互联网与文化danluu.com
随机化 HN：取消进入首页的投票门槛
Randomize HN: Removing the vote threshold for getting to the front page
2013 — 2019，它一次次回到首页
黑点是上首页（≥ 50 分）的那次投稿，空心点是没冲上去的投稿，点越大分数越高。
导读
Dan Luu 讨论 Hacker News 上首页的投票门槛问题，并提出通过随机化让新文章获得曝光机会，而不是只靠早期投票越过阈值。文章分析这种机制如何影响首页内容分布和好文章的发现。
为什么经久不衰：它触及所有投票制社区的核心矛盾——曝光机会与质量筛选，随机化方案至今仍是讨论推荐系统冷启动和公平性的经典参考。
HN 评论精选
把随机曝光看作模拟退火：新文章先获得随时间衰减的随机加成，靠早期互动决定能否滚雪球。
我喜欢这个方案，也想指出另一种看待它的方式：它可以被视为模拟退火。一篇全新文章从“高温”开始，有机会随机展示给一部分人；随着时间推移温度降低，它要么成功，要么没有。也就是说，假设有一个 rand_hash 函数，接收文章 id 和用户 id，产生 0 到 1 之间的随机数；假设 T 是新文章已发布的时间（分钟）。在排序时，我们把新文章视为拥有票数：votes + C * rand_hash(article, user) * e^(-k * T)。其中 C 和 k 是影响随机加成能把新文章抬高多少、以及加成持续多久的系数。这样每个人都会在热门文章之间看到随机挑选的新文章。特别有趣的新文章会自然获得几票然后滚雪球；不有趣的会从随机混合中消失。
btilly查看英文原文
I like this solution, and would like to point out another way to look at it. It can be viewed as simulated annealing. A brand new article starts at a "high temperature", it has a chance of showing up randomly for a selection of people. As time goes on the temperature lowers. And it either made it or didn't. That is suppose we have a function called rand_hash that takes an article id and a user id and produces a random number between 0 and 1. Suppose that T is the time in minutes that a new article has been up. For position we treat a new article as having a vote of: votes + C * rand_hash(article, user) * e^(-k * T) Where C and k are factors that affect how much we want a random boost to be able to raise a new article, and for how long we want it boosted. Then what happens is that every person will see, mixed with the popular articles, a random selection of new ones. The new ones that are particularly interesting will organically pick up a few votes and then snowball. The new ones that aren't will disappear from the random mix.
提出用固定数量审核者的统计阈值来决定新帖是否晋级，而不是依赖公开投票门槛。
HN 可以采用面试式审核。新帖子先给 N 个审核者投 yes/no。标准差是票数的平方根。一旦赞成票数超过 95 分位，帖子被接纳；一旦反对票数超过 95 分位，帖子被拒绝；否则帖子停留在未决定区域直到过期。
vjktyu查看英文原文
HN could implement an interview style vetting. A new post is shown to a N interviewers who vote yes/no. The stddev is sqrt of number of votes. Once the number of upvotes crosses the 95 tile, the post is admitted. Once the number of downvotes crosses the 95 tile, the post is rejected. Otherwise the post stays in the undecided area until it expires.
认为延长 new 页面曝光时间比随机化更简单，因为好帖在 new 页停留太短且少人翻页。
在我看来最简单的解决方案是扩大 new 页面的长度。好东西在那里停留的时间太短，来不及收集足够多的赞成票，而人们很少点到第二页。
jacquesm查看英文原文
The simplest solution in my opinion would be to expand the length of the new page. Good stuff simply lives there for too short a time to gather sufficient upvotes and people rarely click through to page 2.
Reddit 的实践经验：加权 rising 页和首页顶部推荐边缘帖有效，但随机化要适配缓存架构。
在 reddit，我们有两个应对首页问题的策略。第一是增加“rising”页面，它曾是 reddit 默认的“new”页面。rising 页面是加权的新页面。它有点取巧，但一度有效。我看到他们改掉了，我觉得这可能是个错误，不过我已经很久没想这事了。第二是我们在首页列表顶部放置一个崭露头角的链接，这效果很好。这帮助那些边缘帖子获得更多曝光。我以前用 reddit 模拟器测试这类东西。现在那个位置对我来说像是 subreddit 搜索框，所以看起来他们转向了别的方案。随机性是个有趣的想法，不过可能有点难以融入我们缓存的方式。
spez查看英文原文
At reddit we had two tactics for the frontpage problem. The first thing we added was the "rising" page, which used to be reddit's default "new" page. The rising page was a weighted new page. It was a little hacky, but worked for a while. I see they've changed this, which I think may have been a mistake, but I haven't thought about it in forever. The second thing we did, which worked really well, was to have an up-and-coming link placed at the top of the list on the frontpage. This helped those borderline posts get more visibility. I had a reddit simulator I used to use to test things like this. That space appears to be a subreddit search box for me, so it seems they've moved onto another solution. Randomness is an interesting idea, though it might be a bit difficult to fit into the way we cached things.
历次讨论
同类还值得读
如今电脑上几乎所有操作在感知上都比 1983 年更慢（2017）
Almost everything on computers is perceptually slower than it was in 1983 (2017)
2017 年的推文串认为，尽管计算机硬件比 1983 年快得多，日常操作的感知速度却常常更慢：启动、界面响应和网络应用被层层抽象、动画、后台检查与网络依赖拖累。
互联网与文化上首页 5 次 · 2017–2020twitter.com网站肥胖危机
The Website Obesity Crisis
Maciej Cegłowski（idlewords）的经典演讲文稿《网站肥胖症危机》，批评现代网页越来越臃肿：动辄数 MB 的 JavaScript、追踪脚本和广告，把本可以轻…
互联网与文化上首页 5 次 · 2016–2021idlewords.com编辑精选亚马逊那本标价 23,698,655.93 美元的苍蝇书
Amazon’s $23,698,655.93 book about flies
博文讲的是亚马逊上一本关于果蝇的二手书被标到 23,698,655.93 美元。
互联网与文化上首页 5 次 · 2011–2022michaeleisen.org浏览器 User-Agent 字符串的历史
History of the browser user-agent string (2008)
2008 年的经典文章用戏谑的“创世纪”口吻，讲述浏览器 User-Agent 字符串如何从简单的标识演变成层层叠加的伪装：浏览器为了绕过按 UA 分发的网站检测，纷纷把自己伪装…
互联网与文化上首页 5 次 · 2012–2022webaim.org编辑精选最后的问题——艾萨克·阿西莫夫
The Last Question - Isaac Asimov
阿西莫夫最著名的短篇科幻小说《最后的问题》的在线全文。
互联网与文化上首页 5 次 · 2009–2019multivax.com编辑精选