Just-in-time behavior nudging


(Aja Bogdanoff) #1

So obviously this would be for communities with very high standards for behavior/etc.

What would you all think about the usefulness of running some sort of basic semantic analysis (or even just word matching) on new posts, and popping up a little “nudge” when the post is questionable? Like so: (sorry, my attempts to be insulting are really lame)

Would a gentle nudge in the right direction be useful?

Another option would be to automatically catch txtspeak, all-caps, no-caps, etc, and offer the user an auto-correction before posting. Something like: “Hello Aja!  We’ve automatically edited a bit of the spelling and/or grammar in your contribution.  Please take a look over these changes, and feel free to make any additional corrections before submitting your post.  Thanks!”

Thoughts?


(Shawn Holmes) #2

It’s a great idea! We talked a lot about the possibility of providing “gentle education” to the user in early Discourse discussions. I believe where we left it was “probably applicable to a plug-in”, but I’ve always been a strong supporter of some these common “catch-alls” in the core game play.

I remember bringing up the ALL CAPS OMG YOU’RE YELLING to @CodingHorror as an easy win, and he felt that it just wasn’t as common in practice as we thought. I have a skewed perspective with the communities I’m involved in…some of which are comprised of users less than technical than us.


(Pyry Matikainen) #3

I think the last thing you want to do is foster an environment where people think that they can score conversation points by arguing about grammar and posting conventions. These types of rules and ‘standards’ have a habit of becoming ends in themselves, markers of ingroup-outgroup status, rather than means towards good conversation.


(Shawn Holmes) #4

However, if a system were to be intelligent enough to handle the capture/education of this on its own, one could argue that the community could keep itself in check by enforcing the abstinence of back-seat pedantry involving said conventions.

Rather than scoring points on calling out each other’s grammar, the points would be scored for exactly the opposite, reminding each other that the system handles it.


(Sam Saffron) #5

I did a fair bit of similar work at Stack Overflow with @codinghorror.

I like the idea, and in high scale communities it can really help tidying stuff up.

Thing is, to build this kind of feature properly you need to have a corpus of data and problem patterns you are trying to eradicate. Otherwise we would be building this based on guess work that very much risks an unbalanced approach.

Keep in mind we already handle some of the more trivial ones with our entropy checks, however its a system we should be constantly refining based on real-world data.


(Jeff Atwood) #6

I agree with Sam that this needs to be based on actual problems within a community rather than guessing what the problems may be.

And sometimes the detection is hard – for example we say in the FAQ that people should not end posts with signatures, such as:

Hello, how do I make a post in Discourse?

Thanks in advance,

John Q Public, Esquire
http://example.com

How do we reliably detect a signature at the bottom of a reply? How do you reliably detect curse words without creating false positives? There is a limit to what you can automatically process here.

We do enforce a number of fairly narrow, specific rules around topic titles based on noise characters, length, entropy, and so forth – for example, I see topics here that end in multiple question marks, but we should not allow that.


(Jeff Atwood) #7

There is a just-in-time reminder of civilized discourse when new users post a reply for the first time:

And for topics:

Since then, we also automatically fix a few problems with post titles based on real user behavior @geek saw at his Discourse forum. Titles matter a lot, so things like all caps titles, titles with tons of exclamation points, etc we silently fix for the user behind the scenes.

I should mention that the (great) new book by Dan Ariely covers similar ground

Turns out most people steal just enough so that they can still consider themselves fundamentally honest people… but “just in time” reminders not to steal, to be morally upright, had a shockingly high success rate in reducing these kinds of casual theft rates.


(Dave McClure) #8

That was a good book. If anyone wants paperback copy to read, PM me with your mailing address and I’ll send it to you for free.


(Sam Saffron) #9

You know some of us live in Australia :slight_smile: assume you mean send it “locally” for free


(Dave McClure) #10

A deal’s a deal… This copy’s going to Norway though.


(Alexander) #11

Sounds like a plugin that would let admins create their own rules would be a killer feature for some audiences.