I don't have a lot to add in commentary, but almost every time I read one of his essays, I learn something.
Here are some particularly insightful snippets from the article.
> A higher fidelity social product would automatically nip and tuck our social graphs over time as they observed our interaction patterns.
YES! I remember reading about someone advocating for an app to make unfollowing easy (daily, present someone on a social network and if you swiped right, unfollow them). But it'd be even better if the network did it; they have the data, after all.
> Twitter favors pure play Twitter accounts that focus on one niche.
This so much. And it's one of the reasons I struggle with Twitter.
> First, [TikTok] runs videos through one of the most terrifying, vicious quality filters known to man: a panel of a few hundred largely Gen Z users.
LOL.
> TikTok is an interest graph built as an interest graph.
Note that the entire article is built around the concept that western social media companies have used the social graph as a proxy for the content graph, to their detriment.
> It's no surprise that many tech companies install Slack and then suddenly find themselves, shortly thereafter, dealing with employee uprisings. When you rewire the communications topology of any group, you alter the dynamic among the members.
Shout it from the rooftops!
> [On linkedin:] It turns out if you map out the professional graph, not just today but also across long temporal and organizational dimensions, recruiters will pay a lot of money to traverse it.
A friend calls LinkedIn "a rolodex that someone else keeps up to date". I don't think we've seen the peak of LI's value. Such a smart acquisition by MS.
A social network that automatically prunes friends I haven't talked to in years sounds benign. Except I'm quite certain that once they get into the habit of pruning friends without asking me directly, they'll start to distance and/or prune friends for other reasons too. 'The algorithm' deciding who is or isn't a positive influence for me sounds like I'm back in 6th grade and my mother is telling me who I may or may not be friends with. Except 'the algorithm' isn't my mother and I am far less certain that it has the good intentions for me.
a benign version would be instead of the connection being a boolean, have it be a float, with interactions re-inforcing the connection and apathy diminishing it, then your social graph would develop sort of like neurons, links rarely if ever being snapped out right, but connected only tenuously.
The interest graph is built in response to your content upvotes/downvotes. The connections are float values instead of booleans. Here is how these float values are updated:
1. You start off weakly connected to all users and you see generally popular content.
2. When you upvote something - you get stronger connected to other users that upvoted that content before you (you don't connect to those who upvoted after you - this is a proof of work that Eugene wrote in https://www.eugenewei.com/blog/2019/2/19/status-as-a-service). The stronger your are connected to someone => the more weight their other upvoted items have for you => the higher their other upvoted items show up in your feed.
3. When you downvote - your connection to those who upvoted that content goes weaker. And your connection to those who also downvoted it goes higher.
4. Every time someone you are connected to upvotes something, your connection becomes slightly weaker to them. They sort of place a bet that this item is worth your attention and that bet is lost if you ignore that item. So if you ignore those who no longer post useful content or simply post too frequently for you - they will gradually disappear from your list. This prevents the old money problem from the same "Status as a Service" post above.
5. You can put the items you upvote into user-defined collections, so your tech content could go into your "tech" collection and users that are connected to your tech collection will not be bothered by the stuff you put into your "film" collection. This is similar to different boards on Pinterest.
This system is somewhere between popularity based systems (Reddit), subscription based systems (RSS, Twitter) and algorithmic systems (TikTok, Pinterest).
Your connections to other people capture how useful their previous recommendations have been to you (ie, their signal-to-noise ratio) and so can be thought of: how much you can trust them that their future recommendations will be worth your time.
I guess the "easy" solution to that is to have a small fraction of presented content to be randomly selected from outside the echo chamber. Either by presenting completely random content or randomly from contributors that have already been downvoted.
I hope you will forgive my quoting:
"""
You seem to be concerned that the mechanism of LinkLonk will lead to a filter bubble where people would group by opinions they agree with and they would simply be preaching to their own (often misinformed) choir. That is a concern, but I'm not sure it is an inevitability.
Now that you know how LinkLonk works, what type of content would you personally upvote:
1. Content that does a good job at persuading the reader to agree with the opinion you hold, even if that content gives little useful information to you, or even if the reasoning is a bit disingenuous.
2. Content that informs you even if it does not promote "we are right, they are wrong" narrative.
My hypothesis is that you would be more inclined to upvote content (2) on LinkLonk more, than you would in other social systems (Reddit/Facebook/Twitter).
In existing social systems, upvoting content (1) is logical because you are trying to promote your opinion to others. The upvotes are used as a tool to win opinion popularity contests. And you are less likely to post/upvote content (2) because it may be seen as non-conformance to the group. That results in the tribalism you mention.
On LinkLonk:
- The main effect of upvoting content is to get more content from people who also upvoted it. If you upvote content (1), then you would personally increase the amount of noise your "future-self" will see (since it has little new information). If you upvote (2) then your "future-self" will be better off.
- Your ability to influence what other people see is limited by how useful they found your past upvotes. You can easily lose that ability if you misuse it - they can just downvote any useless content you try to push. There is no such feedback mechanism in popularity based systems. Every upvote has the same weight. And so it can be misused.
I'm not saying that LinkLonk will avoid the echo-chamber/filter-bubble problem. I'm saying that LinkLonk has a different system of incentives that it is not fair to apply the behaviour seen in other systems to it. And what kind of system level behaviour LinkLonk will have is impossible to predict just using a thought experiment. That's because this behaviour is emergent. The only way is to run an actual experiment - which is what LinkLonk is.
"""
Personally I'd like the exact opposite. The people who post every day, I'm not really missing much if I don't see their latest posts. But the friend who posts once a year? I want to see that post, even if it's from a while ago already and I haven't logged in since.
I would love this. I think it better replicates how in-person group discussions work. I remember being at a workshop and we sat in small group discussions. There was one person there who was the leader of a famous org in the space, much higher org status than all of us there. He spoke very rarely, and when he did, he used few words. I really paid attention when he spoke. That person on FB, Twitter, IG, and others would probably not even make a tiny ripple in the ocean of noise.
I want to receive no communication from many of my friends and extended family for months, and sometimes years, and then, when I do receive something, I want it to be real and thoughtful and written to me personally, not to a distribution. I want to respond at my leisure to these people in the same way, with no expectation of immediacy. There is often more joy in long absence, and finally reconnecting, than there is constantly being connected. And less pressure. The days of very expensive long-distance phone calls and cheap stamps were great.
> I remember reading about someone advocating for an app to make unfollowing easy
It's not actually hard to unfollow people on most platforms, though.
Twitter shows the Unfollow button when I hover over someone's name in my feed. It's one click to unfollow and I don't even have to leave the page to do it.
I think the real problem is a type of FOMO: People are afraid to remove any connections they've established, lest they miss something important.
That sounds like a bad idea. We want to keep our connections and not be unfriended by the machine.
Instead of fighting human psychology the apps should sidestep it. How about unsubscribing, but not unfriending? Shadow-unfollowing, settings like the Youtube bell for your content-stream?
I don't care to see your status-updates, but I want you in my friend list - there should be an option for that.
It isn't just FOMO, its that when you unfollow someone it is seen by the other person has a rejection and people are loathe to reject others. Especially if the other person is important to you in any way, that is a problem. Muting is better, because it doesn't let the other person know you aren't following them.
>> > A higher fidelity social product would automatically nip and tuck our social graphs over time as they observed our interaction patterns.
This is the one that most concerns me. Most people have some tension between versions of themselves; not just the professional/family distinction and the work/play one, but all sorts of things that people are trying to deal with when they're talking about "be the best version of yourself".
Automatic reinforcement machines make that worse. It's like a bartender finding the drunkest people and offering them more drinks; there's a reason a lot of places will cut you off by law or policy rather than feed your addiction.
Social media at its worst is Pavlov A/B testing different bell sounds on his dog.
One of the interesting things about TikTok was precisely that it had "social capital basic income" / anti-reinforcement mechanisms where it will take lesser shown people and show them around to see if they have anything new or different to contribute. But it still has a very strong interest reinforcement loop.
i dont want my network -optimizing- my social graph, and facebook has done this to a point where I need to supplement my social graph external to their product since i simply have different communication profiles for different (equally important) nodes in my graph.
Facebook optimizes the content from the social graph to try to make the derived interest graph better.
I would love if Facebook pruned my social graph. I still have a ton of FB friends from 2008-2010 when I joined FB and was active. But most of those people don't matter to my life much anymore. Why does FB think they do? Because they haven't pruned my graph based on my interactions.
So FB becomes less useful to me.
That said, I think any network with a monolithic graph (or incentives to have one) will run into the problem you outline (where you need to supplement the network with external tools/solutions).
The "circles" idea from Google Plus was actually pretty useful in theory. If instead of binary in/out pruning, a network suggested moving people between "circles," that might be better than just deleting people unprompted.
> A higher fidelity social product would automatically nip and tuck our social graphs over time as they observed our interaction patterns
I feel creator driven platforms like YouTube and TikTok do exactly this. I'm surprised that social media platforms haven't tried building similar algorithms where your recommendations are based off of what you're currently interested in instead of giving a platform to the loudest voices. I wonder if maybe this is a much harder problem to solve with companies like Facebook and Twitter.
YouTube is super explicit about this, and they even "manage up" in the same way that an employee might handle a manager who wants everything to be the first priority: by maintaining a priority list on the top to couple the action of promoting something to 1st priority to the action of demoting other priorities.
>> It's no surprise that many tech companies install Slack and then suddenly find themselves, shortly thereafter, dealing with employee uprisings. When you rewire the communications topology of any group, you alter the dynamic among the members. Slack's public channels act as public squares within companies, exposing more employees to each other's thoughts. This can lead to an employee finding others who share what they thought were minority opinions, like reservations about specific company policies.
> Shout it from the rooftops!
Shout what? That employees should have a public square? Or that employers should keep people isolated, so reservations about company policies won't surface?
> A higher fidelity social product would automatically nip and tuck our social graphs over time as they observed our interaction patterns.
This should not be so binary - follow/unfollow. PArtly because lack of interaction is not necessarily a sign of lack of interest as much as just different phases of life.
Rather, I think it should be more fine-grained and gradual, kind of depreciating down to lower-level follows, and eventually falling off after a bunch of steps. And, the ability to occasionally scan the depreciation levels to see if someone should be re-activated (oh, I've been meaning to get in touch...)
And of course, the distinction Eugene makes between the Interest graph vs the Social graph is spot-on -- it make so much about the dysfunctional social media system so obvious.
While it faceplanted on so many other levels, I thought Google Plus' Circles feature was an interesting way to try and produce and consume content by interest from a single account.
I think it would be more accurate to say he prefers to be able to choose the extent to which the two are commingled - I largely consume my twitter feed for content, keep up with my relationships via assorted more private methods, and haven't had a facebook account in years because its forced commingling made it largely useless to me for either purpose.
Other people may, of course, have different preferences, but I do understand his irritation.
> Western social apps also rely much more heavily on advertising revenue.
I’m surprised this is only mentioned in passing, as it most certainly relates to the problem of “over-attributing how people behave on a social app to their innate nature”.
To paraphrase, perhaps the single most important influence on users’ behavior in social media is platforms’ existential requirement to generate ad revenue from it?
The author mostly misses this factor in favor of overthinking what technically is its consequences. Seems blindingly obvious that such a business model would shape everything from high-level design to implementation to strategy—all that causes social media dynamics to be what they are.
My wife and I were discussing Facebook vs Instagram vs TikTok the other day. Neither of us use Facebook. My wife pointed out that our friends aren't interesting. Their posts are often their kids or their families or politics or something that just isn't that interesting to us. They're still friends, but they just aren't interesting to follow on social media.
On the flip side, I love TikTok. It's full of strangers! New, exciting strangers doing wild, strange things in trendy and funny and smart and annoying and cringe-filled ways! TikTok is an entire world of social media that could never exist on Facebook. (Or maybe it did back when it started? I can't remember.)
Instagram is the more interesting one for me. I observed over the course of this conversation that if you look at my feed, it's boring stuff I don't care about. I follow my friends and they post pictures of their kids and blah blah blah.
But if you click the "Discover" tab, it's almost entirely three of my biggest interests: horses, fitness, and D&D.
To Eugene's point, my feed is my social graph, but my Discover tab is my interest graph.
Like Eugene, one of the things I like about TikTok is that it's an interest graph and I happen to be connected to a few of my friends. But since I spend most of my time on the For You Page, I never see my friend's stuff. That's totally fine with me!
I'd love more social media apps to understand that connecting with my friends doesn't mean I want to know everything they're doing, and instead I want to meet at the intersection of my social graph and my interests. Those two things don't often overlap, but when they do, it's a recipe for fun and engagement.
FWIW, despite having dramatically increased the number of ads they display, I still maintain that Instagram's ads are far and away better than Facebook's, TikTok's, Twitter's, and even Google's. I see an ad on IG and I very often say, "huh, that looks cool" whereas I would just gloss over it on other platforms.
> My wife pointed out that our friends aren't interesting. Their posts are often their kids or their families or politics or something that just isn't that interesting to us. They're still friends, but they just aren't interesting to follow on social media.
This is a fascinating insight. Facebook is optimized for in groups whereas Instagram and TikTok are platforms for creators. Naturally the latter is going to be more interesting especially if your peer group is pretty stable.
I guess I’m unusual in that I move around a lot (different countries, different parts of the country) and my peer group gets refreshed (or gets reshuffled) every few years. I often get to know people who have very little in common with my previous social groups so Facebook continues to be interesting for me.
There’s also something to be said about continually upgrading or changing one’s peer group. Not practical for everyone of course but for me it just kinda happens.
Point of clarification: our friends aren't boring. We love hanging out with them and doing things with them. But from a strictly social media standpoint, their posts aren't the kind of content we want to consume. I can only see photos of your kid so many times before I've seen enough.
We don't need to "upgrade" our friends or our peer group at all.
The distinction I am (and I think Eugene is) making is that my reasons for being friends with someone has no bearing on the types of media I am interested in consuming.
You and I can be best friends, but that doesn't mean I need to see the photos of your breakfast. And the fact that I'd rather watch Hank Green pontificate on scientific facts in no way diminishes my friendship with you. As Eugene points out in his blog post, the social graph and the interest graph are two awfully different things that may wind up being totally orthogonal to one another, and the Venn diagram overlap may be very small in real life.
If I may try and rephrase your takeaway, it's that our social media landscape needs to be refreshed or reshuffled periodically to maintain our interest, and that can take many forms, including but not limited to finding a totally new peer group. I would agree with that assessment.
But reshuffling your interest graph is much easier than shuffling your social graph, at least for most people.
Thanks for that clarification. I think that’s a better distillation of what I was trying to say.
That said I do believe in changing my peer groups from time to time to get out of my comfort zone, reinvent myself and be confronted with new ideas. But that’s just because I’m in a growth phase. Not everyone needs to do this.
Also for me it helps to have friends who are writers and travelers. They may be the same people year after year but they constantly have new things to say about life or always have new observations/stories to share. I do think there’s a productive way to use Facebook. It’s just that that way only avails itself to certain groups of people like me and the kinds of friends I keep.
> Their posts are often their kids or their families or politics or something that just isn't that interesting to us. They're still friends, but they just aren't interesting to follow on social media.
Plus, the life events that would be interesting to me, I much prefer hearing about once every however long over drinks - and even my heavy-facebook-user friends have commented that they really quite enjoy the fact that they get to tell me about things like that face to face rather than just assuming I'll have read it as it scrolls past.
I didn't find it disorganized. There seemed to be a clear progression of ideas and it kept my interest throughout.
Readership isn't everything. The best way to appeal to a broad base is to say nothing of import; i'm not sure a pared down version would be as interesting a piece.
I got about a third of the way through the article and gave up. If there's anything "top notch" in that mess it's certainly not in that portion of the piece.
When I worked in innovation consulting, we used to look for "workarounds" that people would do to find their unmet needs and desires. Seems like a similar concept. I'm glad he mentioned it and that you mentioned him mentioning it :-)
One example might be influencers buying followers. If simply paying money for bots (or people?) to follow you is enough to create "influence," then that may be a flaw in the system.
Another social media example might be finstas, where people create a separate private IG account to complement their public IG account because IG has binary privacy settings: either public or private.
So basically "you spot flaws in systems by watching people break them". Uh -- duh! Some revelation ... I don't think something that is common knowledge is a secret weapon, but to each their own I guess.
Great essay, very insightful, very smart--but zoom out a bit and we get the old adage: The brightest minds of our generation are working on getting us addicted to social media.
> When designing an app that shapes its user experience off of a social graph, how do you ensure the user ends up with the optimal graph to get the most value out of your product/service?
This is the heart of what's wrong. Nobody running these companies gives a flying fuck if the users get the most value out of their service. They want the service to get the most value out of the users.
Twitter allows editing the list of interests it thinks you might have but the implementation leaves a lot to be desired. It's at Settings -> Privacy and safety -> Ads preferences -> Interests.
The interface is presented as a giant list of hundreds of potential interests, all ticked by default, but which must be unticked one at a time. If you start actually unticking them all or try to automate it with a script (which is the only way to stop Twitter from suggesting tweets from those interests to you, until it comes up with some new ones), the unchecking gets progressively slower until it starts throwing 500 Internal Server Errors, showing the state as changing but silently failing.
"But what if there was a way to build an interest graph for you without you having to follow anyone? What if you could skip the long and painstaking intermediate step of assembling a social graph and just jump directly to the interest graph?...And what if the algorithm that pulled this off could also adjust to your evolving tastes in near real-time, without you having to actively tune it?"
I have to say, Facebook has this working well for me. Their "Suggested for you" posts are 99% interesting to me. I have no idea how they do it.
Facebook never worked that way for me, and I always assumed the explanation given by the article is the reason. The social connections I had on Facebook (family, old classmates, wife's friends, etc.) had virtually no overlapping interests with me. On Twitter I choose whom to follow, and on YouTube I choose which channels to subscribe to. The problem with Facebook for me was that it was too warped by social obligations. An old friend from high school sends me a friend request, and his parents have been talking with my parents about how he isn't doing great, eh, easiest to accept it. Oh, shit, they weren't kidding about him not doing great, and he's consoling himself with some weird ideas.
Twitter for me is 98% non-personal. YouTube is completely impersonal. No social interaction == much better fit to my interests.
I think there is a significant opportunity for an app or platform that allows me to make helpful introductions in the right context. I am not aware of any that allow me to curate my network and suggest who to connect (and provide a context for the suggested conversation). Please leave a comment if you are aware of one or working on it or contact me via my profile.
The fact I had that thought at little earlier in this thread before reading your comment leads me to believe two things, one it's a somewhat trivial idea to arrive upon and that two our patent system is broken that something like that could be patented when it's really not a hard thought to arrive at.
I think social media companies need to build two graphs:
1. interest graph
2. s(he) spends money on graph
There are A LOT of things that I am interested in that social media companies cannot monetize on. But there is also A TON of things they can. While number of companies are fairly great at #1, absolutely not a single on is even remotely decent at the #2.
How do you expect successful implementation of #2 to look like? I'd expect it to be pretty similar to the state of Amazon product recommendations.
The problem is that users and platforms have a very different definition of what a good use of a graph is. This comes from their business model - advertising. For #1, you want to see things that are interesting to you, but the platforms want things that will keep you "engaged" the most (maximizing your exposure to advertising, which they sell as a service). The result is, you get content that, while related to your interest, is mostly a bottom-feeding, lowest-common-denominator material geared towards pushing your emotional buttons.
So with #2, again, the platforms may know perfectly well what you're spending money on, but what they'll show you won't be the perfect deals for you - they'll be the worst deals for you, from customers of theirs who spent most money on getting their offers in front of people.
I do not believe for one second (I could be dead wrong of course!) that platforms know what I am spending money on. It is not that I am not getting perfect deals fed to me in ads, it is that I am getting fed things I do not have a slight interest in and would absolutely never purchase. I play tennis and if I was getting fed ads for racquets that cost $500 or tennis ball machines that cost $3k that would be one thing (not a very good deal and I would never spend that much money) but what I am getting are ads for hoola hoops which I would not buy even if someone was selling them for $1.
BigTech can use the social graph combined with additional knowledge they are collecting on you to shape social outcomes in subtle, but controlled ways. For example, if you begin to be seen as unfavorable by certain people or by some algorithm, the information you see might quietly begin to change: the jobs you see posted, the health care will be offered, the schools you apply to. These would not result in some massively obvious punishment for certain behaviors, but in a way that might be viewed as a marriage between shadowbanning and bias in ML models, which is to say that the bias would be fully controlled and quietly implemented in a continuous way.
Now, LinkedIn and Facebook can track social milieus, but are they using that knowledge to direct social phenomena, so to speak? For example, if a group of people in opposition to BigTech begins to coalesce, are they using social media to discourage, shape, and manage the discourse in a way that diminishes the likelihood that such a group can mount a cohesive threat to BigTech (for example, by increasing fragmentation or cooling by increasing the amount of conflicting information different people in that group receive, or impeding the spread of certain kinds of information within that group)?
The Dunbar number is about individual connections. The exponential scaling comes from the number of possible subgroups. I might have 5 close friends but each pair of friends has a different interest. From 5 friends, we have 15 interests. With 10 friends, we can have 90 interests.
Only read first third but seems like a very good essay, even though not everyone wants to read 10k words on Western social media.
I recently tried to use Twitter and came to the same conclusion - I followed some people cos they are interesting on computer science, but why the fuck would i want to see their nature photos from their weekend walk?
HN doesnt build a social graph or an interest graph but its still interesting to me.
Now if they can just do something about comments. Like just randomly cull a chuck every hour or so and it will be a nice energy generating space to hang out.
If it does mean chunk, I think it means randomly removing clusters on the network graph. I'm guessing the idea is that it would promote healthier networks because you'd periodically be removing portions and only the parts you thought were really worthwhile would get reconnected?
I don't have a lot to add in commentary, but almost every time I read one of his essays, I learn something.
Here are some particularly insightful snippets from the article.
> A higher fidelity social product would automatically nip and tuck our social graphs over time as they observed our interaction patterns.
YES! I remember reading about someone advocating for an app to make unfollowing easy (daily, present someone on a social network and if you swiped right, unfollow them). But it'd be even better if the network did it; they have the data, after all.
> Twitter favors pure play Twitter accounts that focus on one niche.
This so much. And it's one of the reasons I struggle with Twitter.
> First, [TikTok] runs videos through one of the most terrifying, vicious quality filters known to man: a panel of a few hundred largely Gen Z users.
LOL.
> TikTok is an interest graph built as an interest graph.
Note that the entire article is built around the concept that western social media companies have used the social graph as a proxy for the content graph, to their detriment.
> It's no surprise that many tech companies install Slack and then suddenly find themselves, shortly thereafter, dealing with employee uprisings. When you rewire the communications topology of any group, you alter the dynamic among the members.
Shout it from the rooftops!
> [On linkedin:] It turns out if you map out the professional graph, not just today but also across long temporal and organizational dimensions, recruiters will pay a lot of money to traverse it.
A friend calls LinkedIn "a rolodex that someone else keeps up to date". I don't think we've seen the peak of LI's value. Such a smart acquisition by MS.