[HN Gopher] And you will know us by the company we keep
       ___________________________________________________________________
        
       And you will know us by the company we keep
        
       Author : jger15
       Score  : 171 points
       Date   : 2021-10-11 16:06 UTC (6 hours ago)
        
 (HTM) web link (www.eugenewei.com)
 (TXT) w3m dump (www.eugenewei.com)
        
       | strogonoff wrote:
       | > 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.
        
       | mooreds wrote:
       | Like always, Eugene blows my mind.
       | 
       | 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.
        
         | toss1 wrote:
         | > 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.
        
         | xmprt wrote:
         | > 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.
        
           | jjoonathan wrote:
           | 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.
        
         | Androider wrote:
         | 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.
        
           | beamatronic wrote:
           | I want less of a social network and more of a time series
           | database
        
           | jaredsohn wrote:
           | Agree, although this might have the effect of scaring someone
           | from posting again after a long while.
        
         | RobRivera wrote:
         | 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.
        
           | mooreds wrote:
           | 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).
        
             | nitrogen wrote:
             | 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.
        
               | mooreds wrote:
               | Agreed, circles was conceptually great, just too much
               | work (IIRC) and a bit late to the social network game.
        
         | mdoms wrote:
         | So the guy favors content over relationships? No thank you.
        
         | thomasahle wrote:
         | >> 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?
        
         | speedybird wrote:
         | 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.
        
           | bobthechef wrote:
           | "Big Mother"
        
             | saltedonion wrote:
             | This sounds stupid but is indeed as troubling as the big
             | brother
        
           | wolfram74 wrote:
           | 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.
        
         | PragmaticPulp wrote:
         | > 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.
        
           | bongoman37 wrote:
           | 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.
        
           | mooreds wrote:
           | > 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.
           | 
           | Good point. That's why making it automated would be even
           | better. Avoids the Larry David moment the author mentions.
        
           | brazzy wrote:
           | I.e. it's not _mechanically_ hard, but psychologically hard -
           | or even, wait for it, _socially_.
           | 
           | But that, too, is something which a platform's UX design can
           | influence to some degree.
        
           | angelzen wrote:
           | Thus we became compulsive social connections hoarders.
           | 
           | https://www.google.com/search?q=compulsive+hoarding&tbm=isch
        
       | anejrcc wrote:
       | 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.
        
         | burnished wrote:
         | What is a chuck?
        
           | AutumnCurtain wrote:
           | I assume he meant "chunk"
        
             | ChefboyOG wrote:
             | Alternatively, we delete comments by whichever user's
             | screen name reminds us most of the name "Chuck"
        
             | burnished wrote:
             | That makes way more sense. I thought it was something rude
             | but I couldn't figure it out.
        
             | 1270018080 wrote:
             | What does cull a chunk mean?
        
               | burnished wrote:
               | 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?
        
       | Invictus0 wrote:
       | 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.
        
       | bobthechef wrote:
       | 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)?
        
       | bdangubic wrote:
       | 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.
        
         | TeMPOraL wrote:
         | 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.
        
           | bdangubic wrote:
           | 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.
        
       | hbarka wrote:
       | I wish I can ride onto an interest graph without having to wait
       | for the algorithm to decide it for me by inference of who I (have
       | to) follow.
        
       | thatoneguy wrote:
       | Google patented a social graph that decays based on level of
       | interactions awhile ago:
       | 
       | https://patents.google.com/patent/US9020965B1/
        
         | Grimm1 wrote:
         | 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.
        
       | renewiltord wrote:
       | YouTube is another company that builds an interest graph without
       | a social graph. I think I would often like to edit my interest
       | graph myself.
       | 
       | It's not that I don't like being targeted. I just want to help
       | them do it right.
        
         | laurent92 wrote:
         | I've always wondered whether annoying videos (or rather,
         | thumbnails) were necessary as part of making us notice the
         | upsides.
        
       | swamp40 wrote:
       | _" 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.
        
         | dkarl wrote:
         | 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.
        
       | simonebrunozzi wrote:
       | As expected, another amazing post by Eugene. This might be the
       | best sentence:
       | 
       | > One of my favorite heuristics for spotting flaws in a system is
       | to look at those trying to break it.
       | 
       | I use this "technique" quite often and I kind of think of it as a
       | secret mini-weapon, and I loved seeing it mentioned here.
        
         | shoto_io wrote:
         | Can you elaborate on your secret? ;-) maybe give an example?
        
           | Jemaclus wrote:
           | 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.
        
           | simonebrunozzi wrote:
           | I thought it was quite clear: look at people trying to break
           | a system, it highlights flaws in the system.
        
         | dapids wrote:
         | 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.
        
       | Tomorrw wrote:
       | Is Reeds law holding up?
       | 
       | Whats the latest on that? Thought the Dunbar number prevents
       | exponential scaling.
        
         | xmprt wrote:
         | 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.
        
       | himaraya wrote:
       | Eugene could really use an editor for his top-notch insights.
        
         | icelancer wrote:
         | He has one. Perhaps it's not enough.
        
           | himaraya wrote:
           | His writing needs more aggressive editing, for sure.
        
         | bawolff wrote:
         | It's a bit rambly, but that seems more like a stylistic choice
         | than a flaw.
        
         | mdoms wrote:
         | 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.
        
       | themark wrote:
       | Shout out to https://www.trailofdead.com/
        
       | Jemaclus wrote:
       | 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.
        
         | wenc wrote:
         | > 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 an fascinating insight. Facebook is optimized for in
         | groups whereas platforms like 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 is
         | freshened (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.
        
       | da39a3ee wrote:
       | 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?
        
       | skmurphy wrote:
       | 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.
        
       | Lammy wrote:
       | I've never seen anyone use the phrase "social capital subsidy"
       | before but really like it.
        
       ___________________________________________________________________
       (page generated 2021-10-11 23:00 UTC)