discourse software allows for, of course.

3 Likes

Would it be possible to make it fully self-moderated, without pinging the mods or is that a limitation of Discourse? I didn’t even know that was a thing you could do, would make sense to streamline it imo.

I’m not sure how desirable it would be to just let anyone do that. I imagine it could potentially create quite a bit of chaos if people could do that without the mods’ discretion. Discoverability of the feature is of course slightly problematic, but the mods here do monitor the forum actively (it’s not that high-traffic, after all), so I don’t think it’s that much of a problem? If anything, people are being encouraged to flag things that need the mods’ attention.

“likes” are hard. I sometimes balanced between liking someone contribution to the discussion, but also disagreeing with the particular proposition they made.

Similarly someone can make a good technical point while being an asshole and I can’t like that.
I don’t know if discourse supports this but having two separate reactions for this could be great.

Any system can be abuse eg: tagging people you technically disagree with as asshole.

As for the likes counter, I feel a bit silly to get notifications for likes “you got 2 likes !”. But they can also be good signal for core team to gauge interest for specific issue or proposition.

4 Likes

If anyone can create their own post, then why can’t they create their own linked discussion within a post? I could do the exact same thing right now, but manually with extra steps. You’re not giving me more power, you’re adding a convenience feature for everyone.

Completely agree with you, any system can be abused, but if you got “2 likes, 3 respects, 6 informatives” that would give you much more useful info than “7 likes”, and they’re both equally abusable. I think it’s an improvement over what it is now surely.

2 Likes

I’m possibly conflating your intent with how it works right now, but when tangential discussions get moved around, it’s usually multiple comments at once being split out into a separate thread. If everyone could do that, they’d not only be moving their own comments, but also comments made by other people. There’d need to be at least some consent system for that, I think.

I find 1, 2, 3 interesting and potentially helpful. I’d be up for it.

3 Likes

I have not noticed hearts/likes being a problem here on Ziggit. I believe that the best feedback is provided by words. You can reply to the person and explain what you agree or disagree with.

6 Likes

I feel like this might be more suited for https://meta.discourse.org/ rather than Ziggit, as this is more about how to build discussions, rather than how to build software with Zig.

That being said, your post was very helpful to me, as I’ve been wondering about these topics idly, and I am looking at what we know about this things. I’ll read some of the linked papers for sure , but, I am wondering, do you have some survey articles or books to recommend on the topic? I feel I like I can learn a lot from you, <3 :smiley:

5 Likes

I think we can leave retroactive splitting up threads with multiple users to the mods. I was more so suggesting the ability to start a new collapsable micro topic within a topic, or a linked topic from a topic.

a new linked topic from a topic is built into discourse today!

personally i think likes are ultimately fine; the problem is social and probably cannot be entirely avoided. even if likes were removed another version of this problem would arise.

Really? I thought the “Site Feedback” category was about this website’s design my bad :sweat_smile:. Does Discourse not have features/configs/plugins to let the site owners hide likes for X hours, change the like symbol, add more symbols & counters, etc.? I’ve never used it before so I don’t know much about it.

Also you flatter me too much, I’m not an expert on this. Most of the papers I linked are surveys themselves, which I read only because I got curious about the subject. These suggestions are based on my personal experiences using social media/forums, and the research I could find on them. I don’t know any books on this, unfortunately :frowning: .

I think the topic is too new, broad, and not experimented enough with for there to be any definitive conclusions saying stuff like “this is objectively better for a healthier community/forum/app”. What I wrote was based on mechanistic reasoning, with a few empirical studies to support them.

Completely agree that these social issues are unavoidable, but I think they could be mitigated, and feel that the current model is rather incentivizing in the exact opposite direction.

Also how would I make a new linked topic here? When I click “Reply” it only replies within the thread, there’s no option for me to create a new linked topic/micro topic.

You can copy the link of the post, create a new topic and paste it in then write your text about how your topic is related / was spawned from this comment.

When it is only a part of another comment you can select it, press “copy quote” and paste it into your new topic.

Once you create that post discourse will automatically create a link from the quoted post to the newly created topic.

1 Like

better than this, if you click on the reply icon within the reply UI, you get options.

1 Like

I don’t know what you mean.

also works on mobile.

8 Likes