s17e02: How It’s Made; Blocking is Healthy, Actually
0.0 Context Setting
It’s Tuesday, 9 January, 2024 in Portland, Oregon, where it’s wet and getting colder1 until the snow comes this weekend.
Two things today.
0.1 Let’s work together!
My consulting calendar has opened up for 2024 and I’m open to new clients.
1.0 Some Things That Caught My Attention
To make up for yesterday, I’m going to try super hard to make this one short:
1.1 How It’s Made
How It’s Made2, 3 is one of those shows that, well, shows you how things are made. For Americans of a certain age it’s a bit like a whole set of those Mr. Roger’s Neighborhood factory visits4 that they watched as a kid like how crayons5 or macaroni are made, for British people who’re considerably younger, or watch childrens television, or look after kids who like watching television, the equivalent is when Maddie Moate goes to see how glass7 is made, or sticky.
Other examples in the “how things are made” genre that verge into (sometimes light) entertainment are The Great British Bake-Off9 and Grand Designs10, the latter of which is a sort of How Most People Fail To Learn How To Make Buildings, which amuses me as a reshuffling of a book title11.
Every so often, I get interested in what it would be like to make a documentary about how software is made. It would be wrong to say that these don’t exist -- they certainly exist for videogames, Noclip12 in particular makes fantastic documentaries to the extent that I will not recommend one and say that if you’re interested, you should watch all of them13.
I’m sure there are videos explaining bits of software like “hey, here’s how a newsfeed algorithm works”, I don’t think there are any that go through the process of developing well-known software that include access to the development team.
(Which, a quick googling -- see, this is what you get for reading a newsletter that’s written as I type and I don’t go back and edit turns up at least one example, App: The Human Story14, which was screened in 2017 alongside WWDC and features my wonderful friend Cabel Sasser of Panic)
I think what I’m looking for, though, is a series focussing on specific parts of software. There’s a short version -- the up to 30 minute version, I think? -- that covers something like “how a feed works”, and probably doesn’t include development teams, but then there’s also the longer, personality-driven expert commentary version that’s more like Grand Designs. Bonus points from me if it also includes that acerbic and withering British style that is able to point out hubris and occasionally even offers praise for teams who work well and get it right (even when getting it right is getting it least-bad).
I have to admit I haven’t watched much Grand Designs. I don’t know if there’s the vocabulary to understand what’s involved in application/software development.
I would say this, it would be really interesting to do a documentary on infrastructure software, to the extent that as I type this, I’m half wondering whether it’s worth a grant application or talking to some funders. “How It’s Made: An Unemployment Website” would be... provocative, at least? I I think there’s a bunch of people I can probably talk to about it right after sending this.
(Wait, last thought: The Wire, but it’s Product Managers and Executives)
1.2 Blocking is Healthy, Actually
Block Party15 was an incredibly useful app when I was on Twitter, and, well, before Elon Musk hadn’t fucked Twitter over by drastically changing how its APIs were managed and charged for. A tool for automatically blocking abusive replies, while I didn’t have to use it much, when I did have to use it (thanks, right-wing racist Americans), it was fantastic, saving me from even having to see a torrent of abuse.
(Pro-tip: don’t succumb to morbid curiosity when using tools like this to see what was automatically blocked. The whole point is for you to not have to see likely distressing content.)
Anyway. Block Party had to change what they’re doing and are now focussed on tools like privacy; yesterday I was a user research interviewee for them yesterday, so Block Party was front of mind.
One of the useful signals a ranking/recommendation system might use is blocking, but from what I can tell, blocks are rather opaque. (Another negative signal is downvotes, which are visible in places like Reddit).
Recently -- especially on Threads, the strangely lacking-in-affordances-for-threaded-conversations short-form text social network launched by Meta off the back of Instagram -- there’s been some talk in my feed about remembering to block people you don’t want to see.
For the last couple of weeks, many people on Threads have been saying (citation needed) that their experience has nosedived thanks to bad recommendations in their For You feed, the default mobile feed.
(This apparently isn’t because of someone recently fiddling the ranking of posts against the probability of you clicking, replying and reposting -- Threads and Meta as a whole are wisely on a code freeze over some of the holiday period, so it’s unlikely anyone’s been messing with it16)
Anyway, you should block people! Or mute them! At the very least they’re tools to curate [sic] your feed, and the feed provider promises that posts from those accounts won’t be displayed to you.
So I wonder what it would be like if blocking as an activity and its results were made more visible. For a period like the last few weeks on Threads, making visible how many accounts you’ve blocked might remind people that blocking is a useful tool. You could even go so far as gamify [sic] the blocking behavior to encourage people to do it. Why not have a recommended daily block target!
One is that being reminded that there’s a ton of accounts in your feed that you’re blocking is a sign that your feed deliberately provocative. You would expect that blocking a ton of accounts would reduce the number of accounts in your feed that you’re likely to block.
The other is that this might incentivize accounts to show off how many times they’ve been blocked, but at least in the implementation in my head, only the blocking-user gets to see their block statistics. It doesn’t make sense to give people a how-many-people-you’ve-been-blocked-by number, I think, if only because some people will want that number to go up, incentivizing abusive behavior. Trolls are gonna troll, probably best not to encourage them.
Anyway, my point about blocking being used as a negative signal to influence ranking. Meta engineering have got a super interesting post about how the Instagram’s Explore recommendations system works17, but I don’t see anything there about how negative signals like blocking are used. For example, are blocks just used at the end stage of the funnel, to filter out any recommendations? It naively seems like that would be the best (least computationally intensive?) place to apply the filter, since the more inputs you’re using for ranking.
More thinking out loud: most of the writing about this talks about predicting engagement events “e.g. someone liking a post”, which yeah, I get, because you want to predict things that encourage, well, engagement to make engagement number go up. But user experience also includes “not seeing shit I don’t want to see”, which also includes “accounts I am likely to block based on other accounts I’ve blocked”. This is a policy issue.
On the one hand, there’s the (paternalistic?) point of view of hiding accounts from you that a (potentially very coarse and inaccurate) model predicts you’ll block, which comes with the downside of “reducing your exposure to peoples’ speech”. On the other hand there’s the danger in error and misclassification, or rather, the degree to which you think misclassification is acceptable, which, I’m sorry to say, inexorably leads us to The Substack Comparison.
Instagram’s help center, for example, is quite (?) clear on what happens when you block someone on Instagram18. It says nothing about using that blocking signal. Instagram’s How Instagram Feed Works article19 covers both Instagram and Threads, and the explicit signals it lists (“may include”) are:
- likes, shares, comments you’ve made
- other peoples’ interaction (likes, comments, sharing, saving) with a post
- “previous interactions with the person posting”
- activity on threads
Where on the continuum of “we make decisions about what you see” versus “you make decisions about what you see” does a platform using recommendations want to live?
The Substack “free speech” argument (and I think the default position of most American commercial platforms these days) is “you decide”. But this is a reminder that the position on the continuum has always been a choice of platform operators.
Substack gets flack not just because of its position in being totally OK with hosting Nazi content (in fact, by being OK with and implicitly endorsing it despite founders saying “we don’t like it”), but because it’s talking out of its ass disingenuously and dishonestly when it’s using “free speech” as a core principle while, yep, staking out a position on the continuum that isn’t quite as “free speech”.
I digress. Blocking is a signal. Users don’t, I think, have a choice where they wish their experience to lie on the continuum of “don’t show me stuff” and “show me stuff”. They don’t have a choice intra-platform, nor do I think they really have a choice amongst commercial platforms. Very strongly opinionated open source/free software/hardcore Northern European ideology social networks are very much gung-ho on “Simple Reverse Chronological Ordering or Die”.
What if we did have a choice about how blocking was used?
What if how blocking was used was more transparently described as a signal (it may well be! Point me to it if you’ve seen it!)
What if blocking were encouraged as a healthy boundary-setting activity? I mean, restricting screen time and usage is, so why not this as well? (I mean, I know why: you have to prioritize product development)
And as a last aside, I know why commercial feeds default to For You, or feeds that aren’t just Following (Reverse Chronological). I think it’s because for most people, there wouldn’t be much to read and if there’s not much to read, you’re not going to come back, and if there’s not much to read, there’s not much to respond to, which also means there’s not much for your followers to read.
This is an understandable problem to have! You want people to use your social network platform at least a bit, and you need to use it “enough”, which includes “putting stuff on it”. You have targets (which are constrained by your sustainability model -- note, not just your business/growth/profit model). If your model is “lots of people spend time on it so we can make money by showing them ads” then yeah, you want a feed full of content to get people to keep coming back.
Most people don’t post. Most people don’t post that often. Most feeds wouldn’t be high traffic.
That might be fine. It might mean that you just use them less. But if you want a product where there’s always something new, then you’re going to produce something like a recommendation-based For You feed.
(Instagram has started to reluctantly, I feel, deal with this by pointing out when you’ve seen everything the people you follow have posted and explicitly says “yeah, that’s it, now you’re seeing recommended stuff. That’s... good? I mean an alternative is to just stop and say “hey, come back later”).
Put this way, it is weird and obviously a business model decision. It would be like if your email app/provider made money by showing you ads in-between emails and wanted you to spend more time in your email app so... inserted “great emails from other people who also send emails” into your inbox? I MEAN, if Substack had a reader (which it does, and which I don’t use, and thus haven’t checked), then hey, it might insert Newsletters You Might Also Like in between or near newsletters you’re reading in the app? Huh, right?
Anyway. Blocking. I should mock up some designs or something.
Clearly writing short ones isn’t working. Sorry.
How are you doing?
How you can support Things That Caught My Attention
Things That Caught My Attention is a free newsletter, and if you like it and find it useful, please consider becoming a paid supporter.
Let my boss pay!
And we’re all getting older / and the clouds are moving on with every autumn breeze... ↩
Scaling the Instagram Explore recommendations system - Engineering at Meta (archive.is), Meta, 9 August, 2023 ↩