February 14, 2024

s17e11: Context is King; Explicit People Personality Preferences; OS Licensing, but Different

0.0 Context Setting

Wednesday, 14 February, 2024, the day after I made lots of pancakes and ate lots of pancakes.

Quite a bit today.

---

1.0 Some Things That Caught My Attention

1.1 Context is King

Lately, I’ve been thinking about context. Or, I guess, one way is reading the room.

The reply management stickers¹ I made the other day went nuts on Bluesky and Mastodon, they vanished comparatively without a trace on Threads.

The argument against these stickers -- and it’s one with which I’m able to understand the position of the people from whom it’s coming -- is that they’re super tone-policing and that hey, maybe it’s better to manage the replies that are sent, rather than pre-emptively guide conversation.

I mean yeah, that’s certainly an alternative.

But these stickers resonated most with people who’ve experienced the type of replies referenced enough that they would prefer not to, rather than “just ignore them”.

And: there just aren’t great options for managing replies after they’ve been sent. There’s options like “only show them from my followers” or Twitter’s early “yeah, kind of don’t show me offensive ones” which would go into a hidden in the basement inbox, and Blockparty’s (RIP) reply management, but again, that was third-party. Actually, along with Twitter’s example, I think Facebook Messenger used to hide weirdo replies.

But there aren’t user-controllable spam filters for replies. There’s no training for replies. Sure, there’s server-side sentiment analysis and abuse and yeah. I don’t know the details of how that works.

Also, sometimes replies are good! Many times! So there’s the fear that any management system might also hit a false positive (which... you’d probably deal with at the first blush by making sure that replies from mutuals would get through. You’d have a higher confidence, I assume -- but would need to validate! -- that those replies won’t piss you off/be a waste of time/be annoying).

The thing I think the tiny minority of people who were so motivated as to comment/reply about how these were stupid or indicative of an incredibly thin skin on the part of, well, me, and anyone who might use them, was along the lines of “are you fucking kidding me, are you bringing Mastodon’s content warnings and safe spaces to Bluesky, where we just block whomever the fuck we want?”

And, well, no? But also, ish? Again, it’s about being able to guide a conversation, which is something we don’t have great tools or behaviors or practices for in social spaces because of all the dumb context collapse, but even more precisely, that the analogy to the public square is kind-of correct and also familiar. Social feeds are just a big undifferentiated space. That’s their defining characteristic, the alternative is defined communities like Slacks with channels, Discords with channels, Reddits with subreddits, and so on.

But I tell a lie, that’s just one of the alternatives. jwz in my head has a reputation/practice now for needing to include a statement along the lines of “don’t just fucking reply with something you’ve googled, I only want to hear from people who’ve actually used the thing they’re suggesting” whenever he’s asking for help or suggestions about a particular problem. I don’t blame him! If you get a lot of replies and you’ve been around a while, well, there’s etiquette.

Again, this is “trying to manage human social behavior” which, hi, is something we’ve tried to do forever. Some of those tactics range from “scolding” to “yelling” to “ejecting from the community” to “protesting about and ejecting from the community only for the person to return with a triumphant Netflix special”.

But. People have expectations and rooms they want to be in and spaces they want to be in. A thing I’ve banged on about too has been (ha, yet another science fiction reference) Drone auras from Iain M. Banks’ Culture series, where a bunch of machine intelligences display colors/textures to indicate interior states like emotion, or to better/include more dynamic range and metadata (ugh) in their communication.

It’s one of the reasons why I made the stickers different colours. There’s no way, other than through text that uses up your precious character count to indicate that you are tiredly asking a question. Or that you have exhausted many avenues. Or to try to convey in tone that what you’re making is a wry observation and what you don’t want, honestly, are earnest replies. At least, not right then, not for that!

I’m pretty sure I can convey the intent of a conversation through tone. Or, at least, I think I can. Being able to do this explicitly might actually be useful! Back in college, with a bunch of friends who subsequently figured out that they might actually be neurodivergent, there was a ha-ha-desperate-joke that perhaps if you fancied someone and they fancied you², then it would be helpful if you could see a little light above their head like in The Sims. Or even if they were just pissed off with you.

Body language, explicit communication (whether sign, oral, or written) don’t have to be the only ways in which we communicate intended context, especially if these computers are supposed to be so good at breaking down communication barriers, hm?

The shit way of doing this is by inference and automation. I say shit, because it is commonly shit and super good at misclassifying intent, mainly because the people who usually make these systems tend to have narrow exposure to the diversity of human experience, which also normally means “overindex on predominantly white mainstream cis, het communication” in training material. Yadda yadda: why is it that African-American Vernacular English tended to get classified in a certain way?

Sure, the shit way could get better and less shit. But another way would be to allow more explicit setting of context.

“Oh but Dan, that’s just introducing friction!” is what imaginary straw white dude shitty product managers / growth hackers would say to me. “Our goal is to let as many people communicate with each other! Metcalfe’s law! More friction means fewer users and less communication!”

To which I would say: and? So?

And then someone else smarter than me would say: “OK, but that’s because capitalism / the current economic system / the business model is a design constraint.”³

But friction isn’t bad. I could write a whole Medium essay or whatever confidently asserting that Friction is Good, Actually, and deliberately not get across the nuance, which is to say that of course friction is good when it makes sense and the sticking point here is where does it make sense to introduce friction?

In a stunning example of “the business model is the design constraint”, a pattern that continues to exist is “if you want to opt-out of arbitration, you may do so by writing a letter, signing it, and mailing it to this corporate legal department PO box address”. Huh! Friction there, but not in onboarding! WEIRD! I wrote about that YEARS ago! I see designers at The Orange Place complaining about shit like this when latest corporation does it, but again weirdly fewer people saying “hey, so legal asked me to do this and... we’re not going to?”

I will now say something like “the purpose of a thing is what it does” and right now the apparent purpose of most social networks is “let you both spout off and reply as quickly and easily as possible”, some sort of neuralink for the bit of your brain that just comes up with shit to say.

Every so often, the parenting advice of repeating the “Is it kind? Is it true? Is it helpful?” mantra crops up on social media, again, a kind of scold/gentle reminder when, you’ve guessed it, some idiot has not been kind, helpful, or said something true.

And you know what, fine. This gets into a society’s approach to “freedom of speech” and how far along the spectrum you are on “absolute freedom of speech” (i.e. whether you’re someone who actually believes and understands what Elon Musk conveniently doesn’t, but espouses in terms of Mule-like building a cult following) to “encourage healthy speech” continuum. Protocols are supposed to promise us the ability to choose what type of speech we want to produce or encounter. Attempts like “hey, did you actually read this article?” get criticism for ham-fisted implementation, but the intent is at least acknowledged.

So here’s your design challenge. How might you quickly, easily, seamlessly let people add markers to their speech? Of course people are going to want to speak quickly and off-the-cuff. Forgive me that my message was ill-considered, I didn’t have the time to think more about it.

But we do have those physical affordances to create tone and context. ASL is amazing at it, it’s like fucking poetry to watch! So expressive!

Meanwhile the most we’ve gotten has been to include emoji, of which I have to admit is actually a super big deal. You know something in language has made it when there are court cases about what it means. I mean that sincerely.

One suggestion I had was to pair the stickers with iconography, like an emoji. An easy response to that is “but nobody would know what they mean”, but hey: language evolves. People figure out what stuff means through usage. I did, in the footnotes, say I was a descriptivist after all.

That said, pairing with emoji -- especially with facial-expression based might not be the most accessible. It comes back to reading a face, and that’s more difficult for some people than others. So perhaps abstract iconography mapping to intent might be better. Who knows! Feels like it would be fun to find out and do some experiments.

Explicit People Personality Preferences

Another thing I’ve been thinking about for a long time has been emotional range. Or tone? You know what I mean: it’s the thing I go on about with the relentlessly crush-it American/West Coast/hustling tone of voice for notifications and in-app copy, that sort of YEAH YOU CLOSED YOUR RINGS YOU MOTHERFUCKER AREN’T YOU FUCKING CRUSHING IT.

And that maybe, just maybe, for a variety of reasons (which you don’t need to disclose!) you want a different tone.

One way I’ve joked about this is in reference to Douglas Adams’ Genuine People Personalities from the Hitchhiker’s Guide to the Galaxy. There weren’t that many examples in the book, really. Off the top of my head:

• Eddie the exuberant ship computer;
• Doors that just really got emotionally satisfied by opening and closing;
• Marvin, of course, who was depressed and the result of applying Moore’s Law to Eeyore;
• That tank that really liked blowing stuff up but, in true Star Trek style, could totally be logic-puzzled/argued into doing something super dumb;

The other recent example I admire and also consistently reference is the robots TARS and CASE from Christopher Nolan’s Interstellar of which, from the novelization:

When he returned to the farmhouse, he found that a new power supply had been brought, as promised, and so he began the work of bringing Tars back to life.

“Settings,” Tars said. “General settings, security setting—”

“Honesty,” Cooper said. “New level setting. Ninety-five percent.”

“Confirmed,” Tars replied. “Additional customization?”

“Yes,” Cooper said. “Humor seventy-five percent. Wait… sixty percent.”

I really like these. Some thoughts:

I think emotional range might be relatively easy to implement? They are just like internationalization strings, so “just” try adapting the approach. You “just” need to map your emotional states (e.g. “humor level as percentage”) to your emotional range translation.

The other part is specifying emotional range settings. You could set them per app/context (hey Apple Fitness in particular, stop being such a techbro), you could set them bounded by time, and so on.

Apple Fitness has a deal where it monitors your activity level and every so often will offer adjusting it at the end of a week. You could try the same thing with tone? I mean, even the app permissions model on iOS does this -- after a while using an app with location permissions, the OS will pop up a modal saying “hey, Doordash has been using your location for the last 3 months this much, do you want to keep letting it only while you’re using the app, all the time, or turn it off?” That’s nice!

The nice (ha!) thing about this is that it totally creates a bunch of work! I’m thinking of every single bit of microcopy in every single offending web app like new feature bugs that honestly I want an extension that turns the dismiss label into “fuck off”. Just think! Three! Five! Ten versions of every single bit of microcopy, customizable to how irritating or funny you want your app copy to be!

I guess people will create the different versions with LLMs. (BAD LLM! NOW DO IT 10% LESS PUMPINGLY) Now I’m sad.

1.2 OS Licensing, but Different

OK, I was making breakfast this morning and puttering around the kitchen. Making stuff on the stove that needs timers, and:

• yeah, sure, those screenshots of people using Vision Pro and the spatially-situated timers next to the things like pots etc sure look helpful, if you ignore the whole implementation issue of needing to wear a stupendously early-generation face computer⁴;
• ok, but maybe just screens on stovetops and fridges and so on;
• ok, but have you seen the disaster with screens and computers in kitchen appliances? It’s terrible! They’re terrible! Not only does putting a networked computer in something drastically increase the chances of enshittification⁵, but most kitchen appliance manufacturers are terrible at making computers! It’s not a, how do you say it, core competency! Hell, even making reliable kitchen appliances appears to no longer be a core competency, especially of the brand-shells!
• wait, there’s other industries where manufacturers have decided to just give up;
• like cars;
• right, there’s Apple CarPlay and Android Auto;
• the car manufacturers just straight gave up and said sure, we’ll do our own stuff, and we guess it’s also a selling point that people can use their pocket supercomputer with a consistent user experience, apart from if we’re GM or whomever or maybe Rivian or Tesla where we’re actually good at computers;
• this strategy of extending an operating system to another manufacturer is... a bit like the new version of operating system licensing? Like, Apple’s figured out how to get “iOS” on more things without ceding as much control as the clone era?
• so join the dots and imagine something like KitchenPlay, where appliance makers can just have dumb screens that can run iOS applications on them.

Boom.

---

Phooey. How are you doing? I love getting notes. Even when they're just "hi!"

Best,

Dan

---

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!

Do you have an expense account or a training/research materials budget? Let your boss pay, at $25/month, or $270/year, $35/month, or $380/year, or $50/month, or $500/year.

Paid supporters get a free copy of Things That Caught My Attention, Volume 1, collecting the best essays from the first 50 episodes, and free subscribers get a 20% discount.