The original issue was the deficiency in the moderation process of GOOD/BAD(spam/nonsense). If you find this binary process is still workable, and believe that it will always be workable for you, well, excellent.

All I’m saying is that if the borderlines are taking up too much of your precious time (or you starting worrying that you are deleting genuine comments and building up a backlash from newbies who haven’t learnt to avoid sounding like ever more intelligent spammers) there is another process: GOOD(pass)/BAD(delete)/UNSURE(redact).

I’m merely proposing that. It may have fundamental flaws, but it seems like it might be worth a try one day when some moderators have an uncertainty crisis.

I moderate by email, and like Karl said so well, its a stream of spam and my time it spent picking out the ham. So if it looks like spam at quick glance, it gets nuked. If it doesn’t, then I stand by my decision to publish it.

Right now I’m taking the time to reflect on these comments and where I draw the line. I’m still not entirely sure where it should be, but once I do, we’re back to regular business. And regular business is me skimming my overflowing inbox and making flash decisions.

And making false positives.

The point is not to allow spammers in, even if their content is half decent. I don’t care. I check my server logs and I know they’re not just pushing content in, they’re checking if it shows up. Allowing it to show up paints me as a target (yay, more spam!) and helps them improve their mechanisms.

b) Is it clearly nonsense or promotion (at best only tangentially related)? Yes: DELETE.

c) In all other cases redact.

In class (a) there will be some comments that promote a product, some that are boilerplate, some that are remixes of other comments from related sites, some that are FAQs submitted by spambots or Elbonians, etc. It doesn’t matter. If it met your criteria for a good comment, it is a good comment.

If, surprisingly, you have enough time to analyse a commenter’s motivation and objective by researching their name, e-mail and links, you can do so, but I see little point in this activity, except perhaps when the comment inspires a response from you.

That’s why I’m advising people to make sure their comments don’t trigger my spam finger.

The reason I proposed this test is because it’s easy for me to reason about it. It divides comments into two broad categories.

If the comment is valuable by contents alone, then it would be even more valuable if I put it in the body of the post. Not that I plan on doing it, just that most time it’s easy to classify by this test alone.

So far drive-by spam comments fail that test, which brings us to the next category. It now matters who wrote the comment. Again, simple cases, “I love your new theme!” If it comes from a reader, that’s valuable, I want my theme to appeal to readers. What if it comes from a spammer who thinks it actually stinks and hard to navigate, but decided on flattery instead because they just want a place for their ad?

You’re throwing a party at your house and inviting your friends over. How would you feel if someone walked in uninvited and start handing out flyers for online-flowers-cheap, or sell knock-off designer bags, or tell everyone about Elbonia Metals & Glass Inc?

What if they started each sales peach with an insightful quote from the Book of Wisdom, or just told every person to reboot the Windows box regularly?

I am only considering that peculiar case in which you are not sure if a comment is spam.

In the situation you describe you appear to be quite able to recognise the spam nature of these ‘drive-by’ comments. So I’m not sure what analysis you’re expecting from me.

Anyway, I don’t think you make commenters or their comments more authoritative by publishing/not removing coincidentally valid drive-by comments (you may however indicate their comment is indeed valid, apposite, sufficiently interesting, contributes to the discussion, etc.). Readers recognise that commenters’ knowledge and intelligence greatly varies, from bearded experts down to newbie code monkeys – and now down to Elbonian undercover SEO operatives.

Does it matter if a comment is a FAQ asked by someone who seeks only vacuous nofollow-free dialogue rather than enlightenment? If you think it’s worth answering, go for it. What can you lose? Do you feel dirty come the day you realise you’ve been talking to an Elizabot/Moron/Elbonian/Replicant? It doesn’t matter. You are the one who is valued, as is what you say – and this isn’t reduced by being seen to talk to imbeciles or spammers – as long as what you say is valuable. You don’t have to go back and delete the conversation once you’ve found out the skin colour of the person you were talking to, their IQ, or their motivation.

Of course, you may be on a mission to engage in counter-SEO measures because you simply hate the thought your blog might give a tadette more prominence to a spammer in Google’s analysis.

What is your purpose in moderating commments? To deny spammers link love? Or to ensure discussion isn’t distracted by advertising or nonsense?

Let’s assume I can predict what the drive-by comment would say on a given post. Most people read the post but don’t come back to check on comments. If the comment has value, then I can amplify that value tremendously by slapping it to the end of the post. That way, it becomes visible to everyone, and some people will even take it as more authoritative.

So look at the sample comment above, about clearing the cache. That one is a good reflection of most of the drive-by comment I get. It’s not a hypothetical what if, but an actual spam catch. And I already have a couple of cases where, because of multiple drive-bys, I can predict the comment from the post.

So given the benefit of the doubt that it has value, I’m going to amplify that value by adding this piece of text to the next post that I predict will attract a comment like this. Good or bad?

For boilerplate comments that through chance are apposite, we need another plug-in, a Google search showing the moderator how likely it is to be boilerplate (used elsewhere). However, boilerplate comments aren’t really a problem. The value you add is how you respond to them. You don’t remove value from your blog by unwittingly publishing boilerplate. If the boilerplate prompts an interesting response from you then that’s great. If however, your response is to criticise the boilerplate for not making sense or to ask its author to provide a little more evidence then the ball’s in their court. The boilerplater is already losing value through their unthinking behaviour – even if no-one sees the misses because they’re deleted, the duplication of the boilerplate comment will eventually surface and indicate the comment is spam and reduce the reputation of anything it’s trying to promote. This still won’t negate the value of any comments in response, nor make readers think less of you for being ‘caught out’ by a boilerplate comment.

Even if we became overrun with Elizabots that submit querying comments to bloggers (without otherwise tipping the blogger off that it is a bot), the bloggers’ responses are still valuable (because they’ll only be given in response to apparently good comments). A comment that is good by accident is still a good comment.

Everyone still has to verify comments anyway (before acting upon them). The blogger doesn’t lend their reputation to commenters, they only pass comments that may reflect the views of real people and may be interesting to others. It’s not like you’re publishing a peer reviewed journal in which the academic credentials of each commenter must be ascertained to demonstrate they are true peers.

One thing that’s currently different between intellispammers and Elizabots, and is that unlike the latter, the former do not yet have an energy budget sufficient to engage in any resulting dialogue. Even when they do, or when even Elbonians start getting paid pennies for having jejune repartee with bloggers simply to slip in some promotional links or keywords, the blogger still has an energy budget that dictates how many comments are worthy of their consideration or worth replying to.

A far trickier problem is what a human commenter is to do when a blogger specifically rejects their further comments, or even erases their historical comments.

I fancy a far more peer-to-peer+reputation metric based means of blogging – and trackbacks still don’t cut it. Then somewhat like Slashdot, all participants can decide if certain other participants are spammers, religiously/politically unpalatable (subjective), or by default worthy of being considered their peers. Perhaps this is a job for Google? Maybe they’re already working on it?

Here’s a comment I just got on a blog post discussing login issues (not on Labnotes):

“I too had faced this problem initially then one of my friends suggested me that before setting the password I must clear the browser. And after that every thing was fine”

They left name, email address and URL linking to the Bangalore Management Academy.

Real comment, drive-by comment or automated spam?

Eventually I have to make a judgement call. I sometimes sit on a comment for a day or two until I decide (say the target site is down), I can just leave it as moderated. New information may change that decision, but that hasn’t happened yet, so it’s hypothetical.

So what’s bothering me now is not the mechanism, but the decision which will eventually be made.

Say the comment is by Joe and Joe’s URL goes to an online poker site. Obviously spam that deserves to be nuked, right?

What if Joe, real person, real name, actually reads my blog but always links back to their employee, since Joe is reading my blog on company time? How is that different from someone working for IBM or their own startup commenting with a company link?

I worked on a few WordPress plugins. Splogs are still blogs, and blogs use plugins, so what happens when I get a splog commenting on a plugin they might actually be using? (That’s not a hypothetical question)

If, perversely, you explicitly invite fluff/trite comments, as in “What say you good people? Yay or Nay?”, then you do indeed have to count the yays of fluff spammers. It’s a bit of a contrived example of yours. Your unwritten editorial policy is probably to deprecate “I agree” fluff – newbie commenters that indulge in it soon learn that it’s of little worth (until such time as reputation metrics reveal the true import of ‘I agree’).

For those who moderate comments, I’d imagine a redaction button/option would only be used in the situation where you aren’t sure and would rather not waste time angsting about whether the potential offence and backlash caused by deletion of an off-topic/trite/fluff/possibly spam comment outweighs the irritation of its presence to other readers. Redaction would convert it to a weeny icon, with alt-text hover excerpt, and with javascript onclick popup for the entire comment (links auto converted to text).

I think you recognise that you are going to have to deal with this problem (even though you don’t want to). I’m just trying to suggest that the quickest way of dealing with it is to avoid it, and simply redact the borderline cases. The other cases take a moment anyway. Redaction is a way of also only taking a moment.

A) Fair comment (publish)
B) Spam, abuse, obscene, etc. (unpublish/delete)
C) Hmmm. Not sure… (redact)

Then, because the redacted comment retains its integrity (links despite being in text form can be reconstituted), it’s also an easy thing to have an Unredact option, e.g. when you discover that it was an important dignitary who left an apparently off-topic comment under what they thought you’d recognise was a well known alias. Perhaps George Clooney might submit a vaguely relevant comment about styling his new online casino website [link] under the alias Clooney Tunes?