TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Protect me from what I want

188 点作者 lolsoftware超过 2 年前

31 条评论

andrewmutz超过 2 年前
I completely agree with the article. There is a vast soup of user generated content that could be displayed and there&#x27;s always an algorithm of some sort employed, even if that algorithm is &quot;show me only content from people I explicitly follow and sort it chronologically&quot;.<p>If you wanted to give the social media companies the benefit of the doubt, you would say &quot;they just want to have a feed that the user enjoys and the best metric they have to tune their algorithm is is engagement&quot;. If you didn&#x27;t want to give them the benefit of the doubt, you might say &quot;They want to surface a feed that keeps you hooked so they make the most money on ads, so they use the metric of engagement&quot;.<p>I also completely agree with Tim that a marketplace of algorithms is the best way to solve this problem. If users can choose their algorithm, they can choose between the &quot;sugar high&quot; content that is optimized to maximize engagement and time-on-platform if thats what they want. If they&#x27;d rather choose one that maximizes for other metrics, they are free to do so. Examples of metrics that a given user might prefer: &quot;A monthly survey that the user fills out that asks how good the recommendations are&quot; or &quot;a daily survey asking questions related to mood&#x2F;mental health&#x2F;etc&quot;.
评论 #33783398 未加载
评论 #33785406 未加载
gfaster超过 2 年前
&gt; And anyhow, those algorithms are just showing you what you want. Don’t try to deny it, if it wasn’t what you wanted you wouldn’t be doomscrolling so much, would you?<p>I believe this to be based on a faulty assumption: people do what they want. While it may be true that at the reptile-brain level we do want what the Algorithm feeds us, that statement is equivalent to saying that someone who is addicted to opiates wants to feed their addition. It&#x27;s only true at the most shallow of levels.<p>The article at large seems to realize that, but I think what is needed is to distinguish &quot;want&quot; as what we as rats in the Skinner box want and &quot;want&quot; as what we humans who do not wish to be prey to that Skinner box. That is the promise of platforms without the Algorithm, just that it wont try to take advantage of users.<p>(As a side note, I am an advocate for using capital &#x27;A&#x27; the Algorithm to refer to the content-aware, black-box recommendation engines that run social media sites. That let&#x27;s us continue to talk about lowercase &#x27;a&#x27; algorithm to refer to sorting and such)
评论 #33779918 未加载
评论 #33780817 未加载
mgerdts超过 2 年前
An algorithm that I can tune by having private +1, -1, and &quot;hell no&quot; buttons would be helpful. If I follow too many people and&#x2F;or hashtags, I will end up with so many posts&#x2F;toots that I spend a bunch of time reading through things that really aren&#x27;t that interesting. If I had a way to easily mark content as interesting or not, an algorithm could construct my feed such that the things that I find interesting are more likely to be read. There have been people that I followed because they had an interesting post or two, then the other side of them started posting that is not interesting to me.<p>As an example, one person posted that they were working on end-to-end encrypted messages. That was really cool, nearly instant follow. This person identifies as a furry and posts <i>a lot</i> more content related to their furry identity than E2E encrypted messages. I&#x27;m not really into the furry scene, even as a spectator. Nothing against this person - I&#x27;m happy they can be who they are. I just wish there was a way that I could tune into the once in a while updates about the tech project without the part that is of no interest to me. That person may be equally annoyed with my feed if I posted regular updates of my love for mac n cheese.<p>Importantly, the +1&#x2F;-1 that I give should be used only for my feed with no feedback to the poser. My -1 does not mean I disapprove and I certainly don&#x27;t want it to hurt anyone. I don&#x27;t want it to influence others&#x27; feed. A -1 needn&#x27;t mean that I never see something like that, but a &quot;hell no&quot; should.<p>Such a system would allow me to follow more people and be exposed to different ideas that come from more exposure. It&#x27;s then up to me to tune my feed to match my interests and available time.
评论 #33783133 未加载
评论 #33781544 未加载
评论 #33783571 未加载
nostromo超过 2 年前
Most Twitter users, as best I can tell, aren&#x27;t actually complaining about what <i>they</i> see. They are complaining about what <i>others</i> see.<p>In their view, they have the &quot;correct opinions&quot; and have not been biased by any algorithms or moderation. Meanwhile, the people that disagree with them are the ones being duped by algorithms, bad moderation, and bad actors.<p>That&#x27;s why there&#x27;s a huge censorship movement right now in the US. People aren&#x27;t saying, &quot;protect me from what I want&quot; -- they are saying, &quot;protect others from what they want, but leave me alone.&quot; Which is, of course, entirely hypocritical.
评论 #33778844 未加载
评论 #33780219 未加载
评论 #33780709 未加载
评论 #33780236 未加载
评论 #33780490 未加载
评论 #33778370 未加载
评论 #33782497 未加载
评论 #33781210 未加载
评论 #33780246 未加载
评论 #33778960 未加载
评论 #33778687 未加载
评论 #33779528 未加载
lcnPylGDnU4H9OF超过 2 年前
I read this as a problem with how the term &quot;algorithm&quot; is used by &#x2F; when communicating to the layperson (basically, the general problem of &quot;buzzwords&quot;). It&#x27;s just being simplified; possibly overmuch but that&#x27;s not relevant to the layperson&#x27;s complaints. I think it&#x27;s fair to say that the layperson understands that they don&#x27;t want an algorithm which manipulates them to spend more time &quot;engaged&quot; than they otherwise would.<p>I do think there is value to algorithms which <i>incidentally</i> increase engagement as argued by the article (e.g. capturing and isolating email spam), but there is understandable push-back against an algorithm that is, additionally or not, being optimized for pupils pointed at screen.
KaiserPro超过 2 年前
Instagram, youtube &amp; tiktok are a reflection of what you click on.<p>But, to make them nice, you need to give them both negative and positive feedback at the very early stage.<p>If you see something that you don&#x27;t like? Don&#x27;t ignore it, hit the &quot;...&quot; and select the &quot;get tae fuk&quot; option. Otherwise it&#x27;ll keep trying that genre of what ever is popular with a similar persona to yours.<p>The problem is, the &quot;I don&#x27;t like this&quot; button is often hidden. (in youtube I&#x27;m not sure how well weighted the downvote button is).<p>For mastodon its hard, whilst the data is out there for most people to grab, the models need to be stored and trained somewhere. Your model is a PII risk. Also, there isn&#x27;t much info out there on how to make a good recommendation engine (for obvious reasons) Unlike object detection&#x2F;text to speech&#x2F;OCR and other freely available models, recommendation models <i>are</i> as valuable as the dataset. This means they are rarely made public.
评论 #33781919 未加载
eatonphil超过 2 年前
Even if Mastodon doesn&#x27;t want to implement or play with &quot;algorithms&quot; (ignoring as is correctly pointed out that everything is&#x2F;has an algorithm), all Mastodon content is ActivityPub content, right? So someone can put together a Mastodon alternative that supports pluggable algorithms, no? No need to convince the Mastodon devs who don&#x27;t want this sort of thing.
评论 #33780392 未加载
lifeisstillgood超过 2 年前
One other thought is that most algorithms don&#x27;t need to be very sophisticated.<p>So DuckDuckGo is (mostly) predicated on the idea that if you type &quot;mens running shoes&quot; into a search box you can confidently sell ads for trainers all day, and possibly other sporty goods. The bet Gabriel Weinberg is making is that the accuracy &#x2F; profit for selling those ads is not significantly worse than using the text plus a thousand data points about that persons pregnant daughter and bathroom habits. Inwoukd love to see some study on that (not sure how - maybe have google return duckduckgo results and see what they buy. Hell I would bet they have already done this !)<p>Anyway I suspect that I would get as much out of twitter if I just get what David Attenborough reads on twitter (ie a manual curation from someone interesting and well versed in the world)<p>Sometimes the best algorithm is a much more educated and experienced human
andrewmcwatters超过 2 年前
The shock people experience when you tell them any sorting, especially sorting chronologically, requires an &quot;algorithm&quot; is disappointing.
评论 #33778014 未加载
评论 #33778394 未加载
评论 #33778425 未加载
评论 #33780193 未加载
评论 #33778105 未加载
评论 #33779785 未加载
评论 #33778072 未加载
评论 #33778487 未加载
Pxtl超过 2 年前
Yeah, this is pretty much what I&#x27;m hoping for on Mastodon. On Twitter I would occasionally turn on the &quot;home screen&quot; instead of the &quot;latest&quot; feed to see what&#x27;s going on in a larger world. Including such a facility (or even multiple alternate implementations of such a facility) would be nice in mastodon.<p>The other part of the post - the abuse? That part is harder. Fundamentally, commercial social networks spend a crapload of money and traumatize working-class moderators all over the world trying to keep spammers, scammers, illegal pornography, hatespeech, and worse off their platforms. I don&#x27;t know how sustainable it is for hobbyists to implement that.<p>Realistically, I&#x27;m expecting that if Mastodon truly takes off, it will gradually bifurcate: 1) Free instances that are hives of scum and villainy 2) Paid instances that are well-moderated and pleasant<p>Fortunately, Mastodon servers have the ability to federate with other servers in a &quot;second class citizen&quot; sort of way, so members of group 2 will be able to follow people on group 1, but the mayhem of free instances will be purely opt-in on a per-user basis.<p>I mean Mastodon has a lot of problems. Its tech stack is a nightmare to admin, and it is notoriously wastefully chatty when it comes to server-to-server messaging. But those faults can be fixed.<p>The social problem is much harder.
moron4hire超过 2 年前
It seems to me that there is a &quot;solution&quot; that goes completely unspoken in these conversations.<p>There&#x27;s the one side that says, &quot;don&#x27;t do anything opaque, just show me a chrono-timeline, I&#x27;ll deal with the deluge myself&quot;. And another side that says, &quot;you delusional to think you can deal with a deluge&quot;.<p>But maybe there is a third position: don&#x27;t get into a deluge in the first place. Maybe the answer is actually that we shouldn&#x27;t be trying to follow thousands of people on social media. Maybe there is no meaningful way to keep track of that many people and still be able to existentially understand them as, ahem, <i>people</i> anymore.<p>So far, my experience on Mastodon is bearing this out. I have almost exactly 10% of the followership on Mastodon than I do on Twitter. Yet I&#x27;m easily having 5x more conversations. The quality of those conversations is very significantly better, but I&#x27;ll leave that hairy ball of unquantifiability by the wayside for now. So far, &quot;engagement&quot; on Mastodon is 50x better than on Twitter.
nonameiguess超过 2 年前
It would be nice if there was a way to talk about this subject without immediately devolving into partisan political comparisons. This is a far more fundamental issue with recommendation systems, a conflict between immediate gratification and addiction mechanisms versus deeper nourishment and satisfaction. You can see this in so many arenas of human consumption.<p>Consider diet. If you log everything I ever consume and then measure the probability I consume the same thing again in the next 30 minutes, what are you going to find? You&#x27;ll be recommending I eat nothing but donuts, hard liquor, and potato chips.<p>Consider video. If you log everything I ever stream and watch and then measure the probability I stream and watch something similar in the next 30 minutes, you&#x27;ll be recommending nothing but ChiveTV-style fail videos and short-form hot takes.<p>But are these the things I actually most want? Humans are complex creatures with desires and preferences that don&#x27;t always express themselves in terms of quick re-consumption of similar items. If you ask me my favorite film, I&#x27;m going to say Apocalypse Now, not an epic fail compilation. But if you measure what I&#x27;m more likely to watch on repeat for 8 hours, it&#x27;s going to be the fail compilation. If you ask my preferred foods, I&#x27;m going to say lean meats and vegetables, but if you measure what I&#x27;m most likely to repeatedly eat for hours without stopping, you&#x27;re going to find a bunch of dessert foods and snacks.<p>The types of things people consume that are most nourishing and satisfying, since they actually provide nourishment and satisfaction, are not things that lead to immediate re-consumption of the same thing. The types of things that lead to immediate re-consumption are things that are insubstantial, don&#x27;t require thought or reflection, don&#x27;t lead to satiety, and things that are addictive.<p>I have no idea what sort of solution to this scales and will satisfy people that don&#x27;t like gatekeepers, because the reality is, I&#x27;ve found what films and television shows and music albums I&#x27;ve most liked from top 100 lists curated by experts, and I&#x27;ve found what foods best nourish me and lead to the long-term health and physique outcomes I&#x27;m trying to achieve by the same method, expert recommendations from people well-versed in science.<p>You can&#x27;t automate this, but to work, the public has to trust a curator, and it seems most of the public doesn&#x27;t trust anyone to do this, or if they do, they&#x27;d rather trust the Critical Drinker and Liver King instead of the American Film Institute and FDA.
PaulDavisThe1st超过 2 年前
Once again I come to sing the praises of browser add-on Tweak New Twitter.<p>I see just tweets from people I follow, with their retweets available on a separate tab. No trending, no random tweets. Twitter the way I originally imagined it was supposed to be.<p>No need for any algorithm beyond what this offers, unless you&#x27;re really keen on discovering random people to follow, and the people you already follow do not retweet. In which case ... you have my sympathy.
fleddr超过 2 年前
If one was to be a purist, you&#x27;d remove the boost option.<p>With a boost you want to increase the visibility of a post by somebody else. Probably because you agree with it. Something has to happen with that boost signal, otherwise boosting is pointless. Hence, in some way a boosted or often boosted post is promoted over ones that are not boosted.<p>As the boosted post gets more eye balls, it will get even more boosts. Not necessarily because it&#x27;s so awesome, simply because it&#x27;s the post that is shown more prominently.<p>This simple snowball effect has viral potential, hence it&#x27;s just as corruptible as Twitter. It can and will be used agenda-driven. Power always consolidates, hence soon you&#x27;ll have the elitist layer with large followers boosting each other&#x27;s posts. The other 95% gets zero traction or engagement.<p>The other downside of boosts, retweets, quote tweets is that it discourages writing your own original posts. More than 80% of what Twitter calls tweet activity is simply retweets, not new tweets.<p>Finally, if you want organic social media, people to follow should never ever be automatically recommended. This triggers the exact same snowball effect.
评论 #33780807 未加载
评论 #33780831 未加载
评论 #33781941 未加载
jasonlotito超过 2 年前
&gt; Don’t try to deny it, if it wasn’t what you wanted you wouldn’t be doomscrolling so much, would you?<p>This is the most ignorant and idiotic[1] take I&#x27;ve seen on something like this.<p>Don’t try to deny it, if it wasn’t what you wanted you wouldn’t be smoking so much, would you?<p>Don’t try to deny it, if it wasn’t what you wanted you wouldn’t be drinking alcohol so much, would you?<p>Don’t try to deny it, if it wasn’t what you wanted you wouldn’t be taking oxy so much, would you?<p>Don’t try to deny it, if it wasn’t what you wanted you wouldn’t be seeing ads so much, would you?<p>&gt; These ML models know what you want and that’s what they show you.<p>No, they don&#x27;t. This is 100% wrong, and coupled with the previous comment, shows a complete lack of knowledge in this area.<p>These ML models don&#x27;t know what you want.<p>They know what you engage with. What you engage with isn&#x27;t necessarily what you want. And <i>ANY</i> suggestion otherwise is wrong. Simply put, engaging with something doesn&#x27;t necessarily mean you want to see more of it.<p>[1] Yes, that&#x27;s a hard stance to take, but I stand by what I said. But you read this comment, which means you wanted it. Don’t try to deny it, if it wasn’t what you wanted you wouldn’t read any part of it, would you?
评论 #33780067 未加载
评论 #33780042 未加载
compiskey超过 2 年前
What criteria are being used to determine this is what we want?<p>People also want heroin and we use literal impact on society to curtail use. Freedom of reach is what we’re talking about limiting; right and left seem to both demand they have access to my beliefs, and leverage big corp as a tool to hide their political agenda to that effect as government social programs are accountable to public oversight.<p>I mean I’m pretty much convinced at this point human languages lack the nuance to make any conclusions of sound logic at this point except in contexts they came up in; Anglo social norms. The correctness of a colloquialism comes along with a command to use it in a certain emotional way.<p>We’re just engaged in spiraling emotional pettiness; how dare you offend thy sensibilities, good sir!
评论 #33779258 未加载
throwaway290超过 2 年前
&gt; Mastodon introduces a feature where you can download and install algorithms, which can be posted by anyone. They are given the raw unsorted list of posts from people you follow and use that produce a coherent feed. You might have to pay for them. They could be free. They could involve elaborate ML, or not. They might sometimes pull in posts from people you don’t follow. They could be open-source, or not.<p>That, or just use a separate ActivityPub client developed with this feature in mind. After all the point of it all is open and standardized APIs.
PaulHoule超过 2 年前
It bugs me that people call these &quot;algorithms&quot; when they really are &quot;heuristics&quot;.<p>An algorithm for sorting an array sorts it correctly 100% of the time, at least it is supposed to<p><a href="http:&#x2F;&#x2F;envisage-project.eu&#x2F;proving-android-java-and-python-sorting-algorithm-is-broken-and-how-to-fix-it&#x2F;" rel="nofollow">http:&#x2F;&#x2F;envisage-project.eu&#x2F;proving-android-java-and-python-s...</a><p>Something that ranks your feed is a heuristic, a rule of thumb that works right frequently.
nullc超过 2 年前
I have to protest the assumption that if people are spending time on it is what they want.<p>It can be and often is what people fear-- what they DON&#x27;T want. It&#x27;s a pretty good survival technique, in general, to pay attention to things that threaten you.<p>Companies monetize this human tendency, same as any other.<p>Playing on people&#x27;s wants may waste their time or make them dull, but playing on their fears also harms them by producing constant stress.
winReInstall超过 2 年前
I think a happier person produces, does not consume. So a ideal algorithm would be shoving the user towards being productive, creative. To accomplish that, the feed would have to redirected to people who &quot;educate&quot; people to become creators.<p>Take that nice little austrian post-card creator living in bavaria i follow on twitter. Every now and then a political rant, but nothing.. oh No... oh No No No..
评论 #33784740 未加载
jimmytidey超过 2 年前
Am I missing something? I have Twitter set to show the timeline of tweets as they happen, which although I accept it requires a deduplication algorithm is broadly non algorithmic.<p>I realize there is also another &#x27;curated&#x27; timeline setting, but you can always switch it off.<p>There are all kinds of problems with Twitter, but I&#x27;m surprised to see so many people saying &#x27;the algorithm&#x27; is one of them.
评论 #33785712 未加载
kornhole超过 2 年前
If we had a succinct way to explain or logic to point to, it would help clear the confusion about the massive difference between algos built into Mastodon SW and the ML type algos deployed on platforms.
thomastjeffery超过 2 年前
We end up obfuscating too much of the subject every time we say &quot;algorithm&quot;.<p>We are talking about filtering and sorting. There is no need to be more abstract than that.
djmips超过 2 年前
It&#x27;s not what I want - it&#x27;s what an AI model thinks I want and I feel like that&#x27;s modelled after the average &#x27;want&#x27;.
thereald0tt超过 2 年前
No, it isn&#x27;t showing you what you want. It is showing you what it thinks will get you to spend more time on the service, that&#x27;s all
unicornhose超过 2 年前
A good post, I like the direction and will follow bottoms-up ranking design with interest. :)
tensor超过 2 年前
I&#x27;d love my own personal machine learning filter&#x2F;ranking algorithm.
hdjjhhvvhga超过 2 年前
&gt; being used on Twitter to combat Nazis and incels<p>There&#x27;s something unsettling about putting these two terms next to each other.
lifeisstillgood超过 2 年前
I think of this as does Twitter &#x2F; other social media act as a common carrier or not? If twitter just hoses you down with your followed accounts in chronological order then they probably get to be called a common carrier. If not then they are doing some level of curation, choice, publishing and that is &quot;the&quot; algorithm - and yes it might be tweaked to be &quot;better&quot; (good luck defining that) but it always exists.<p>Whatsapp is I think a common carrier by any sensible definition - and yet look at the anti-social effects ascribed to it - murders in India, right wing election issues in Brazil.<p>There is no algorithm controlled by whatsapp on who someone adds to their group and shares pro&#x2F;anti political messages. There is just the real world social graph being imprinted on the technology<p>And I think this is the problem - people.<p>Tech does not mindlessly control us, it simply provides us with our bubbles.<p>I think I am violently agreeing with tim bray.
californiadreem超过 2 年前
Ah, the halcyon days of thrusting off the shackles of capitalism to have freedom in a decentralized digital environment where a new Social Contract is being negotiated! There&#x27;s no way our idealism will feed a new class of nouveau riche that will sacrifice community ideals for profit.<p>Or to be more clear, Google (&quot;Don&#x27;t Be Evil&quot;) was founded in the same year that John Perry Barlow published A Declaration of the Independence of Cyberspace (1996). Barlow is dead and so is his dream. The Empire, long divided, must unite; long united, must divide.<p>Which is to say, federate the web all you want, but a Hamilton is inevitably going stomp on your Jeffersonian dream (and vice-versa).<p><a href="https:&#x2F;&#x2F;www.eff.org&#x2F;cyberspace-independence" rel="nofollow">https:&#x2F;&#x2F;www.eff.org&#x2F;cyberspace-independence</a>
评论 #33779009 未加载
progrus超过 2 年前
“Stop trying to make Mastodon happen. It’s not going to happen.”