Literacy-slop

I found I had so much to say about Emily Segal's latest Nemesis Memos post, that I not only transferred it from Thought Shrapnel to here, but I think this is going to have to be a couple of posts.

Segal is talking about the world of aesthetics as it relates to AI, but it's absolutely relevant to the world of digital literacies.

Taste is not really a property of various objects. It is a socially validated relation between objects, people, histories, scenes, and timing.

If we swap “Digital literacy” for “Taste” then it's a socially-negotiated relation between people, tools, practices, contexts, and communities.

Segal discusses the ways in which AI-generated designs look almost right, but feel a bit hollow. The uncanny valley is produced by AI knowing what the visible signs of taste should be (e.g. a Dieter Rams book on a shelf) but they're extracted from the social relations that give them meaning, and redeployed generically.

I would say the biggest thing people are missing about taste across the board is that it is relative, contextual, and social. Taste needs to be socially validated. There is no such thing as taste if it falls in a forest.

While the object looks tasteful, and the curation looks intentional, it's contextless. It has no meaning.

This isn't a new problem. Proponents of New Literacy Studies (e.g. Gee, Street, Lankshear and Knobel) made a similar argument about reading and writing. Literacy isn't a “skill” you carry around in your head, but a social practice which only makes sense in context. My doctoral thesis, and work I've done around digital, web, and AI literacies after it, was an attempt to think about that in the digital age.

The eight elements (cultural, cognitive, communicative, constructive, civic, critical, creative, confident) that resulted from my meta-analysis were never meant to be a checklist or a framework. Instead, they describe the baselayer of a kind of habitus of norms, dispositions and tacit know-how obtained by someone who is literate in a particular digital domain.

Tasteslop, you could argue, is what happens when the habitus is automated to produce signs without substance.

Automating the classifying function

Segal explains that:

Tasteslop is what happens when the classifying function is automated, overly explicit, or reduced to spitting out rote taste tokens.

I've made a similar point about digital literacies when talking about how the point is not simply to operate tools more efficiently, but to learn to notice structures, question defaults, and understand the incentives underlying those tools.

Segal's collaborator Greg Fong puts it more succinctly:

Nobody can have taste unless somebody else can see it. The LLM is not a person. It isn’t subjective either. You can subjectively like it, and it can be weird enough to be stylish, but it doesn’t actually have taste; it can only look for data indexes of taste...

That phrase, the “data indexes” of taste, has a parallel in education. It annoys me that academics wring their hands about students submitting AI-written essays, as if their institutions haven't consciously created data indexes of literacy and competent performance. Essays were long ago extracted from the social practices that gave learning meaning.

So I'm going to give this a name: literacy-slop.

Literacy-slop is the credential without the community of practice; it's the qualification without the learning; the skills certificate for getting an AI agent to click through a self-paced module on digital skills. It looks like literacy, satisfying the classifier. But it's just curation without a social body.

Literacy-slop fails in the same way tasteslop fails

Let's take three of the eight elements of digital literacies to see what's going on.

Critical element

Segal cites her “Bourdieu agent” (presumably an AI tool) as saying:

If taste classifies (and classifies the classifier), tasteslop is what happens when the classifying function is automated.

The critical element asks people to interrogate their, and other people's assumptions. Who is doing the classifying? With what values? In whose interest? And to what end?

Tasteslop is a kind of fluency without interrogation, where you can deploy Dieter Rams as a taste-marker without knowing why Rams' ten principles became canonical and not someone else's. Equally, you can generate a “digitally literate” essay response without understanding anything behind it.

Critical digital literacies aren't about being suspicious of everything, but about having the tools to ask relevant questions when they matter. Literacy-slop never asks them.

Cultural element

The three things I always make sure to say about my work on literacy are:

1. Literacies are plural
2. They are context-dependent
3. They are socially-negotiated

This almost exactly maps onto what Segal says about taste:

The biggest thing people are missing about taste across the board is that it is relative, contextual, and social.

The cultural element of digital literacies foregrounds the fact that digital practices only make sense in context. Competence, skill, or fluency in one setting might be relevant, or actively counterproductive in another.

This element is about understanding the norms, histories, and values of the particular digital environment in which you find yourself. The way that people interact on Instagram is different to LinkedIn, is different to the student forum on Moodle.

Tasteslop decontextualises. Segal talks about how the same Trader Joe's tote bag is vulgar in California and briefly high-status in Tokyo. A Togo couch is tasteful until overexposure makes it obvious. It's all surface, no depth.

Literacy-slop does the same. It looks at surface competencies such as “the ability to evaluate online sources” or “understanding privacy settings” without the contextual knowledge that makes those competencies functional or useful. Evaluating online sources means something very different in a professional scientific community compared with a secondary school classroom, or a political campaign.

Without taking care of the cultural element of digital literacies, you are just running the classifier and creating literacy-slop.

Communicative element

Meaning is socially-negotiated. Taste requires an audience and a community sharing enough reference points to validate the judgement.

Nobody can have taste unless somebody else can see it.

AI can't participate in that process; it can only simulate its outputs.

The communicative element that I identified in digital literacies isn't just about about writing clearly or being able to share things. Rather, it's about the relational dimension of meaning-making: the fact that I'm writing this on my blog makes my thoughts visible and valuable to others who can recognise and respond to it.

That's very different to an AI-written essay that's automatically graded. It's not really literacy at all – never discussed within a community of practice, never contested, or refined, or validated by other humans. It's just a surface-level data index.

Cultural capital after extraction

I like Segal's summary of tasteslop as:

Cultural capital after extraction, after it's been through the blender.

In other words, platforms and AI tools identify the aesthetic outputs of culturally-situated tastemaking and decontextualise them from the relations that give them meaning. Like the front of a house on a film set, the surface is there, but there's no actual substance behind it.

I'm not against AI in education, but I'm bored of seeing the same things presented as somehow revolutionary:

• Learning content generated by models trained on Open Educational Resources and textbooks
• Automated grading based on surface-level markers of competence
• A focus on decontextualised “skills” with the thinnest layer of evidence

This is literacy-slop served at scale, shorn of community interaction and the norms of the various disciplines. It's the logical endpoint of an education system that treats competence as something that can be somehow identified, extracted, certified, and redistributed independently of the communities of practice that produced it.

AI just accelerates that tendency and makes it cheaper. But it's the same error: mistaking the index for the practice.

My eight elements of digital literacies were an attempt to push back against that tendency. As I mentioned above, they describe not a set of skills but a habitus. There exists a whole complex of knowledge, dispositions, and social relationships that makes someone capable in various digital contexts. This is why I've always been interested in the Open Recognition side of digital credentialing, rather than just slapping a badge on something.

What are you saying, Doug?

You could dismiss the above as a misguided claim that you can't measure learning. That's not what I'm saying.

Instead, I'm making the narrower claim that tasteslop and literacy-slop can be seen as symptoms of the same extractive logic. And that recognising that logic is itself an aspect of critical digital literacies.

• A question for tastesloppers: “Oh, you've created an impeccably-tasteful AI-generated moodboard? Great: tell me what you know about Dieter Rams.”
• A question for literacy-sloppers: “Oh, you've created a beautifully-produced AI-generated learning module? Great: tell me what you know about constructivism.”

AI is a tool that allows us to create shortcuts, but the point is to be able to judge whether the shortcuts take us anywhere useful. Just as humans are the ultimate arbiters of taste, so we are the ultimate arbiters of literacy. And that can't be sloptimised.