We should've stopped--but let's keep going and see what happens.

HN/Lobsters: Good News, Bad News, How to Win Big (Part 2)

In part 1, we talked about HN/Lobsters-style sites and how their karma mechanics promoted good behavior and helped produce civility and discussion.

Next, we’re going to cover how those same mechanics promote toxic behavior, and how a bad-faith poster can cynically exploit the system in order to gain standing.

Note that this is all going to be behavior accomplished by a normal user. No sockpuppeting, upvote ringing, or anything else sketchy needs be happening—this is just playing the game straight with a different value perspective. As such, this is also all behavior that plausibly a normal user may accidentally engage in.

Bad News

We’ve shown how to be a good poster, and how to really improve the community and get decent amounts of karma for little time investment.

Let’s assume though, in typical paperclip-making machine fashion, that we only care about increasing our karma. This assumption also means that we have extra time (perhaps a couple of hours a day) to spend…that being the case, every little trick we can use will help ratchet us to greatness.

The bad news is that there are people on HN and Lobsters that are like this and that, worse still, the same mechanisms that make normal posters good make bad posters really really toxic. These toxic posters take advantage of at least some of the following anti-patterns:

  • Strategic shitposting

  • Junk submissions

  • Post order manipulation

  • Post manipulation

Strategic shitposting

Strategic shitposting is a catch-all for purposefully making posts that gain karma without contributing to the discussion. Examples of such posts would be some of my earlier work on Hacker News. This behavior relies on throwaway jokes or snarky replies to comments, almost always able to produce a few more upvotes than downvotes. Properly formatted (in my case, leaving out all punctuation and capitalization to make it easier to parse as less thought-out than my normal work) these posts somehow manage to amuse more than they annoy, and in so doing they can reap a tidy profit.

The posts in question always need at least a little relevance, in order to prevent flagging (Lobsters) or downvoting (HN), but their main purpose is to get a visceral reaction (typically, humor and amusement) in the reader. Once that reaction occurs, they’ll probably get an upvote.

How does this work? Let’s talk about voting.

How voting works on HN/L sites

Why do people upvote things?

  • They just want to click on the little arrow.
  • They agree with the point the author is making.
  • They disagree with the point the author is making, but agree with the writeup of the point.
  • They want to reward everybody who has participated in a thread of conversation.
  • They want to manipulate the rankings (more on this later).
  • They assume that whatever the person is saying is correct (patio11, tptacek, pg) and want to reward them.
  • They want to downvote, but they missed on the touch interface. HN won’t let you correct a mistaken vote, though Lobsters will.
  • They want to reward the poster whenever possible.

Why do people downvote?

  • They disagree with the point made by the post (primarily HN, due to a misguided public policy statement by Paul Graham and parroted thereafter).
  • They believe the post is rude or uncivil (formerly on HN, now primarily Lobsters).
  • They want to pile on downvotes for already downvoted things (any posts that are greyed out tend to mark themselves as “downvote me”).
  • They meant to upvote, but missed on the interface (again, mainly an HN problem).
  • They want to manipulate the rankings (more on this later).
  • They want to punish the poster whenever possible.

In Hacker News, for a long time, downvoting was de facto reserved for marking comments as excessively hostile or rude or off-topic. Lobsters to this day tends to function that way and even has a more nuanced downvote system for showing why a given post was downvoted.

Using voting mechanisms to profitably shitpost

This standard means that—for a poster without a reputation, and for a post which is otherwise not abrasive—if a post gets any reaction, it’ll probably be an upvote. Thus, the winning shitposting strategy is to write posts that can cause a strong reaction without being too difficult to write. A few different types of posts have enhanced survival characteristics by leveraging that truth:

  • Posts that are primarily jokes, using the context of discussion as source material.

  • Posts that signal simple agreement with whatever the current zeitgeist of discussion is. They need not be long, but people react more strongly and much more positively to these than short disagreeing posts (which they may even downvote!).

  • Posts that are minor corrections (technically correct, the best kind of correct). The reaction is due to the thought process of the reader checking that the post is in fact correct, rewarding the correctness, and moving on, even though the post itself contributes little to the conversation. Think grammar or URL corrections.

  • Lampshaded off-topic rants that are tangentially related. Most HN/Lobsters readers seem to forgive a blatantly off-topic post if it contains an upfront admission that they are so.

  • Griping about injustice or outrage, the more entrenched the better. Pandering to injustice and impotent outrage evokes strong reactions and such posts can be easily tailored to match the overall views of the hivemind.

None of these posts actually tend to elevate the discussion or reveal new truths, but people will almost always upvote them more than they downvote them.

And that’s the source of their toxicity: shitposts do at least as well as quality posts, they don’t increase the signal of the community’s nominal area of discussion, and they are very easy to crank out even by idiots.

Junk submissions

Junk submissions are articles and stories that do well in karma but that again dilute the focus of the community. I suggest that the voting patterns for submissions mostly follow those mentioned above for normal posts and comments, so I won’t rehash that here.

To pick a good junk submission, you basically want something that:

  1. Doesn’t require an in-depth understanding of the subject matter to comment on. Some random user who sees an article on type algebras statistically isn’t going to be in the population who cares about such things, whereas bike-shedding on some Algol-derived language is fun for the whole family.

  2. Doesn’t take a long time to read, because people will be less likely to upvote it. The longer somebody takes to read an article, the more likely they are to move on to something else (or to disagree with it, or ignore it entirely). This is precisely the opposite behavior from giving your submission upvotes.

  3. Evokes a strong reaction in the reader. As with comments, the only bad reaction from a user is no reaction—so, pick a topic that demands to be felt strongly about. So, “C is Dead” is going to be a better article than “Subtle memory aliasing issues with legacy C99 code”. Similarly, “How We Prioritized Diversity” is going to do better than “Surveys of Hiring Practices and Diversity” or mundane bureau of labor reports, even though it may have objective information.

  4. Is likely to create a lot of discussion. Remember, the ranking algos slightly favor more discussion over less discussion. You want to pick an article that is hilariously one-sided, or that leaves out key details so as to provide speculation. Anything that is cut-and-dried is probably not going to get a lot of comments compared with proper bait.

In other words, you want news, gossip, and/or controversy. HN and Lobsters both have some measures in place to lower submissions that have a lot of discussion but few upvotes, but those aren’t always effective. Additionally, done properly, you can use a junk submission to create a loud discussion and then profit off your own posts and comments within that discussion even if the story itself gets torpedoed. Loud arguments may take several days to slide both off the front page and out of people’s minds.

Gossip is attractive but obvious

Gossip tends to get shut down on HN, mostly because the best gossip in our industry tends to run counter to the explicit goals and interests of the folks at Hacker News. They understand that frequently appearing as a clearinghouse for muckraking damages their reputation and their community, so mods will kill off stories that are primarily gossip.

Gossip about the community site itself (meta as it is known in other contexts like Stack Overflow) trips the same breakers. On the useful side, though, meta discussion easily hits all of the 4 criteria for a junk submission, least of all because it usually creates a lot of discussion and everyone has something to say.

Controversy is a reliable standby but self-limiting

Controversy is often overlooked by users as needing to be curtailed, but may fall prey to overzealous moderation. The same thing that causes people to upvote controversial articles tends to make them complete jerks in the threads, and rapidly the civility of such discussions tends to die out. Pre-emptively, mods may hide or nuke such threads to prevent them from dominating the front-page. While the articles rage on, though, you can expect a lot of votes—especially if you picked a submission that doesn’t blatantly appear to be using this tactic. It can often help to seed the discussion with a comment to highlight or bring out the controversy, and in so doing get better mileage out of a safer article that doesn’t at first glance appear incendiary.

As of the time of this writing, the most reliable form of outrage submission invokes social justice. Everybody on HN or Lobsters has known somebody who has been discriminated against unfairly, because we really do have a history of that in our industry. It’s a topic getting a lot of coverage in other sectors.

(Aside: I don’t mean to use social justice as a pejorative here: it’s a term I’ve seen used by both sides a great deal at this point, and it tends to be a good shorthand for the category of discussions of privilege, diversity, gender, equity, and similar issues. I think those issues are important to examine, but here we are only concerned with their utility for gaining karma.)

Let’s look at it under our four criteria:

  1. Social justice outrage is typically reported in either news or blog form. There is a large (and often impenetrable to the untrained novice) body of academic work on social justice concepts, but the majority of what is discussed online today is short-form and written to make people aware. Similarly, everyone has a least some opinion on the topic, even if it is as simple as “I hate political correctness!”.

  2. Social justice outrage often is event-based, using social media. Because a lot of material is sourced from Twitter or Medium, it tends to be pretty short and quickly-digestible. You don’t have to grep through more than a few tweets of content to get the meat of many posts.

  3. Social justice outrage lends itself exceptionally well to provoking a response. Because of the previously-mentioned history of discrimination in tech, and because everyone has felt excluded or slighted at one point or another, any user with any shred of decency or empathy will find some resonance with any material involving those topics. Additionally, there are people who are angered or feel persecuted whenever social justice articles show up—and they tend to either upvote an article to get it discussed or downvote so conspicuously as to garner sympathy upvotes from users not sharing their preferences.

  4. Social justice outrage always results in lots of discussion. Discussions on these articles tend to involve a lot of personal suffering (in which case people commiserate, question, or express sympathy) and argument (because people tend to hotly debate facts or deeply-held convictions). Even though the discussions themselves typically don’t enrich anybody’s life, they do generate a lot of opportunities for either quality commenting or shitposting.

All the same, a problem with these sorts of submissions is that are usually heavily moderated once the site admins catch on. HN, for example, seems to heavily monitor and remove these sorts of submissions once they pop up. The counter to this, of course, is to start with articles that aren’t directly about this, and slowly ease in more and more inflammatory social justice articles as the community standards normalize in that direction.

News articles are your secret weapon

News articles tend to be best form of junk submission.

What makes them so great? They map directly onto the four useful traits we talked about above:

  1. News articles are, by construction, easy to digest. To gain the widest-possible circulation and reach, news writing tends to be simple and accessible. For our purposes, that’s perfect.

  2. News articles are almost always relatively short. It’s hard to keep the TechCrunch pipeline flowing if every article is suitable for print in The New Yorker. It is also harder to keep on the bleeding edge if you post long analyses, and frankly it’s hard work. So, shorter articles with links to sponsors and relevant information carry the day.

  3. News articles are, by definition, novel. In addition to whatever feelings the articles promote, news always have a sense of novelty and discovery to them. Almost any user gets a mild rush from not having missed out when reading an article or by having their curiosity rewarded.

  4. News articles tend to create discussion. Everybody loves to show off their grasp of the nuances of a news story that others may lack, and they’ll happily chat about that. Or, they’ll complain about the rest of the story being left out. Either way, they’ll be talking.

There is a deeper toxicity to news, though, and one that only Lobsters so far even seems to acknowledge (through its tagging system):

Every time you post a news article, product announcement, or current events thing, you are creating a minor data point that says that the novelty of the article, and not the content of the article itself, is acceptable for the community. This is an important precedent to set, because it means that as you have a harder and harder time finding good submissions, you can fall back on novelty over quality.

Over time, this means that the front page (say, of HN) becomes filled with things that would otherwise be considered off-topic, including advertising and product releases and whatnot. This is not postulation—this has happened on HN.

On a site like Lobsters, this process absolutely destroys the uniqueness and educational value of the site.

Additionally, as this rot sets in, there is no going back—once the name of the game is “What is the most novel thing we can post?”, only news articles will be commonly successful. They have an entire industry of people behind them optimizing them for that niche, and simple links to other things have little chance.

Post order manipulation

By carefully upvoting and downvoting, as well as picking what you reply to, you can also improve your odds of getting karma.

Observations about how people read posts:

  • Early posts tend to get the most reactions, and stay wherever they’re ranked.

  • People tend to pay attention most to the top-most posts.

  • People tend to pay the next-most attention to the bottom posts, especially in long threads.

  • People skim over subthreads which are more than a couple of levels deep.

  • People may downvote entire subthreads if it turns into one or two posters bickering.

  • People notice early-on if the same poster is replying in all of the subthreads.

These observations suggest some tactics.

  • If a post is near the top of a thread and the thread is long, it is better to reply to that post than to create one at a sibling level. If yours is the only reply to a post at the top, you’ve basically assured your spot as the second post people encounter.

  • If a post is near the bottom of a thread, reply to it. This is again prime real-estate, and bottom posts tend to be blatantly bad—thus, a reply correcting them or mocking them tends to get additionally upvotes.

  • Don’t engage in pointless back-and-forth with another poster. Watch your per-post karma if available, as in Lobsters) and stop once you only are getting a vote or two per reply.

  • Early on, any post that isn’t yours that people agree with should be downvoted if it looks “strong”. Alternately, you should upvote a strong post and reply to it immediately, to help ensure a good spot in the reading order.

  • Resist the urge to reply to everything in a thread. Done too much, or too frequently, you will antagonize the other posters, and eventually they’ll downvote you on principal everywhere in a thread.

The reason that this behavior can be toxic to a community is that while it can help people act more civilly (as in the cases of avoiding back-and-forth and posting everywhere), the behavior itself can cause neglect of middle-quality and late posts. Rankings of posts also become suspect, because a post that is at the top may have only gotten there by the author downvoting more relevant posts and discussion baiting. Subthreads can also become cluttered up by virtue of people using the “post to a top thread” heuristic instead of the expected “engage with posts worth engaging with” metric.

Post Manipulation

Manipulating and editing your posts after the fact is an old technique, but still sometimes productive.

One thing to do is to edit posts for content after seeing a similar post get eviscerated. For example, one may tone down a post from being too aggressive after watching a somebody else get downvoted.

Another reason to edit posts is to remove content that others are using to call you out. This is doubly effective if a replying poster gets really agitated—if you remove inflammatory content after somebody has replied in kind, it appears to the outside observer that that the reply post is being overly mean.

Finally, and perhaps even in decent faith, editing to append additional context or corrections and acknowledgements to a post that was originally incorrect or poorly worded can actually result in more karma. People who may otherwise have been annoyed may say “Aha! They’ve corrected themselves and owned up to it—I shall reward this with an upvote (or perhaps remove a flagging).”)

This is toxic behavior because it tends to destroy the archival value of discussions—for this reason (I assume) Lobsters and Hacker News both disable editing of past posts. That said, if people delete old posts they can achieve similar outcomes. Additionally, it can be really gnarly when people are trying to follow a conversation or point and suddenly the original words that led to a disagreement are gone or are different from what they used to be.

Thoughts on negative strategies

Frankly, these strategies range from annoying to downright harmful. I’ve used a handful of these at various points in my posting career, and honestly I regret my small bit of negative influence on the communities where I practiced it. When I look at the practices on places like Hacker News nowadays, I can’t help but wonder how many people employ these or similar methods, and how much of the behavior is considered acceptable now because of the precedent set by me and others like me.

The worst part is, these strategies appear to work and work reliably. The first two, shitposts and junk submissions, are almost disgusting in how reliable they are after a bit of practice. They don’t educate anybody, they don’t elevate the conversation, but damned if they don’t increment that karma every time bit by bit

Hopefully, the techniques presented above will give everyone a clearer idea of the sorts of things they shouldn’t support and condone in their sites, and also help make it easier to detect when these things are being used in their presence. Ideally, if we are all so informed, we can stop ourselves from being exploited.

Also, as you may note, a lot of these behaviors with only slight modification can result in really helpful and engaging discussion. Posting articles that are relatable and educational instead of being clickbait, making jokes that rely on a technical knowledge and help illustrate points and teach—these are things that manage to help instead of hurt, but can so easily go the other way.

In the final part of this series, I’ll discuss the things that I think site maintainers can do to help encourage good behavior and to help limit the viability of these techniques.