Purely algorithmic feeds work for TikTok because people want to shut their brains off when they use the app. It's like a drink after work. Copy-pasting that logic to longform text misses the differences in the mediums, and how people interact with them.
When I'm reading, I'm trying to be thoughtful, not titillated. And part of being thoughtful is consciously choosing what to read.
One might say that sites like reddit and hacker news are just archaic versions of the same recommendation engine, but the friction is actually an important feature, and keeps the experience from devolving into lowest common denominator clickbait.
There's social modeling suggesting that simple ideas (anger etc) dominate in overly-connected networks. That's why the friction you mention is important. This has bothered me for a long time, so I made the opposite of Artifact a while back: https://www.recents.cc
You follow a few people, and you get whatever links they post, no algo or frontpage. You repost good stuff, so a link can travel several hops to someone, but it must have been filtered through a friend. Kinda like real life. In the time that I've used it with my friends, I've seen a nice balance of mainstream and niche stuff in my feed.
I think it is more about an exploration-exploitation problem. I think you're focusing on the exploitation aspect too much. Algorithms like TikTok's do well with respect to exploration because there is a low cost for a bad recommendation. The quick content allows the algorithm to update quickly to the users current mood. This isn't going to be true for articles, where you are going to have spend more time and energy on the content being provided. This means that you have to more quickly model a person's mood/energy levels when they are using the app (you can determine this faster with quick content because you are iterating faster). Exploration is an important part of the overall reinforcement strategy because as humans our cost function is constantly changing, we do crave novelty (just in different levels). If you don't have the exploration side, and only have exploitation, then you just end up in a bubble (a bit can be good, but life needs spice). Worse, you'll end up bored.
I tried Tumblr, and it felt like a way better version of Twitter. Had me follow some trending accounts for a few topics. Then I get three feeds, default being those I follow, others being the algo or my tags. I can like, share, or repost stuff. The UI is clean, it doesn't lag my laptop, and signup took me about 1/3 the time vs signing up for Twitter.
So the social difference from Twitter is they kept the chrono following-only feed? It's nice but not what I envisioned for Recents. I wanted to avoid any kind of global "trending" aspect and focus all interaction on reposts. I'm probably reinventing a wheel somewhere, but not one that I've seen.
Can't wait for ActivityPub support so I can control the UX better with a different client.
Come to think of it, this what User Agents were supposed to be, but we just added a few more layers on top (for that sweet sweet ad revenue juice), and now are stripping some of it with the likes of Mastodon.
Tumblr was very close to the perfect social media network. It's a shame it succumbed to corporate greed after the Yahoo acquisition. I really hope someone manages to recreate it as a decentralized protocol
It's no longer owned by Yahoo! but by Automattic (WordPress) now. As for decentralized, they've posted recently about their intent to support ActivityPub.
The problem is not about the broadcast capabilities, they already support RSS.
The problem is about their content censoring. I don't reckon Automattic reversed the "sexually explicit" purge initiated by Yahoo. I'm not interested in a social network that censors ~30% [1] of the global art production. For now Twitter is the better place for relatively free content sharing.
[1]: Number pulled out of my hat, but the human nude being the most popular drawing subject ever, I wouldn't be surprised if this is close to reality.
You might enjoy "Rage Inside the Machine" by Robert Eliott Smith which describes how systems like social networks can be manipulated by introducing a small group of highly motivated, but strategically placed actors (human or algorithmic).
And now, unsolicited UX feedback, Hacker News style™!
1. put the content of "?" in the front page for users who are not logged in, instead of "you're not following anyone"
2. split the 3 key points you mention there into three columns with titles, so the user (no need to do anything more complex, just throw in a flexbox in CSS)
I like the idea of the site, but 1. would make its purpose a bit more obvious/save you some time explaining the idea, and 2. would make the now obvious easier to digest by a new user/scan visually
Thanks for the tips. As you can maybe tell, I'm pretty new to UI dev. A lot of those divs are using fixed pixel sizes calculated by hand in React because I lost my patience with CSS, but yeah I'll put the (?) page into a flexbox.
> social networks can be manipulated by introducing a small group of highly motivated, but strategically placed actors
I should read the book and see if there's anything I didn't think of, cause one of my top concerns was making something hard to exploit.
Someone should one day write a desktop RSS aggregator that finally makes use of <rating> element. (with a website producing the feeds)
It's a bit of a pain figuring out PICS rating and rating services but I think it is a fantastic idea (that found only stupid implementation in parental supervision and porn filtering)
It seems to close to what you describe to not mention it.
This is a great idea. I totally agree. It's the same design we've chosen for the social network in Peergos. Two main differences are we enforce no server-side algorithmic curation by being e2e encrypted, and posts can be anything, text, images, videos, links. If you're interested you can read more here:
Impressive! Yeah, I think this is the natural evolution. Chat apps were the hot thing after social networks, and things are appearing in the middle. Though you're going for ambitious security.
Thanks! The security is partially a defense against future owners/operators of your server. But I'm totally sold on the idea of only seeing things shared by people you follow.
This is basically how I use twitter. I follow a few friends and some stuff I like. I retweet stuff that I like and that I think they might potentially like. They do the same. Add to that some microblogging, and conversations that emerge from that, and you have a great social media. At least for me and them.
Heh, maybe I should've posted the link to follow my own account (hotgril2) on there, but I didn't want to seem too advertise-y. This also makes me follow you back. I post a mix of tech, history, world politics, and music.
I agree with you. While I love RSS and use it frequently, there are two things that I find that could be improved:
1. RSS doesn't help me find new interesting blogs and writers. One of TikTok's few virtues as a platform is giving algorithmic reach to creators who have no social reach.
2. My RSS collections have always been one of either I'm following a highly curated list that leaves me hungering for more content or I'm following a firehose that makes me miss the content I prefer.
I stumbled across a website that does something like what the article describes, but for just tweets[0]. It's surprisingly compelling and gives me the feel of using StumbleUpon. I am eager to see more things like it.
To help with discovery of new RSS feeds I use this mechanism for https://linklonk.com - when you upvote or submit a link, you connect (subscribe) to all users that upvoted and to all RSS feeds that posted that link.
As you rate content you get connected to more and more sources.
The way it solves the oversubscription problem is - whenever a user/feed posts something - your connection to them goes down slightly. As a result you see content from sources with the highest signal-to-noise ratio first.
The differences from Artifact is:
- Artifact uses implicit signals such as read time; LinkLonk uses your explicit upvotes. I think implicit signals are fine for entertainment content, but not for informational content. Time spent is not equal to becoming better informed. I think only you can be a judge of that.
- Artifact uses an opaque AI algorithm optimized to do what they want; LinkLonk uses a transparent algorithm, where each recommendation comes from a feed or a user that you have co-liked items in common.
> My RSS collections have always been one of either I'm following a highly curated list that leaves me hungering for more content or I'm following a firehose that makes me miss the content I prefer.
Filtering helps but "i just like some type of content from that person" seems to be forever problem on every site. Like on youtube, you watched whole three videos of a guy cooking ? I will bomb your feed with random crap from his channel for next month or two!
I don’t think TikTok innovated by merely having an algorithmic feed. Facebook, twitter, YouTube have algorithmic feeds. I think they innovated in giving users what they actually want in the algorithmic feed. My understanding is that TikTok is much better at acting on information from users (ie what they seem to like/dislike).
I don’t know how much of figuring out what users want transfers from many short videos to text. For instance, dwell time seems an obvious metric but I find I ‘dwell’ longest on either things I find enjoyable or on things I find terribly boring. I hope it does transfer well – exposing people to more of the kind of writing they want seems generally good to me.
>I think they innovated in giving users what they actually want in the algorithmic feed.
People underestimate this. TikTok's algorithm makes the app enjoyable to use while most other algorithmic feeds make their respective service less enjoyable. I imagine that is because TikTok factors in other criteria beyond active engagement. At this point we all know it is easier to trigger active engagement through negative emotions. By simply prioritizing any engagement, algorithms like the ones used at Twitter and Facebook end up prioritizing content that delivers us those negative emotions in order to get us to engage. TikTok's algorithm also factors in passive engagement like the dwell time you mentioned. That ends up creating a more positive user experience because it actually learns what we like instead of just what makes us reply angrily.
I wonder if part of this is that TikTok is less focused on engaging with others and more focused on receiving content from others (I don't use TikTok...). When on platform communication happens it can be more personal as it is done through videos rather than text, where it doesn't seem like there's a person on the other side.
I think some of it is just that lots of algorithmic feeds take things like view count as too strong a signal and just promote generic content that has mass appeal rather than more niche content.
Algorithmic feeds have totally destroyed Facebook for me. And are now doing the same for Instagram (suggested posts). These are really the only social media I use. Facebook was pretty good when it just showed my friends on a timeline. When they switched to a feed it became a disaster and I had to leave it.
I don't like user-generated video content so I rarely use YouTube and I've never touched TikTok. But I don't think algorithms really help. I really prefer the user selection here on hacker news.
I miss stumbleupon dearly. I suppose that was sort of the experience Tiktok gives people now; I had no control over what the next site would be, I just clicked a button and got something. My "feed" was largely dominated by nerdy webcomics and I loved it. Eventually, I found the quality was degrading and I switched to reddit in about 2010 or maybe 2011, though it was a while before I actually made an account there.
And TikTok. It really depends what Tok you end up on, but there's plenty of corners of TikTok that are exactly the same mechanics as Twitter: quick bursts of infuriation that activate you and leave you wanting more. I think it's easier for a video-based content network like this (these are not social networks) to have content where you turn off your brain, like TV, than it would be with text. But they both can and are incentivized to do the same thing: titillate.
> there's plenty of corners of TikTok that are exactly the same mechanics as Twitter: quick bursts of infuriation that activate you and leave you wanting more
If you start following tags or certain people you'll visit the dusty corners, but the FYP is remarkably good at filtering out incendiary stuff, at least in my experience. I get basically all good vibes, nothing like Twitter/Reddit, and I'm never angry or high strung after watching TikTok.
I'd argue that text has an advantage over video: speed of skipping past content you don't want. People don't talk anywhere near as fast as they skim headlines
You're right about turning off your brain. But not everyone wants that thing turned off
I'm not sure that's accurate. I used to be a huge advocate of text, but don't forget the whole "a picture is worth a thousand words" thing.
If I see a random blog with text, I have no idea of anything regarding it without first reading (or skimming) it. But within one second of watching a YouTube video, for example, I can know a lot about the video - how professional it is, the "genre" it's in (e.g. BookTube vs Fitness), etc.
Plus, jumping around a video of an explanation on how to do something with the computer is actually fairly easy. It's easy to jump past the "here's how to install the app" part and straight to the "here's someone with a code editor open editing the thing I care about".
Twitter only works because of the character limit. And even then the “for you” page is pretty bad and widely disliked, at least as a stated preference.
That's the idea, of course. But the definition of 'article' is pretty vague, which opens the door for all kinds of garbage content. If you shut your brain off and just let the algorithm feed you, you're going to end up swimming in mindless drivel. Maybe this will be better, but I think of lot of people are just going to find themselves in propaganda echo chambers.
It would be better were that true. But sadly, there's are hair-think veins of incredibly useful discussion happening there as well. I have tried in the past to construct in Twitter a timeline that showed me just the very informative or thoughtful comments but it always ends the same way - my worst interests gratified, my better ones unsated.
i think you're just describing addiction. there's always some elusive "value" that you're chasing.
i don't know about twitter, i've never used it, but i find it hard to believe there's anything "incredibly useful" on there. interesting, i could see, but useful?
I've had pretty good success with Twitter. I am careful to limit my followers to avoid cluttering up my feed with stuff I don't want to see.
I mostly follow the math teacher/early education crowd and regularly get inspired with fun math games and other activities to do with my kids. Plus I've gotten a lot of neat books from the library after seeing someone mention them on Twitter.
So for me it's been a good signal-to-noise ratio, unlike e.g. Reddit where I can definitely relate to that elusive value-chasing you describe. Even on HN, the amount of relevant stuff I learn is low compared to the sheer amount of time I've spent here over the years.
> The app opens to a feed of popular articles chosen from a curated list of publishers ranging from leading news organizations like The New York Times to small-scale blogs about niche topics.
Not Twitter but many of these articles already have headlines catered to a certain method of consumption (aggregation done by Reddit, HN, etc.). I didn't need to read this article to know that Instagram's co-founders are making TikTok for text!
I get that the slot-machine nature of the algorithmic feed is what makes TikTok work, both from a user perspective and commercially. It's nice to shut off your brain and just consume during downtime. But lately I've been thinking that I want the opposite (no, not TikTok for text): Something where I know exactly what is coming next. There is a ton of engaging and useful content on YouTube for example, and I think making it more accessible would be a superpower.
Remember how in the olden days there used to be manually curated web directories? They didn't scale so they were replaced by search engines. But imagine a huge, meticulously curated directory of all video content (maybe sans memes). It would use machine learning to tackle the scale, which was not possible before.
You could be watching a video of a recipe, and then go to variations of the same recipie by different cooks. Or the cook is using a technique like deglasing, and you could go to other videos describing it in detail. You'll be able to drill down to niche content. For example "Woodworking > Furniture Building > Jointery > Ornamental Joints > Butterfly Inlays". Contrast with YouTube where the interesting videos are juxtaposed with clickbaity "I bought this silly tool from Aliexpress, look what happened next" and you are tempted to go off rail all the time.
I mean it is a question of optimizing for clicks and commercial success, versus for usefulness and sanity. The former is heavily incentivized in our economical system, but I really hope that we'll be able to build products that fit in the latter category. Even if they remain in a niche.
I do something similar with youtube. I choose a video from a content creator I like. For example vox or kurzgesagt. Then from the first video I choose I make a playlist with usefull youtube recommendations or search result to similar themes.
What I want to say is to make a curated playlist is something that is made realy fast for oneself.
Spotify had once this killer feature that you could start a radio based on playlists. So i mixed some crazy playlists and the radio did get some interesting results. Something similar for video would be nice
Edit: there is something called couriosity stream. Its a website from some youtube creators. I gave it a try and it was nice. But many content missing
Exactly. The comparison also ignores the distinction that TikTok is user-generated, whereas this sounds like it's leveraging publishers. This is an insanely crowded space with a landscape littered with competitors: Apple News, Flipboard, Post.News, SmartNews, and so on.
Exactly. When the article mentioned that it curates content from popular sites like the NYT my eyes rolled to the back of my head. The last thing I would ever want is to read yet another article from the NYT. They (and hundreds of other similar sites) don’t need boosting, curation, or even more distribution.
>Purely algorithmic feeds work for TikTok because people want to shut their brains off when they use the app. It's like a drink after work. Copy-pasting that logic to longform text misses the differences in the mediums, and how people interact with them.
Twitter seems to get users just fine (before the musking at least).
Hell, people come to "brainless fun" subreddits just fine too
To be fair, there isn't just one type of reading - casual, inspectional, analytical, synoptical and etc. I also read in multiple channels and each has its own merit.
There is a wide spectrum of intents, and you read (text) / watch (videos) with an intent. Some people go on YouTube to binge watch but some people go on it to learn. It all depends on your intent.
I would say that this can help with discovery of content that you're interested in. Similar to hacker news, just an alternative approach.
I'm interested to try it and have joined the wait-list. I'm looking forward to how it differentiates itself from other article curation apps (Feedly, Google news, etc.).
I am skeptical of this line of thinking, even for something like Hacker News. It is the same thing that people say about TikTok, and pretty much every other preceding social network/content distribution app. "Discovery of content" implies to me that there is some kind of search. It implies an amount of self-directed, self-motivated activity. I would think that a PhD researcher or a programmer on stack-overflow or a person in a library is "discovering content"
However, on these algorithmically informed apps, the thing doing the discovering is the app, not you. It is discovering the bare-minimum amount of interest that it can provide such that you don't leave. If the app showed you only things that you are the most interested in right away, that would mean that each new piece of content was less interesting than the previous. You would open it up, spend 5 minutes, and then leave. Instead, it learns how to intermittently reward you with interesting content and figures out how much filler you will put up with without exiting.
At least on hacker news and other forum-based sites, you have the ability to only click on links that you want. I regularly open the site see there is nothing for me and leave within 20 seconds. Facebook is more controlling as you can only see a few posts at a glance. TikTok is completely controlling as there is no realistic 'browse' experience. You can see someone's profile, but since the content is video you can't really preview anything or know what it is going to be like without diving into the one-after-another-no-breaks feed.
The over-all point here is just to remember how passive a process using any of these applications are, and to remain clear-headed about it.
> "Discovery of content" implies to me that there is some kind of search. It implies an amount of self-directed, self-motivated activity. I would think that a PhD researcher or a programmer on stack-overflow or a person in a library is "discovering content"
This confused me until I realized that you're using "discovery" like a lawyer would:
discovery (noun)
1. The act or an instance of discovering.
2. Something discovered.
3. The compulsory disclosure to the opposing party of factual information or documents relevant to a lawsuit prior to trial.
I don't know, curation is hard. I think that 90% of what I read is average at best.
Newspapers quality has been going downhill really fast with the shift to digital first, the pressure to generate clicks and the ongoing move to video content. My few journalist friends lament that they can't do their job properly anymore. Long form is probably the last remnant of genuine quality in the field.
Aggregation platforms tend to work quite poorly. If I'm honest I mostly come to Hacker News to mindlessly waste time and because watching a virtual tally rises feels good to me. Encountering interesting and unexpected links happen but are rarely and I have a hard time remembering the last time I had a trully insightful conversation. Meanwhile, Reddit target demographic skews quite young and the time of life where I enjoyed having long winded discussions with students convinced they are experts is far behind me now.
I think there is a place for a well curated feed. They will live or die by how good it is however and the applications that came before like FlipBoard don't inspire confidence in the viability of the idea.
One might say that sites like reddit and hacker news are just archaic versions of the same recommendation engine, but the friction is actually an important feature, and keeps the experience from devolving into lowest common denominator clickbait.
I'm not certain about this seems like it would depend on the algorithm in question. In any case not certain I understand exactly what the friction you're referring to is, and how it helps devolution.
A form of this that leverages the social graph would be great. Reddit is a little too democratic, any bot can vote.
I'd like a system where the comments are weighted such that I'm more likely to see comments from users I upvote/find insightful. Extra points if the users overall views differ from my own.
Reddit is also ignoring the time that users spend on the content/article itself which is a very important signal.
Idk, I’ve seen some peoples fyp that are primarily text with maybe some background music. So I could see this artifact thing working for those people, and maybe catching on enough to work for others as well
I think “we” might be able to turn off our brains when we consume shit tier media.
For most people, in my most humble opinion, that is the base case and they are now addicted to these apps.
Turning the brain on is hard enough. I know it is for me since I’m not a smart guy. It took a lot of work to get better/disciplined and I know many people like my peers from the past just don’t think this way.
It’s sad really. These tech companies mushed the brains of so many people. This is probably our generations’ opium crisis. Ironic that it was done in reverse to the west.
> Purely algorithmic feeds work for TikTok because people want to shut their brains off when they use the app.
Yes, if I'm half-freezing and my train is like 15min late something like that might actually come in quite handy... but that's about where its usefulness ends for me.
>but the friction is actually an important feature, and keeps the experience from devolving into lowest common denominator clickbait.
This is a great insight, I think you are 100% spot-on here.
There are needs where a focused recommendation algorithm like TikTok can still be helpful. And that is to go through a daily flow of financial news which might or might not affect your portfolio and/or might provide additional opportunities. Yes. it is possible to read all of this through an RSS feed or Bloomberg, or Twitter, or Reddit. But none of the have useful recommendation algorithms that can surface personalized information.
> When I'm reading, I'm trying to be thoughtful, not titillated.
This might not be universal. I read for fun, I guess titillated is fun too. There are certain niche topics on which I don't think I can find tiktok videos. But I still want to know about them and many times there are articles, blogs etc available. Not all reading needs to be about the gung ho race to better yourself. Reading can be trashy too, and I love it.
> Purely algorithmic feeds work for TikTok because people want to shut their brains off when they use the app.
I and many like me are actively engaging our brains while we use TikTok. We are intentional about training the algorithm through all the various signals and use it to bring us the content we want to discover. Personally, I would never use it with my brain shut off.
It could work really well for short-form text. I'm addicted to reading comments. I even read your comment before reading the article. A feed with short articles curated to my interests and attention span would absolutely get me hooked.
How many times has a friend sent you a wall-of-txt, and even though this is your good friend, you still say "aint nobody got time for that" and you dont read it.
This is a stupid, knee-jerk attempt. Lets hope Im proven wrong as an old curmudgeon...
When I'm reading, I'm trying to be thoughtful, not titillated. And part of being thoughtful is consciously choosing what to read.
One might say that sites like reddit and hacker news are just archaic versions of the same recommendation engine, but the friction is actually an important feature, and keeps the experience from devolving into lowest common denominator clickbait.