AI to use grammar syntax for auto-flagging comments

Question

We were having a titillating discussion in the Teacher's Lounge about toxic comments when it came up that "imperatives" could indicate rude comments. I won't defend this, but I'll explain the background...

Take this made-up comment:

Edit your post so it has sources.

That's "imperative", instructing someone what to do. To new users, that may seem bossy. We mods on BH.SE use comments that sound more like suggestions. Albeit, different sites have different cultures... tl;dr

Then, we discuss more...

Adding "please" doesn't make it less bossy, but seems more friendly. Is it?

Please edit your post so it has sources.

Then, we discuss more...

Adding conditionals appeals to logic and feels less bossy. Consider what I've seen others post, and myself:

Please edit your post to include sources so that others don't vote to delete it.

...The imperative alone (first example) may be perceived differently than the imperative with the conditional (third example)...

Maybe you don't agree. Great! Maybe you do agree, but before we get married...

tl;dr

Using pure grammar in our considerations, alongside other considerations, could be very helpful. We could look at meaning, buzz words, et cetera. But, also step back and take a look at the pure grammar itself.

That could be useful because:

1. Grammar is objective and thus is a better reason for flagging, voting, deleting, taking action, et cetera.
2. Grammar could work with AI to more easily auto-flag possible abusive comments.

I don't want to argue whether asking a user to edit a comment is abusive. I want input on using pure grammar itself as a possible way be more objective, both for community human decisions and for AI helping out with the grunt work.

We are discussing this because we want to be more welcoming to new users, in the minds of high-reps on more than one site, as the network can be perceived as toxic. Perhaps running grammar through AI algorithms might help identify trends from comments in the past that were deemed abusive. This presumes that we are looking for ways to be more welcoming.

Albeit, comments would be easier to implement, but once refined it could prove useful for flagging posts and beyond.

Answer 1

You use three imperatives in this very post:

Take this made-up comment

Consider what I've seen [...]

Step back and take a look [...]

That no-one would consider these examples rude in the least must already cast considerable doubt on the notion that the use of the imperative mood might provide a purely syntactical indicator of rudeness without regard to semantics.

A cursory survey of comments on the site I moderate suggests these kind of what you might call rhetorical imperatives are very common. Most other occurrences of imperatives aren't in instructions to do something or other on the site, and many of those that are appear in implicitly or explicitly solicited advice (e.g. how to get a certain symbol with MathJax).

Here's an example where someone is being instructed to take a particular action:

Please do not give new information only in comments, edit your question to add the new information. We want posts to be self-contained, comments can be deleted, and anyhow, information in comments are not well organized. Also, many people do not read comments.

Even if there were only the first, imperative, sentence, it'd hardly be rude; & it wouldn't be more polite when recast as a question ("Would you please not [...]"): what makes it considerate, though, is taking the trouble to explain the reason in the following sentence. You allude to something like this, I think, in your discussion of conditionals,^† but again the grammatical form employed to this end seems barely relevant.

Not to throw the baby out with the bathwater: there are cases where using the more direct imperative rather than a more indirect form can seem rather bossy, but they're very context- (& culture-) dependent. And directness is sometimes called for: it's not a desirable outcome that someone ends with up their question closed/account suspended for ignoring "If you ask me, it might be a good idea if you were perhaps to [...]".

For someone already set up for training M.L. models on S.E. comments labelled as rude or not, I don't suppose it'd be all that much effort to investigate a little, if there's an off-the-shelf P.O.S. tagger that can identify imperatives. I don't know if there is or not: the one in OpenNLP, for example, uses the Penn English Treebank, & would tag an imperative as a "base form" of a verb, along with infinitives. If there isn't, it's a question of whether this looks a promising enough avenue of investigation to develop one specially—in view of what I've written above, I don't think so.

---

† Your example, however, comes over more as a threat or a warning than an explanation:

Please edit your post to include sources so that others don't vote to delete it.

More welcoming is:—

Please edit your post to include sources so that interested readers have something to follow up (or skeptical ones can verify your claims!).

Answer 2

This is not snark. I'm flat serious.

Easy-to-misunderstand Questions may be a greater priority

This Question may be off-topic since it presumes what some might call "hyper-geek" knowledge, or specialized knowledge, which is why I put out a feeler for asking on AI.

As a writer, I'm supposed to make sure others understand my words. By that standard, this is a terribly written OP because it gives some readers the wrong impression that:

• it proposes applying rules with a heavy hand (it doesn't)
• it would exclude all other considerations (it wouldn't)
• it is about Questions and Answers (it isn't, that's possible future)

For example, this comment probably should have been inside the OP:

So, in preparing the AI algorithms and telling an AI what to look for, throwing grammar into that mix would give AI programmers a lot of very useful things to focus on building out. In their programming work, it would never be the only consideration at all, nor would it be right for any kind of first and final decision; it would simply be a useful factor for them to explore over years and decades.

The system could have picked up that I talk about AI, but that is a technical industry that not everyone understands. So, a post for people who are experienced in fields other than AI would benefit from this explanation on a site that is not about programming, (like Meta, Worldbuilding, and Writing.) But, it might be patronizing on sites about programming (like Stack Overflow, Server Fault, Unix & Linux, and especially the AI site).

That might have been really great to have the pre-question system give me a notice like

"Some users might not understand your Question because it assumes specialized knowledge from another field. Consider providing a brief explanation."

The proposed OP may still have some priority

The proper answer from a researcher might be that AI should search "deemed-abusive" comments for grammatical patterns, looking for any trends, only then could we know whether it is useful.