A new trust level: The Helpful member?


(KajMagnus) #1

Hi,

I’m thinking Discourse forums might benefit from one more trust level, namely the Helpful member or Considerate member. (Considerate means like kind & helpful.)

I’m thinking: Some people might visit the forum infrequently, and post helpful answers and stuff the community likes, and cast helpful flags. However, they are short of time and don’t visit often. They won’t become Regular members. But when you observe them over a long period of time (like a year?), it becomes clear that they are indeed kind and helpful people you can trust, and it could make sense to give them their own trust level and more power to help keep the forum a good place.

Now you might ask: Why have both Considerate and Regular members? Well, the difference is that the Regular members visit regularly, and thus are up-to-date with everything that happens in the forum. But the Considerate members don’t visit regularly, so they might not know what’s going on right now. And, since the Regular members are up-to-date with everything that’s happening right now, they could/should have a bit more power than the Considerate members, because the Regulars are in a better position to make goo decisions. (And only regular members would get access to the “frequent flier’s lounge”.)

A regular member that leaves and returns a year later, becomes a Considerate member (and thus loses some of his/her super powers, but not all of it).

(The reason I’m posting this topic, is actually not that I’m hoping/suggesting that anyone will implement it. Instead, I’m about to reuse Discourse’s trust levels in a project of mine, and I’m thinking about including this additional trust level too, and thus wonder what others think about that trust level. Plus, I think this discussion can be useful for the Discourse community — if it turns out Discourse could actually benefit from that “missing” trust level. Namely the kind-and-helpful members that visit infrequently only.)

(“Considerate” could perhaps be changed to “Helpful” or “Well-behaved” or “Kind” or …)

Anyone interested in this? What are your thoughts? Would you want this new trust level in Discourse?


(Andrew Waugh) #2

I think I know what you are getting at.

Our forum has a few of these, typically they use the email digest to see what is going on, but don’t post all that often. When they do, it is useful and accurate content.

We also have what you might call “magpies”. They are very active, comment on damn near every topic which appears, quote and distribute likes as if there is no tomorrow, but other than causing thread drift, they don’t actually do much.

I would say just bump your helpful members to the higher trust level.

The magpies… well, that is a different issue.


(KajMagnus) #3

Ok, yes that’s a good idea and that would work. At the same time, it’d take time for me to do that — I need to try to remember these fairly infrequently visiting people (which might be hard since they don’t visit so often and I’m afraid I’m a bit forgetful). And I would spend time manually looking at their statistics and perhaps comment history — time I could have spent doing other suff. Also, I’m a human and inherently unfair. I might look at someone’s photo and think “yes clearly a helpful member”, then I look at someone else’s photo, someone with the same statistics & behavior, but conclude “obviously shouldn’t be promoted” :- / … Also I’ve read that after one has slept + eaten, one is a lot more likely to judge other people favorably, in comparison to at the end of the working day (hungry & tired). — So, me evaluating people = takes time & is unfair. The computer = … better, I hope.


(Andrew Waugh) #4

There is a human factor in the evaluation, both in the sense that the system can’t distinguish between magpies and useful contributors, and that we humans aren’t always objective.

On top of that, our forum includes a lot of content about which I am not qualified to judge. It is a car forum, and much of the discussion is about arcane details relating to models which I am not familiar with.

If anything, I would like to see something like moderated trustlevel promotions. I.e. the system looks at the activity levels and suggests “X looks like a contributor worthy of promotion, do you agree?”.


(Dave McClure) #5

Perhaps an automatic badge based on a well devised badge query could fill the gap here?


(KajMagnus) #6
  1. You mean that moderators would regularly review people who have recently gotten that special badge? That sounds like a good idea :- ) Then, a Google Calender reminder, say repeating once per month, + this badge, + manual promotion to a better trust level, would mostly address the “problem” this topic is about.

  2. Or can badges be associated with side effects? That’d be a bit cool, if there was e.g. an “Accurate-Flagger” badge, and when someone got that badge, his/her flags thereafter had a little bit more weight. — Then badges would really “mean” something.


(Dave McClure) #7

Option 1 is closer to what I meant.

Discourse does not offer fine grained permissions, so option 2 would not work.

However, things that could potentially be automated include:

  • raising their trust level to an existing trust level
  • adding them to a group
  • showing a special title and/or “flair” on their avatar
  • giving them access to a special category (e.g. the lounge)

Some of the above may be possible out of the box (I’m not sure). Otherwise, it could be automated via a plugin or something that sits aside Discourse using the API… or it could be done manually as you suggest.


(ljpp) #8

The current logic/algorithm may be putting too much emphasis on activity and could focus more on quality.

One thing that comes to my mind is likes vs time spent. The non-frequent yet insightful members might rank higher with this approach.

With the current approach the hyper active members will get to level 3, even if the value in their posts is not too great. Just by writing a lot one will bag enough likes to get promoted.

Edit: Or likes per post ratio.


(Dave McClure) #9

I hear you.

I think the current algorithm does a pretty good job of finding folks who are both active and high-quality. When coupled with quality, activity is something you want to reward or distinguish

But you can probably adjust the TL3 requirements in your site settings to make adjustments that fit better given your sense of things.
.
That said, to me your point reinforces the OP’s request. Maybe there would be some benefit to having a TL that sits between 2 and 3.

I used to be TL3 here, but I’m not as active anymore. I think it’s cool that I got “demoted” accordingly.

But maybe there would be value in distinguishing members like me from those who just crossed the TL2 threshold.

And looking at our own site, I could see the benefit in this too.

The badge system and the controls for TL3 requirements are good ways to adjust the defaults here, but defaults are powerful, and maybe there is something to the OP’s idea.

Maybe it’d be interesting to brainstorm a little together - If there were a TL2.5, how should it work? How would it be gained/lost? What additional permissions should it have by default?


#10

This started as a quick post on the mobile app but quickly became much bigger. Let me know if you don’t like the length. :slight_smile:

It seems like the right place to add a trust level

My paraphrase of the OP’s initial idea is that their could be a new trust level between trust levels (TL) 2 and 3. It is called TL2.5 for convenience.

That position is significant because TL0, TL1 and TL2 are achieved by activity without reference to a time period. Whereas TL3 requires a level of activity within a specific time frame. So TL2.5 makes sense in terms of its rank in the trust levels.

For each trust level I’ve summarised the main features released at each trust level and the promotion threshold which is initially automated and measured in days:
Automatic assignment
TL0 has limited posting features with new user restrictions
TL1 has core posting features after 10 minutes active over at least 1 day
TL2 has inviting others after 60 minutes active over 15 days
TL3 has basic topic moderation features after 50 days active over 100 days
Manual assignment
TL4 has complete topic moderation features after manual promotion

There are also staff roles which don’t have trust levels but I have assigned them numbers to denote the progression:
‘TL5’ Moderator can view user info and have no rate limits
‘TL6’ Admin can impersonate users, create user groups and change site settings

But the proposal subverts the trust level structure
Two measures are proposed to promote users to this new trust level:

  • their exemplary good behaviour such as kindness, helpfulness which over a longer period of time suggest trustworthiness
  • the value of the content they contribute.

The problem is that neither measure directly contributes to the stated purposes of the trust levels.

This proposal would subvert the stated purpose of most trust levels which relate to contributing time spent and gaining experience over time. This time spent/contributed and activity performed leads to a greater role in moderating the forum and hence increasing abilities for that purpose.

So why should users who show no inclination (or ability to find the time) then be assigned a higher trust level?

What additional features would TL2.5 provide?

I am a TL3 user at the moment but I don’t expect to retain it because it was only one well supported post that pushed me over the threshold. Normally, my likes received are somewhat below the threshold.

I’ve grouped the TL3 features into those that might be attractive to a helpful user and ranked them according to those I prefer:

Useful for TL2.5:

  • have all their links followed (we remove automatic nofollow) - useful for bloggers and others wanting traffic or reputation for their own web pages?
    
  • make their own posts wiki (that is, editable by any TL1+ users) - useful for authoritative topics but probably less likely to be posted by an infrequent visitor? Maybe allow their posts to be improved when they are not around to follow the topic?
    
  • access a private "frequent flier's lounge" category only visible to users at trust level 3 and higher - more useful to a regular visitor?
    
  • recategorize and rename topics - again, these users are less likely to create topics?
    

Not useful for TL2.5

  • spam flags cast on TL0 user posts immediately hide the post - unlikely to spend their time on this?
    
  • flags cast on TL0 user posts in sufficient diversity will auto-block the user and hide all their posts - ditto?
    
  • Daily like limit increased by 2 × - unlikely to hit the limit anyway
    

If there is no TL2.5 then what would most easily and usefully move to TL2? I think that making their posts wiki would be easiest to move and possibly the most useful. They could leave a post which others can amend and add to when they are not around.

So I would probably propose that the ability to make posts wiki should be moved from TL3 to TL2 instead of creating a TL2.5.

Other ideas about trust levels also suggest evaluating content value and good behaviour

There appears to be a common theme to these proposals: evaluate user behaviour and content. So maybe there is something in this.

Even so, adding the proposed trust level or these other attributes will muddy the simplicity and clarity of the current trust level structure. For this reason alone, I don’t support the proposal unless it comes with a comprehensive reformulation of the stated purposes of the trust levels or an alternative structure to support the new attributes.

That’s why it is worth looking at other discussions about the purpose of Defining trust levels . In that article there were a couple of perspectives that I appreciated which are somewhat similar to the OP here.

@ChrisHanel suggested something which has potential. And if others were to see it as useful then he might be the type of user that the OP is speaking about. In summary, he adds to moderation good behaviour and valuable content and provides a graphic example of how moderation and content could fit together. However, good behaviour doesn’t make it onto the chart.

He proposes:

  • your good behaviour
  • your value to the community - differentiating between two types of contribution:
    • the value of your management or moderation of the forum
    • the value of the content you contribute.

Continuing the discussion from Defining trust levels:

@frandallfarmer in Defining trust levels also suggests two measures of user value which can be likened to Chris Hanel’s additional measures:

  • evaluation quality which could be likened to good behaviour
  • content quality

Personally, I can’t see how content value can be used with the trust levels which emphasize moderation contributions.

The best moderators aren’t always kind and gentle. Likewise, it is easy to find people who are terse and judgemental with others and yet capable of well-drawn technical explanations.


Put TL4 in the "moderator/admin" category instead of trust level
(Erlend Sogge Heggen) #11

I think this is the main user group we’re failing to properly acknowledge. But rather than creating a new trust level, I’m leaning more towards changing the criteria of TL3. People’s schedules and interests are different, and we could be more accommodating of that.

If a user focuses on our #dev category and only checks in every month or so, they can still be an immensely positive force in the community, fully deserving of the trust and recognition that TL3 represents.

TL3 currently requires regular check-ins, but I’m not so sure that’s a fair condition. Here’s the current default requirements:


:pencil: Here’s my suggestion:

Regular (3)

Regulars are the most active readers and reliable contributors over long periods of time.

They can get to trust level 3 by…

This is time and activity based, unlike the previous levels. In the last 65 days… (a quarter)

  • a likes-to-posts ratio of 2:1
  • must have replied to at least 10 different topics
  • must have received 20 likes, and given 30 likes.*
  • must not have received more than 5 spam or offensive flags (with unique posts and unique users for each, confirmed by a moderator)
  • must not have been suspended

and one of the following:

  • of topics created in the last 100 days, must have viewed 25% (capped at 500)
  • of posts created in the last 100 days, must have read 25% (capped at 20k)
  • must have viewed (or replied to) 25% of all topics in any given category (minimum set to 20, capped at 500)

* These likes must be across a minimum number of different users (1/5 the number), across a minimum number of different days (1/4 the number). Likes cannot be from PMs.

All of these must be true.

Unlike other trust levels, you can lose trust level 3 status. If you dip below these requirements, whether from flagging or the passage of time, you will be demoted to Member. However, in order to avoid constant promotion/demotion situations, there is a 2-week grace period immediately after gaining Trust Level 3 during which you will not be demoted.


The purging of inactive TL3 users makes most sense in relation with our #lounge category, which is supposed to be this special hangout space for the most avid members of your community. But in reality we’ve hardly seen any communities create this type of meta-community within itself.

What we should do about the #lounge category is a different topic, but what I’m getting at is I don’t see the point of demoting TL3s. I’ve seen several discussions of users finding the automated demotion stressful; I hate the idea of our tool causing people stress.

As an admin, I’d much rather:

  • Receive a notification when a new user reaches TL3
  • Receive a notification when an existing TL3 is no longer fitting the criteria

…and leave it up to me as an admin to demote that user if they’re no longer bringing value to the community.

This is something we’ve considered adding to the soon-to-be-core admin statistics report.


(Jeff Atwood) #12

I am absolutely opposed to any sweeping tl3 changes but tweaks to make tl4 work better seem sensible.

Also 25% of any category leaves many holes; for small categories this would be a backdoor.


(Erlend Sogge Heggen) #13

Hence the minimum condition:

Aside: do we have any way to output a list of users who were at some point in time a TL3 user on Meta?


(Mittineague) #14

I may be missing something, but it seems to me that a Staff member could do a “hey, this members is great” and lock their Trust Level at 3 to by-pass any auto-demotion from happening.

A bit more manual work, but easier than trying to come up with an algorithm to let code do it.


(Neil Lalonde) #15

We create a record in the user_histories table when trust level changes, so we could find anyone who had TL3 at some point.

Some of the discussed options are already possible by changing the many site settings.

  • “tl3 time period” can create a much larger window of activity
  • “tl3 promotion min duration” can be increased to give a higher grace period before demotion to TL2 happens. If you don’t want demotion to happen, set this really high!
  • “tl3 requires days visited” can be more forgiving to people who follow activity from their email inboxes
  • “tl3 requires likes given/received” can increase requirement for quality
  • etc…

There are probably factors that can’t be controlled, so we can try to find what those are.

I think the automatic loss of TL3 is important, especially when conditions like receiving flags should cause the demotion. And it’s easy to change. Setting “tl3 promotion min duration” to 365 would give people a full year of TL3 without worrying about demotion, for example.

I like your ideas @KajMagnus because I follow a few forums (Pebble, Gearbox, Sublime Text) via the summary emails and watched category notifications, but I’m not convinced that I should be granted special privileges above what TL2 gets that other TL3s on the site deserve. For example, if I want to recategorize a topic, I could send a PM to a moderator or a TL3 person to have it done. If they know me as a really helpful and trusted person on the site, they’ll probably do it. It could be formalized with a badge or group membership with some flair on my avatar, but I dunno about a TL2.5… Hmm.


(Sander Datema) #16

I like the idea of this thread and it touches my own experience. I’ve been TL3 a couple of times here on Meta, but lost it every time again because I wasn’t online enough, but I daresay I did add something to the community.


(KajMagnus) #17

Hi again (5 days later),

Thanks everyone for an interesting discussion & insightful ideas. I think I personally am fairly much done with this topic … below, I’d just like to write something about my personal conclusions. And then reply to individual people.

Ok, after reading the discussion here, it seems to me a TL 2.5 makes sense if designing something from scratch.

But Discourse already exists & works fine, and I suppose this new trust level wouldn’t do so much useful, in comparison to how-long-it-takes-to-discuss-and-implement. I guess there are more important things on the roadmap. … But perhaps worth revisiting a TL 2.5, a few years later, when there aren’t so many other important things on the roadmap any longer :- ) (because they have all been handled).

Hmm, I’d like to recap what I see as the 2 benefits with a TL 2.5:

  1. Saves time for staff & Leaders & regulars — because TL 2.5 supposedly lets more trustworthy-&-good-judgement people help out with moderation.

  2. Might make these infrequently visiting, but trustworthy-&-good-judgement, members feel a little bit more appreciated — someone has thought about them and given them their own trust level.

***

Below follows replies to individual people.

And @erlend_sh also suggested looking the likes-to-num-post ratio.

I think that’s a good idea. Also, all like-votes are not the same. A like vote from staff + Leaders because you wrote something insightful, could be worth more, than [many like votes by fairly new members because you posted a joke].

(I have some more ideas about this, but might be better of in a separate topic.)

***

This was a bit hard. I’m thinking most or all TL3 permissions would make sense also for TL2.5, except for “Daily like limit increased by 2 ×”.

If a TL 2.5 was similar to TL 3 permission wise, then, perhaps a TL 3 would be useful, mainly as something you need to have had for like 90% of a year, in order to be considered for promotion to a Leader. So TL 3 wouldn’t add so many new permissions, it’d instead be a step on the way to becoming a Leader. — Perhaps a trust level isn’t needed for this. A badge might work, too, hmm.

How TL 2.5 granted? I’m thinking via a fairly high like-votes vs posts-posted ratio, + not shown signs of bad behavior.

***

Hmm I guess I’m thinking there could be another purpose, too, of the trust levels — to reward good contributions & helpful flags, even if happens infrequently only (but consistently over a long period of time).

I think the image you linked from the Defining trust levels topic, the one with Assistant Mod and Scholar, is interesting. I hadn’t thought about there being two kind of “high-trust” users, namely both Assistant Mod and Scholar. I’m thinking Discourse a bit takes this into account, because if one casts good flags, then, the weight of ones flags, will increase. To me this looks as if D. in that way a bit automatically gives more powers to good flaggers, which a bit seems to corresponds to Assistant Mod, hmm.

I agree. The name I initially suggested, Considerate = kind and helpful, was just the best name I could come up with, at that time. I suppose a good moderator also can be brief & terse, and should be able to make “tough” decisions.

***

That might work, yes. The problem is that it takes time for the staff — that works against what I see as the main benefit with a TL 2.5, namely to save time. Hmm, yes easier to do, than coming up with an algorithm. However, one would need to do it over and over again. The algorithm would be “invented” just once, and then help “everyone”, “forever”.

***

Yes, I think in Discourse’s case, perhaps that’d work almost equally well. (I do think it seems/feels like a good thing, though, to differentiate between infrequently-visiting-helpful-people & very-regularly-visiting-helpful-people.)

The way I look at it, TL 2.5 would be auto-lost too, if the like-vs-num-posts ratio drops too much, or because of flags.

***

Ok :- ) Interesting to hear from someone in that particular situation, i.e. losing TL 3, although probably still being a good contributor & “helper” (just not as often as before).

***

/ KajMagnus


(Chris Hanel) #18

Randomly stopped by the boards for the first time in awhile, and I just wanted to say that I’m pretty happy that my original post/graph is being resurrected so much time later.

Still digging through this new thread, but I might have more to contribute in time.


(Graham Perrin) #19

There’s much to absorb in this topic and its relations so forgive me if I overlook something.

If the outcome will be a new trust level, then will it appear alongside the four at e.g. Discourse Meta where everyone in a community can see who has reached the new level, and can see why it was reached?

Imagine a community where a handful of members, below a trust level, provide useful and/or accurate content that either:

  • can not be judged automatically (by the system) for stats; or
  • is naturally thankless.

The human touch – an expression of gratitude for contributions that slip through the net – is ideal.

However: if that expression is to be badged – at /badges – I’m not (yet) convinced that publicity/badging will be ideal. If there’s public reward to most of the handful, whilst overlooking one or two individuals, there’s the potential to alienate the few who are (for whatever reason) overlooked.

Personally, I’d send an occasional private message to thank any person who is identifiably helfpul. YMMV.


(Jeff Atwood) #20

This “natural thanklessness” seems very unlikely to the point of contortionism. The :heart: under each post works beautifully in practice, on every community I’ve ever looked at.