Hacker News new | ask | show | jobs
by A_D_E_P_T 1091 days ago
Site admins should "buy upvotes" a few times, at different hours of the day, and aggressively ban every account that's a repeat offender.

But, really, upvotes are always easily exploitable and often lead to a sort of herd mentality among commenters. This is why Reddit is the definition of "echo chamber." HN is much better, but it would still be improved if upvotes and downvotes were done away with entirely.

8 comments

I’m struggling to see how HN without voting would be anything other than a simple timeline of posts with comments.

If you compare the level of “interestingness” of the front page and the first page of new, it’s night and day for me. I still go to new sometimes, but when I do, I feel like I’m doing community service to try to surface the most interesting items for the community.

I read hn through hackernews.com, which is kind of the same? Yeah it's "frontpage" only, but it works well for me. I like to scroll through everything I missed.

It would be more than difficult to figure out a way to do this without a "frontpage" but anyone who cracks that, would have me as a user.

> I’m struggling to see how HN without voting would be anything other than a simple timeline of posts with comments.

I feel like voting for the content itself but keeping the comments section flat and sorted by time would lead to less gamified and much better comments sections than what we have now.

Maybe keep upvotes around as a positive reinforcement without having an effect on the position of a comment, and obviously mod flagging for things that break rules, but upvoting and downvoting being equivalent to "I agree" and "I disagree" leads to users gaming the system to hide things they disagree with.

> but upvoting and downvoting being equivalent to "I agree" and "I disagree" leads to users gaming the system to hide things they disagree with.

I've elsewhere said that useful/not useful and agree/disagree should be two different buttons, similar to how LessWrong does it. However, it is a community with even higher SNR and a very different look and feel compared to HN, so I don't know if the same thing would work here.

I mostly meant "upvotes and downvotes on comments" -- as opposed to on topics -- but now I wonder if all upvotes and downvotes could be done away with. Perhaps there's a different way to gauge "interestingness," such as clickthrough rate and comment volume?
HN doesn't have downvotes on topics/submissions, which IMHO is part of the problem. You can only flag. But there are a lot of stories that don't rise to the level of "this is totally inappropriate/spam", but also aren't very good either.

In any case, I don't get the hate for comment up/downvotes: yes, some people will try to game it, but overall, it tends to get good stuff to the top, and is better than having to wade through comments randomly.

HN was originally looking for user entrepeneurialness. Interestingness is a poor goal, if it even is a goal. For what purpose are we looking for interestingness? "Why?"

I suppose we can start by splitting up introvert and extrovert brands of interesting. Entrepeneurialness for example comes in both pondering alone time and the gathering of crowds.

If we could succeed at creating or uploading a digital dang people would be able to bench their publications in advance and create the SEO click bait hell all over until eventually (like google) success has it that the only remaining signal of quality is how high it was ranked previously back when the algo still worked. Counting click-though is specially terrible there.

Only setting real goals and tuning against exploits could produce hard numbers. It's much easier if the goal is selling a product but even then there is the short and the long game.

> HN was originally looking for user entrepeneurialness.

I don't know what the original site guidelines said, but the earliest version I could find in the wayback machine (July 2, 2014) starts exactly the same as the current content. [1]

  Hacker News Guidelines

  What to Submit

  On-Topic: Anything that good hackers would find interesting.
[1] https://news.ycombinator.com/newsguidelines.html
Click-through rate optimizes for clickbait titles and controversial topics. Comment volume optimizes for controversial topics as well, since people would be arguing in the comments in those. It also optimizes for comments that don't add value.
I suspect Reddit's echo chamber is because of

1. "shadow moderation" (Robert Hawkin's term, see https://www.reveddit.com/#spread_word)

2. "The masses" entered. When the site was smaller, diversity was expected (as well as the average technical proficiency of the users). With larger numbers, diverse thought held by minorities is harder to hear.

Or don't ban the accounts, but secretly build a detector of posts which receive bought upvotes.

Then, don't show such posts on the frontpage.

The thing is downvoting isn't that visible on HN due to scored being hidden, just post hiding, which feels useful for bad posts which get it and deserve it (spam, useless comments, and the like)
It’s hard to attribute Reddit’s rot to voting when it’s had so many years of administration alternating between absentee and inept, and when HN gets such dramatically different results. I generally agree that the trivialization of sentiment is probably a bad thing, but there’s also a lot that works here and one should be careful not to break. I’d certainly be curious to see how it plays out. I wonder if one couldn’t disable voting for only some threads…
i use one of those filters that only show me stories that have at least 50 points. i only read those interesting to me, so i don't know if any stories with fake upvotes make it through, but without that filter i'd be overwhelmed.

i suppose number of comments would work as an alternative measure, but that doesn't work for stories that are interesting but don't generate much discussion. and it potentially would just push people to post fake comments instead, creating more moderation work.

Decentralized hackernews/reddit of the future (which can't do this) will need some sort of weighted vote system...
This is actually the only challenge. Building a [say] p2p search engine is much easier than people think but ranking results is a pain.
I think this is solved by "authenticated atomic action", where users sign individual interactions instead of trusting login servers. In this case, each upvote, downvote, and comment must be cryptographically signed.

Once user action is signed, trust must be assigned to networks of users. (Of course, individual users can be trusted, but it would be easier to manage if users are bundled into networks somehow.)

Key management is the hard problem for this to work. User need to be able to have many keys that are revocable. Blockchains like Ethereum are advocating for chain centralized solutions, but I see no reason why that must be the case. It can be done, from top to bottom, in a totally distributed way. From the blockchain perspective, a "multi-chain" solution is the right one.

Why do you think they'd use the same accounts every time?
They do what works, if using the same accounts works they will do it. If they can use new accounts they will do it. On fb, reddit, twitter, google etc they gradually develop accounts sometimes over years. If those are detected by the tens of thousands it would drive up prices.