The Affordance Loop

A black-and-white postcard photograph of more than a dozen men standing on tall stilts wearing berets, many in rustic sheepskin coats. The stiltwalking French shepherds in the marshes of the Landes are incredible, and I love this postcard—which you can buy from a nice person on eBay—in particular.

Affordances, in the simplest terms, are what an object offers or provides to a specific individual at a particular moment in time. This sense originates with James and Eleanor Gibson, a pair of celebrated researchers in psychology. James Gibson lays out a wonderfully informal definition in one of his later works:

The affordances of the environment are what it offers the animal, what it provides or furnishes, either for good or ill. The verb to afford is found in the dictionary, but the noun affordance is not. I have made it up. I mean by it something that refers to both the environment and the animal in a way that no existing term does. It implies the complementarity of the animal and the environment. The antecedents of the term and the history of the concept will be treated later; for the present, let us consider examples of an affordance.

If a terrestrial surface is nearly horizontal (instead of slanted), nearly flat (instead of convex or concave), and sufficiently extended (relative to the size of the animal) and if its substance is rigid (relative to the weight of the animal), then the surface affords support. It is a surface of support, and we call it a substratum, ground, or floor. It is stand-on-able, permitting an upright posture for quadrupeds and bipeds. It is therefore walk-on-able and run-over-able. It is not sink-into-able like a surface of water or a swamp, that is, not for heavy terrestrial animals. Support for water bugs is different. —The Ecological Approach to Visual Perception, p. 127)

(Support for water bugs is different!)

In UX/interface design, conversations about affordances tend to revolve around the cues designers can offer to reveal what a system actually offers the people who use it. Those conversations, paired with mass use of the online and mobile interfaces coming out of especially Apple, Microsoft, and Google resulted in the modern interface design vocabulary we can take for granted until we’re doing phone tech support with someone who doesn’t distinguish a webpage from a home screen from a preferences panel.

On the social internet, people who have used the biggest platforms and networks enter new ones expecting to find ~standard affordances and expecting that familiar interface cues will map to familiar affordances.

When newer systems and tools confound those expectations, people get, at best, confused. At worst, they try to walk across a solid-looking but sink-into-able surface and get stuck in a bog, while stilt-walking locals familiar with the marshy ground cluster around to chide them for acting like—well, water bugs, maybe.

Both online and off, I am comically bad at spotting secondary or subtle affordance cues. If there is a door with a horizontal bar and a sign that reads PULL,” I am going to run full-tilt into that glass. Vertical bar and a little push” label? I’m gonna stand in the hallway pulling like Babe the Blue Ox and wondering aloud why the door is locked—real ninety-ninth percentile performance on not getting it. This, plus my design-adjacent background, inclines me to listen to other peoples’ complaints about confusion even when I personally find a system easy to use.

The class of design problems I (literally) run into manifests when an object’s actual affordances don’t match the main visual/perceptual cues it offers, as Donald Norman lays out in his foundational and genuinely great Design of Everyday Things. Norman’s humanist approach to these problems was a major contributor to the emergence of user-centered design (now human-centered” design, which is nicer). It also resulted in the doors I run into getting tagged Norman doors in his honor, which seems like poor thanks for a lifetime of good work.

The people who make casinos, malware, chum-box ads, and other scams use Norman door patterns intentionally, to deceive, trap, and exploit. And although people in the US-based culture I come from condemn those uses, we tend to consider accidental deceptive affordance cues merely thoughtless, much as we prefer to punish those who kill people intentionally much more harshly than those who kill by accident or negligence, even though the person is just as dead. In practice, this emphasis on intent produces cultural systems in which no one is responsible and lots of people die.

Legally, we balance our hyper-valuation of intent with the duty of care, but in daily life, duty of care gets ideologically clotheslined by personal responsibility and liability-reduction theater, like the tiny-print waivers we sign to do anything mildly risky, or the Montana school district that chose to rotate students around the classroom every 15 minutes during covid restrictions to avoid triggering quarantine rules. This orientation saturates the tech industry, which responds to more active constructions of duty of care with things like foot-dragging cookie warnings.

It’s not surprising that few commentators apply the idea of the duty of care—even informally—to social platforms and communities, but once you start assessing online conflicts and disasters through that lens, it gets hard to un-see the breaches.

In an interview with Justin Hendrix on the Tech Policy Press podcast The Sunday Show, Dr. Johnathan Flowers digs into affordances on the social internet:

…by affordances I mean the resources, the tools, the very structure of the website—things like hashtags, retweets, quote tweets, and so on. The ways that a given community forms by means of the affordances of a digital platform structures some of the nature of that community.

As André Brock notes in his book, Distributed Blackness, Black Twitter as a phenomena emerged by means of the affordances of Twitter, through how things like quote tweets, retweets and hashtags allowed Black users to engage in Black digital practices, which are the performances of Black offline culture by means of the affordances of the digital platform of Twitter. So things like call and response, playing the dozens—which are culturally-mediated practices within the Black community, to engage in information sharing, in community building—were all enabled by the features of Twitter such that Black folks could engage in digital practices that made present their identities as members of a community.

(Disclosure, I’ve very lightly repunctuated the transcript excerpts for readability, based on my repeated listenings. Flowers speaks in dense paragraphs and makes it sound easy, and I want to represent his work as well as I can here.)

This is one of several I think highly relevant takes on affordances in the interview, and I encourage interested readers to listen to or read it in full. In this post, I’m mostly going to talk about specific results of specific affordance—and affordance-cue mismatches—but I want to start by nodding toward two subtler points Flowers makes. The first pulls in Sara Ahmed’s work on the way whiteness functions in the world to talk about the underlying cultural patterns that shape online spaces:

…if you have a space that is predominantly populated by white persons regardless of their other identities, if you are in a space primarily populated by white persons, the norms, the habits, the very structure of that space will take on a likeness to whiteness by virtue of how the majority of people participate in that space. As I said, Mastodon is a very white space. It is not unlike other tech spaces where whiteness is predominant. Insofar as this is the case, the norms, the habits, the affordances of the platform will inherit whiteness.

Inheriting whiteness” here has direct bearing on the kind of norms-based conflicts that occur and recur when white Mastodon users try to discipline Black Mastodon users for discussing racist oppression—or even mentioning race. Here’s Flowers:

Insofar as the majority of the users on Mastodon are white, then they take up the kinds of ways that whiteness organizes space—including an entitlement to freedom from say, understanding one’s complicitness in racism, or freedom from engaging with experiences of racism as made present by users of color.

The last point I want to pull out of Flowers’ interview is that the underlying cultural patterns of Mastodon influence not only the affordances that get built, but also the norms governing how people are permitted to use those affordances. This shows up in what is socially required, as in you must use content warnings for mentions of your experience of racism,” and in what is socially forbidden, as in linking to another post is just the same as a quote-tweet and we don’t do that kind of thing here.”

So: Patterns, affordances, and norms.

Some of what’s on my mind is about Mastodon/fedi, which has lost tempo on a lot of the parts of the Twitter migration that I’ve cared most about—Black Twitter, literary Twitter, radical journalist Twitter, while its someday-maybe-rival Bluesky has imported at least some of those communities despite obvious problems.

Mastodon’s userbase remains divided on whether the network’s niche status is a good thing. Plenty of people on Mastodon are committed to making both Mastodon and the broader fediverse (the apps and services interconnecting via the ActivityPub protocol) easier for a mass audience to use and understand—but the opposing point of view is also very present, and often very heated. The view that e.g. Twitter users have proven themselves to be Nazi sympathizers who should be excluded from the fediverse is a minority view, but it shows up like clockwork. So do claims that anyone who wants greater ease of use on Mastodon is really just mad about not having a billionaire overlord to complain to—or that finding Mastodon confusing is a smokescreen for clout-chasing and influencer culture and loving capitalism.

A little further behind those complaints, you find the people making extremely forthright claims that Black people who discuss race are the real racists and don’t belong.

One of my very first exchanges in my 2022 re-attempt to get situated on Mastodon was with a scrupulously polite young man who hoped that Mastodon could repeat the successes of Gamergate in gatekeeping” the gaming world to keep it healthy. In hindsight, this was heavyhanded foreshadowing that patterns of intentional, active, and probably coordinated exclusionary tactics were already entrenched in the fediverse—and not only on the known trollfarms and abuse hubs that even moderately well-run instances block.

There are subtler and better arguments about the way security through obscurity (non-derisive) and interface friction can provide provisional refuge for people targeted for abuse. I think people making those arguments usually have extremely valid concerns, especially but not only when they talk about fears of Meta’s upcoming federation. For today, I’ll just say that remediation work on confusing Mastodon features and flows should always either preserve the existing affordances that offer shelter and refuge to existing users or replace them with better ones that provide the same benefits. (The best way to do this would probably be to engage in large-scale user research and participatory design, but I say that about everything.)

As I initially wrote back in April, there are plenty of reasons to avoid Bluesky, the most obvious of which is that Jack Dorsey was involved in the initial funding and remains on the board. Nevertheless, I came in a little bit optimistic about the infrastructure Bluesky’s AT Protocol might be able to offer intentional communities in a federated future. Since then, things have…not been going great.

Last week on Bluesky, the company wrapped up a string of own-goals on the trust and safety side by 1.) failing to run a username denylist that included technical terms but not even the most common list of racial and other slurs and straight-up Nazi signifiers, 2.) going silent except for GitHub comments and PRs after a Black user brought obviously awful usernames to light, and 3.) eventually publishing a super-anodyne statement that their policy has always disallowed racism and harassment and reflects their values. Katherine Alejandra Cross has a good overview of the situation on Wired:

When reached for comment, Bluesky’s press office mostly repeated the language of its posted apology on the platform itself. However, unlike in their thread, Bluesky admitted a mistake” occurred. The company added that an incident report to increase transparency and accountability” is in the making and will be published soon.

If this is true, it could represent a small step forward in restoring trust with a user base that, for now, is disproportionately made up of marginalized communities badly burned by the failures of moderation on large platforms. But skepticism is warranted. What Bluesky corporate has presented up to this point is communication that could have been generated by ChatGPT—the vague, anodyne language in its vanishingly rare public statements on the matter verge on the embarrassing.

My only quibble with this characterization is that nothing the company or its CEO has posted about this most recent incident qualifies as an apology. In the days since I initially drafted this post, a Bluesky protocol engineer posted an actual apology that I thought was a starting point. Otherwise, it’s been mostly silence in the face of broad pressure for the company’s leadership to acknowledge that the way they’ve handled this and previous moderation controversies has left a lot of Black (and non-Black) users feeling—I think rightfully—deeply pessimistic about the future of the platform and the protocol.

Maybe Bluesky reforms itself and hires a benevolent assassin to publicly lead trust and safety on their beta instance and handle those same dimensions in a culturally literate way in the protocol itself. Maybe not! But a lot of people want to use the beta, and if the team makes it to federation, far more people will be affected by the cultural patterns the protocol reinforces, enables, and restricts, so I’m rooting for them to get it together.

Ultimately, I think it’s going to be near-impossible to build out the right patterns for a prosocial big-world protocol without being scarily prescient about ethical hazards, deeply versed in how common abuse patterns, demonstrably determined to prevent every bit of preventable harm, and absurdly great at egoless comms and course correction. I’m not seeing that yet.

But here’s the thing: despite the Bluesky-the-company repeatedly tripping over its own feet, the beta actually is getting moderated, and on a reasonably snappy cadence—I’ve seen reported racist and transphobic posts and accounts get smacked down much more quickly than they were on, say, 2021 Twitter. And Bluesky-the-beta-platform is still where I’m finding a lot of the most on-point and sophisticated discussions of network norms/trust and safety/moderation work, along with all kinds of conversations between Black and brown journalists, public figures, writers, and artists.

An Afro-Boricua writer posting as Bryanna, Angry Noodle summarized the dynamic on Bluesky, quoted here with permission:

I post because I like attention. The positive kind, not the negative.

I think the moment we all admit we just want to be liked, life will be a lot better for all of us.

And it’s sad because social media platforms know this and capitalize off of it. They know very well that we’re desperate for an easy way to connect, and that for a lot of us Twitter fulfilled that role.

And they know that being semi-good means that they can still be mostly bad and won’t lose us.

This brings me back around to the Tech Policy Press podcast episode, to this exchange between Hendrix and Dr. Flowers:

Hendrix: I hear you almost mourning something that you do see as legitimate and worth preserving inside of this flawed environment, which is that, of course, these networks that have been built up sometimes in—well, I don’t want to use the word spite—but sometimes in opposition to the dynamics of that platform. Is that a fair assessment of your argument?

Flowers: Yeah, that’s a fair assessment of my argument, I would say that in spite of the platform” is actually a pretty good way to put it. When we start talking about Mastodon, one of the things that I’m going to try to make clear is how this in spite of” nature functions as a result of the ways that platforms inherit whiteness or inherit structures of oppression. But in spite of” is actually a fairly good way to put it, because it is in spite of the privately-owned nature of the platform, that it has become a digital commons.

I think those interlinked dynamics—that people want connection so badly that they’ll accept mostly bad” spaces, and that the history of the social internet includes oppositional and defiant uses of network affordances in spite of” their patterning in modes of capitalism and white supremacist assumptions and norms—belong at the center of our conversations about what better online spaces could be. I also think they’re signal flares that fediverse and other new-school network advocates should attend to, even when they come from or point toward networks those advocates find repellent.

What does it mean if the mostly bad” on Bluesky or until-very-recently-Twitter works for proportionately more Black people and other folks of color than the purportedly healthier systems at work on Mastodon?

If I had to sum up the whole entangled ball of predictable failures and unlikely successes across Mastodon and Bluesky at this exact moment, I’d say that neither place is covering itself in glory, but in terms of the provision of refuge, more familiar and cherished affordances with clear cues are attracting mass audiences better than high-friction tech structures designed to be beneficially anti-viral at the cost of being fun and easy to use.

Again, that’s all despite the ongoing moderation and communication problems, and even though pretty much everyone I follow on Bluesky is openly skeptical that it will work out.

I will also say that in spite of the problems I’m about to get into in exhaustive detail, Mastodon and the fediverse offer a big, motivated, often super-generous ecosystem of toolmakers, devs, and instance operators who really, really want to nail big-world + semi-sheltered networking that is inclusive and welcoming. I share that goal—it’s the only reason I’m spending so much of my time and energy wrestling with this stuff.

Back to affordances, then. I’ve said everything I’m about to write on Mastodon already, and lots of other people have too, but finding or re-finding things there is a nightmare, so I’m going to write it all down here as a reference. If you’re not into the details, do skip down.

Some of the expectation-afforance and cue-affordance mismatches on Mastodon are super-intentional, like the omission of built-in quote-posts and attendant notification systems, and the super-local and usually only semi-functional search. These omissions were meant to reduce virality and increase the healthiness of federated conversations, and Mastodon-the-organization has made a public commitment to working on careful ways of eventually integrating versions of those features.

Then there’s the other stuff, which starts before you get onto Mastodon at all. Signing up means picking an instance, and the instance you’re on will, through its moderation actions and policies and its federation and defederation choices, define much of your experience on the fediverse, but instances currently distinguish themselves primarily by location, size, or nominal shared interests, rather than by a detailed and transparent discussion of those actual major experience differentiators.

The missing stair actually matters, here: If you’re a Black gay dude, say, and you stumble into choosing an instance that federates with known trollfarms and harassment hubs, and you make a post that reveals anything about your identity, you’re very likely to get flooded with some of the worst abuse anyone ever gets on the internet. None of this will be clear to you before you sign up.

Assuming you escape that fate, then once you’re on, you’ll also have to individually find your people who’ve made it onto Mastodon—but probably not by searching by name or partial username, because search is flaky. If you google them, maybe you can find them, but clicking through the search result will bring you to their instance, which probably isn’t your instance, so you can’t follow them from there, because that’s not where you’re logged in. So you have to get their whole usernames—maybe from their Twitter bios if that’s allowed this month?—and search for them on your own instance, then follow.

Maybe you also want to follow people back, so look at your followers list and pop open some profiles. If you’re using the main web interface (or many apps), those profiles will open on, you guessed it, those people’s home instances. Try to follow them there and you’ll get hit with a login request, but you can’t log in there because there isn’t where your account lives. Unless you’re on the same instance, in which case it does.

You can follow straight from your timeline, though, so if you see a cool post, maybe you open the person’s profile in a different tab, decide if you want to follow back, and then re-find their post on your timeline so you can actually follow.

I’m not sure why you find this difficult,” say the people running around on marsh-stilts. Maybe you just love billionaires and chasing clout.”

Now let’s say you made it onto Mastodon and you find some of your people or you just make new friends or whatever (this is literally a strategy I’ve seen advocated for across Mastodon). Cool. This brings us to the next thing, which is that replies on Mastodon have strong affordance cues familiar to us from other networks, but the actual affordances are different:

  • If you make a post and people reply, you will see all the replies directed to you with any level of privacy (public, unlisted, followers-only, mentioned-people only—the quasi-DM of masto) and from any instance unless you’ve blocked or muted the replier. (Or if they’re sending you a quasi-DM and you don’t follow them and you have DMs turned off for people you don’t follow, in which case their reply will go into a black hole and neither of you will know it.)
  • If you are looking at replies to someone else’s post, you will only see the replies that come from instances that your instance already has a federation relationship with. Someone replying from a small instance that no one on your instance follows or is followed by? You’re not going to see their reply unless you actively open the post on its home instance and read the thread there.

This maybe sounds wonky and unimportant, but in practice, it means that tons of people who are replying are only seeing a fraction of the previous replies—which, in turn, means that the original poster often gets tons of near-identical responses/the same joke over and over/other annoying things, and that nearly everyone in the conversation is participating with a slightly different set of replies as their context. Add a couple language translation glitches and you have a recipe for deep social weirdness.

So if you’re that gay Black dude who accidentally landed on a server that federates with internet demons, your replies are eventually going to be filled with the worst things anyone can imagine sending to you, and none of the people on well-moderated instances will see the abuse you’re getting because their instances already yeeted those assholes into space. (Not even getting into the way followers-only and quasi-DM posts also reduce the visibility of abusive, threatening, creepy, and criminally tedious replies.) So when you then talk about the abuse you’re getting, most people will have no idea what you mean, because they don’t see it. Some of them will think you’re making it up.

Honestly this is so complicated I feel like I am making it up, but alas.

All these janky affordance/cue problems on Mastodon are a pretty crystalline example of what I’ve tried to get to in more abstract ways in earlier posts this year. People with specific goals and orientations made the system and laid down the cultural patterns that govern acceptable uses. Other people later joined in, and some of them made it through the gantlet of sign-up, got lucky on instance choice, were able to find friends, and brought in different hopes and purposes that are orthogonal to those of the initial makers. But unless/until those people and their communities hit critical mass, their attempts to hip-check the tooling and the norms will encounter so much resistance—some of it explicitly vile and organized, some not—that a lot of them drift away.

This is the affordance loop: Communities shape tools that shape communities, surrounded by everything happening in the world around us.

And another loop, online and off: Bad outcomes happen in spite of good intentions. When the fact of initial good intentions result in hyper-defensive responses to critique, the outcomes get worse and worse and worse. This is the second reason I try to force myself not to think too much about intention and motivation: When people get critiques laced with statements about their intent, it’s easy for them to respond with Nope, that’s not what’s in my heart” and throw out the whole thing. In my experience, critiques that relentlessly focus on explicit statements and specific outcomes are a little harder to dodge. (The first reason is that I am really bad at understanding other people’s inner lives.)

We used to call intentionally deceptive affordance cues in software dark patterns.” I think now they’re just a seamless part of the pervasive scam culture we accept online: Enter your email for 10% off! Lol now we have your email but you have to give us your phone number to get the code.

But let’s go back to the unintentional deceptions, like the Norman doors. Let’s say you build something that looks like it’s private, either because it’s described or visualized as locked” (like Mastodon’s followers-only posts or messages) or behind a login wall (like the closed Bluesky beta), but really it’s not private, or not very private, or not private for long and you totally said that in the docs.

People will still treat the thing like it’s private and post private information, including their bare butts, all over it. When this material is later made explicitly public, some of those people will be upset. Others will then berate those same humans for stupidly trusting the affordance cues; those people are wrong.

If you make something that looks and feels private, you’re making a commitment in the honorable and ancient language of affordances to behave accordingly. No amount of getting cute about not posting anything on the internet you wouldn’t want in a newspaper changes that. Slapping the equivalent of yellow plastic FLOORS SLIPPERY WHEN WET signs might get you through until you can fix it, but anything less means you’re building a trap, even if you didn’t mean to.

Now let’s say you build something boat-shaped, slap some high-viz paint on it, and drop it into a sea next to a sinking passenger cruiser. You might even get the chance to haul up some of the fleeing, floundering passengers, which seems nice!

Congratulations (genuine): You’ve made a lifeboat. This is true whether you call your boat LIFEBOAT or closed beta or jkjklol don’t hurt me. It was true from the moment you pulled someone out of the water, and it’s inextricably tangled with the duty of care.

A 1909 Los Angeles Herald Sunday Magazine article describes the stiltwalking peasants of the Landes (the ones in the header image for this post) as:

…a gay folk, who always seem to make the best of their hard lot, and manage to remain cheerful although they are generally miserably poor and frequently fever-stricken. In the more remote parts of the country the postman goes his rounds on stilts, and can thus do long distances in a short space of time. Mounted on his wooden legs, he is independent of brooks, marshes and hedges, and goes straight on, checked by nothing.

The 1908 tone is really something—if you open the page, beware the hair-raising bit about British roadbuilding in Africa—but there’s still something about simple, independent technologies that permit freedom and flourishing under difficult circumstances that probably resonates with a lot of people who want to build much better networks, preferably out of free and malleable pieces of software.

It seems relevant that the Landes region wasn’t marshy before humans cleared its forests around 600 AD—and it isn’t marshy now, because of a drainage and monocultural tree-planting program that kicked off in the early 1900s, in the process disrupting a thousand-year-old way of life adapted to the marshes’ challenges.

I was thinking about this earlier today, when I read Melissa Johnson’s funny, gross, sometimes wrenching account of a trek through the Guatemalan jungle for the wedding of her friends, an Arab-American woman and Mexican-American woman who met in the Army and then got out to live safely together. Why travel to a country where same-sex marriage is illegal and violence against LGBT people is a real threat—and then haul themselves to the top of a pyramid in the ruins of El Mirador—for their tiny ceremony?

Johnson writes that:

…neutral ground doesn’t exist for Angela and Suley. When they announced their engagement back in the States, members of their families cried—and not in the happy way. Despite getting their marriage license in California, the couple didn’t feel safe having a public wedding during the first year of the Trump administration. Choosing peak rainy season has assured them of precious privacy. We have not seen, nor will we see, another tourist the entire week. This is what a history of trauma yields. When you’ve been forbidden to be yourself for so long, a lost city feels like home.

The networks I want will feel like a refuge—including for people who need different affordances than the ones that keep me afloat.

Coming soon: Home and sanctuary. (How should a lifeboat be?) Why I’m talking about refugees at all.


20 July 2023