Reactions count as likes?

I made the silly assumption that other reactions were treated as likes, because I’ve only been testing positive ones. :man_shrugging:

Good idea @volanar about positive and negative reactions.

Or maybe even simpler: Some reactions can be flagged as negative and are excluded from calculations. The other reactions are treated like likes - as in my eyes there is no difference.

Is there any progress for this topic? I think that many Discourse communities which use reactions see more reactions used over likes. So the calculations or trust levels will be heavily affected already.

2 Likes

Yes, it is indeed a good idea. Let’s look at Understanding Discourse Trust Levels. Among the credentials for being admitted to and for maintaining membership in Trust Level 3 – Regular is this:

  • Must have received 20 likes, and given 30 likes.*

… that is within the past 100 days.

Receiving likes does reflect well on the recipient. Granting 30 likes during the prescribed period could also reasonably be considered to constitute constructive engagement in the community, but that might be a bit tricky if the grantor covets that gold-level Crazy in Love badge. On the other hand, receiving negative reactions, if counted, could point toward holding down that user’s trust level.

However, what can reliably be regarded as a negative reaction? For example, I personally would be quite reluctant to aim an angry reaction at another user. If I were, hypothetically, to use such a reaction, it would likely be to express agreement with the author of a post that a situation described in that post is a vexing one, for example one where the user identified rampant misuse of a large language model (LLM) to generate posts on behalf of a user, so that the user may quickly gain credentials. In that context, the angry reaction would rightly be considered a positive one, appreciative of the author’s having identified a problem that needs attention. Some users other than myself might use negative reactions differently. Due to that ambiguity, I am currently inclined to avoid using them entirely.

Note the ongoing discussion Enhance About page website statistics. Perhaps we can consider how we can enhance the statistics feature that is the subject of that discussion in order to offer convenient access to information that might help us with the issue of this discussion.

1 Like

I agree. But negative votes can make messages dimmer, as less useful and to reduce visual noise.


I think that negative reactions should not lower the user’s trust level. It is better to influence his messages.
Reactions also indicate the value of the message. As a rule, negative and useless messages for the community have negative reactions.

I think all reactions should be counted equally. Just because those negative ones are used as sign of empathy quite often.

And the big question is are negative reactions sign of receiver’s or giver’s negativity?

Because we don’t know and there is no way to understand context automatically, we must take way of social media: action is action is action.

3 Likes

On the upside, the system of requiring the receipt of 20 likes, and the giving of 30 likes during the last 100 days for Trust Level 3 – Regular is simple, transparent, and avoids the ambiguity of some of the reactions. A downside is that it ignores those reactions, which constitute data that could reflect upon the quality of the concerned posts.

Would it be possible, and not overly cumbersome, for someone to quantify the number of likes and the number of reactions that have been applied on this forum since the time that reactions have become available as an option? If this can be done, it might help us judge whether ignoring reactions for the purpose of awarding trust levels is a significant issue.

There is the possibility of using AI to apply sentiment analysis to replies to posts in cases where the person who replied has also invoked a reaction that might otherwise be seen as ambiguous. But my sense is that this might entail a large amount of work that delivers only a small amount of benefit. In addition, sentiment analysis inevitably would incorporate some systematic cultural bias into the process of awarding trust levels. This is also true of assigning numerical values to reactions.

So, the above leaves me inclined to wait for itemized statistics on the use of reactions before my forming an opinion. That brings us back to the issue of the discussion, Enhance About page website statistics. Implementing the proposed feature could help us here.

There is another route too: we don’t count reactions/likes toward trustlevel. I choosed that way — but sure, my situation is different what perhaps most of the world is living in (with? Damn those prepositions are confusing).

I don’t have spam problem. My forum is low traffic. So I can give TL1 right away. I don’t need higher than TL2 because it is an incentive, hiding ads, and one gets it via creating content, not spreading smileys. And higher trustlevels — well, moderation is not users’ job.

I know. That is not reality for all and there is forums where reactions are needed to prove ability to behave.

But my vague point is thay every admin should think a bit deeper what is actually needed and what is just… hype. And I’m not totally sure if counting reactions and/or likes proves anything. It is pure situation easy given quantity goes over quality.

And yet I would like to see reactions counted as likes. Just because if a forum uses likes as a metric for trustlevel, and users are using reactions instead of likes, and that’s why users may loose earned trustlevel, what signal we are giving then?

So, I see only two options:

  • we don’t count any kind reactions at all
  • we count every reactions (equally)
1 Like

This all seems too complex. Why not merely have the thumbs up and down emojis demonstrate support or lack thereof respectively, and all other reactions be cosmetic? That would allow usage of all emojis to be implemented immediately without any consideration of its effect upon this issue, and provide simpler UX to the users.

Because it breaks metrics of trustlevel. No other reasons.

1 Like

In my opinion it is not too easy to gain TL3. At least in our community, which is not as crowded as maybe here (where it looks like the default values are not in use).

Anyway, I read through the trust level concept and also thought about the default values there. And I think, they make sense for communities like the one I am running Discourse for. But they do not match anymore when reactions are active. So admins would need to check the percentage of likes and reactions and adjust the numbers. And readjust after some time and checking again the percentage. Also readjustment is needed when adding more reactions and so on. And, at the moment, SQL queries are needed, which leads to Enhance About page website statistics again. :wink:

So, I still cannot wait for having the reactions having their part in the trust level system and other parts of Discourse, where likes play a role. Optional for sure, but this is just a setting to turn on or leave it off and the traditional way. Also there should be some way to treat negative reactions, which I personally would keep as simple as possible (like flagging a reaction as negative and to choose if negative reactions are counted or not) - every community is a bit different.

Well, the trust level concept and also likes being a part of it was one of the main reasons for us to migrate to Discourse. So no, this would be the least I want. But as mentioned, every community is different and it is good to keep reactions out of it. :+1:

Fundamentally mixing likes (with impact in other parts of the site) and reactions (entirely cosmetic) has been confusing for users, in my experience. We added reactions because people wanted to show a variety of emotions. A big one is simply demonstrating care. So we have the :sob: reaction to demonstrate the reader agrees with the author about a sad situation. It’s not about expressing displeasure about the author or the quality of their writing. It’s the same instinct as pressing :+1: or :heart:, but with a different nuance based on the message.

If I were designing the system I’d either:

  1. Make every reaction work like a “like” or
  2. Move cosmetic reactions to a different part of the interface so that you can react and like a post.

The current system gives users the choice of either giving a bland/default response (thus helping them gain trust level) or the appropriate response (that’s just cosmetic). That’s assuming people even understand the difference (my users generally don’t). The current situation is far from ideal, but adding a downvote on the same interface will make things worse, in my opinion.

1 Like

I believe that this is the sole viable option to prevent a system like Facebook’s whereby certain reactions are contextually awarded to demonstrate negative reaction to the post, yet boost it algorithmically.

If modifying the GUI to demonstrate the difference between cosmetic and functional reactions were however implausible infeasible, I would propose that cosmetic reactions be deprecated. However, demonstrating a difference between them is immediately feasible, and appears to solve this problem. I believe that it should at least be temporarily implemented until and/or unless a more complex but superior solution is implemented.

I doubt that any other solution is superior, however - any UX for it would be needlessly complex.

Note, here, that the statistics provided indicate that the total count of all reactions, excluding likes, was about 7.2% of the total count of the likes plus the count of all reactions. So including reactions as likes without any weighting, for the purpose of awarding trust levels, would have a minor impact. Unfortunately, it would complicate the process and make it more opaque. Accordingly, I would lean somewhat against it.

1 Like

Well, you would need couple curves to show how using of reactions is growing. And is using likes coming down. Reactions here is relaticely new and users knows how TL-system works and like it.

So… could we have here a bit high bias :smirk:

2 Likes

If the trend continues, an adjustment to the current system might be needed. The appropriate threshold is admittedly a matter of judgement, so my leaning against it is only a small lean. :wink:

I just checked in our community for the ongoing month of November: The quota here is 94 likes vs 68 reactions. So that makes 42% reactions vs 58% likes. None of the reactions were negative (cry or angry). Also the other months show that negative reactions are barely used (8 for the total year 2023). November is a “normal” month, but I can also come up with the 2023 numbers if needed.

That’s the reason why I am not satisfied with not including the reactions into the trust level system or handling completly seperate to likes.

1 Like

Aha, so that hypothetical trend has already materialized, at least within some venues. 42% is a significant plurality. Accordingly, within your community, and perhaps in general, an adjustment to the current system of awarding trust levels does seem merited.

That is interesting. Presumably, you are referring to the Sidemount-Forum. Within a community such as that, which engages in exploration via diving, reactions to reports of spectacular marine and aquatic creatures and other discoveries are to be expected. Those reactions are likely to be highly positive, and it is understandable why you would want them to count. Negative reactions would be uncommon, except perhaps for occasional fright.

On the other hand, a different forum dealing with politics or other controversial topics might experience a significance of negative reactions, with some directed against particular opinions expressed, and with other negative reactions offered in agreement with a post, perhaps that some political figure has committed a misdeed (all too commonly :wink: ).

With the recognition that the meanings of particular reactions may vary between communities, it might be a good idea to enable the fine details of formulae for computing trust levels to be configurable.

1 Like

Yes, correct, I am talking about the Sidemount-Forum. :wink: And I also think that this community is different to many other which also are different to others. Perfectly normal. Even if divers in our forum can disagree a lot (e.g. about diving equipment, the way to use it and so on), it seems not to be very common to use negative reactions. I would think, that this would not be the case in a forum about politics, yes.

But that how we come back to a suggestion, which seems be pretty simple and which I suggested, too. And which makes a lot more sense for many communities than just ignoring reactions.

  • let the admin decide if reactions should be counted or not, default is “no”
  • if yes, should be differentiated between positive and other (negative) reactions - if yes, negative reactions would not count - like now
  • if yes, let the admin choose which reaction should not count.

That’s it. :wink:

/EDIT: I did not mention the most important part: Without counting reactions it is worse for many forums than counting and not be 100% accurate or well thought out. On the other side, reactions bring so much more into many communities, so it does not make sense to disable reactions to have 100% accurate like counters. I hope I was able to explain what my thoughts are about this.

If the user’s content receives negative reactions, then such content is not liked by the community or it is useless. If it is impossible to block a user with bad content, since he has not violated the Rules, then you need to influence the user or his content.
With the help of reactions, community members will be able to manage content and lower or increase user points depending on its value. And that’s right :slightly_smiling_face:

1 Like