The Ambiguist (or, What Comes After Science Fiction?)

SF is the canary in technology's coal mine. But can we take the useful service it provides, and institutionalize it?

How do we reign in the accelerating upheaval of our lives driven by new technologies? Remember, technology is legislation. Introducing a new technology—social media, drones, or LLMs, is equivalent to imposing new global economic, legal, or societal policies with major consequences for families, companies, and you and me. The problem is we don’t know in advance what that policy is going to be. Social media was supposed to bring us closer. Drones were a cool way to photograph the top of your head. And AI… well, the less said the better. Instead, it’s like we spin a social-consequences roulette wheel every time we commercialize new tech. Could be a winner; could get you hauled away by the bouncers.

So what can you do about it?

In a recent post called “Alien Politics” I introduced the idea of abeyance. Here I’m going to expand on that by talking about what kind of person does abeyance, and what kind of institution might practice it.

Abeyance is the conscious refusal to decide that you understand something too soon. This is hard to do, because at the very least, we are expected to have an opinion on everything; for many of us, our jobs require us to be able to authoritatively state what we know, and deploy that knowledge. And as voters, we’re expected to comprehend current political issues and options. On a personal level, the human brain is hardwired to come to conclusions, as a survival strategy. The result is that for nearly everyone, the three most difficult words to say are, “I don’t know.” In “Alien Politics” I use the legend of the Bed of Procrustes to illustrate the dilemma: if a visitor was too short for his bed, Procrustes would stretch them to fit; if they were too tall, he’d cut them down to size. We constantly, automatically and unconsciously do this with new events, social systems, and technologies as they enter our lives.

It’s not just that it’s too soon to say how current crises, such as the U.S./Iranian war, will be resolved. For much of what’s happening, we don’t even have the right words to describe it. Premature closure is the Procrustean act of naming new situations with old words so that we can react—or at least, voice an opinion.

Often, premature closure manifests as premature labeling. For example, is social media really social? Seems antisocial to me. Maybe we should have waited a few years before settling on a term for it.

In “Alien Politics” I proposed an institution, the Department of Abeyance, whose function is to maintain productive contact with that which resists existing narratives, while not defining, categorizing, or dismissing it.

We need such a function in society to deal with rapid, unprecedented change. Arguably, we already have it, but it is not (directly) connected to the wheels of progress. Science fiction can say, “yes but—” to the introduction of driverless cars or personal robots, but we’re thought of as entertainers, so what we say is seldom taken seriously. Strategic foresight is a thing, and it does have access to power. It has strategies for abeyance, such as defamiliarization (also called strangemaking), but these are mostly upstream activities, inputs to a process that feeds decisions about what counts as real and actionable. Foresight feeds into decision-making but lacks the power to defer the decision about what counts. To make this concrete, I’ll use the example of the thing we have prematurely labeled as “AI.”

When it comes to how we could react as a society to new phenomena such as AI, there are two issues. Who performs abeyance, and, what would abeyance look like at the institutional, political, or legal level?

Let’s start with who does it.

The Ambiguist

First of all, it’s not AI.

“AI” is a perfect example of premature labeling. Large Language Models are not Artificial Intelligence—they don’t reason. My favourite test for this is to say to a circa 2026 LLM: “I want to wash my car. The car wash is only a hundreds steps away. Which makes more sense to get there: walking, or driving?” Guess which answer you’ll get. No, LLMs are not thinking machines. They are something else, something no science fiction writer anticipated, and our mislabeling of them regularly leads us to attribute qualities to them that they don’t possess.

The person who does the work of abeyance, the ambiguist, can’t be an AI because AIs are trained to minimize perplexity. In a literal, technical sense, that is their optimization target. They are machines for reducing surprise. The child of LLMs, fully reasoning AI, is certain to be similarly optimized. The mad prophet on the hill, on the other hand, is there to protect surprise—to maintain the perplexity of signals that haven’t yet been identified. The ambiguist’s role directly opposes AI’s function, both in the way it works and in its social effects.

By this framing, the ambiguist is not just a science fiction writer with a wacky idea. They stand in a directly adversarial relationship to Artificial Intelligence. They oppose our natural human cognitive tendencies toward premature closure.

In a Moebius-strip sort of way, our problem is not that we don’t have people who perform this role; it’s that, just like the issues the ambiguist deals with, we’ve had no name for them as a distinct role. Right here, and right now, we need that—hence my coinage.

The ambiguist has to exist as a specific kind of person, who draws a visible line in the sand declaring that we will not prematurely decide what some things are. Neither will we cede this closure of our worldview to AIs that someday will learn to reason but will be all the more dangerous because they’ll have been designed to always provide answers, never to shrug. The ambiguist legitimizes the shrug.

Exercise: Imagine the opposite of an LLM: a system deliberately designed to maximize perplexity. The intellectual opposite of an LLM. Something that flags where existing categorizations are under the most strain, without resolving that strain. That would be a genuinely new kind of tool, and a royal pain in the ass. (First impulse here: maybe LLM hallucination is a feature, and not a bug?) My suspicion is that they would act like Zen masters, providing us koans to break our cognitive chains.

So who becomes an ambiguist? We have several obvious candidates: academics, whose job is to worry away at ideas like dogs with a chew-toy; artists in general and science fiction practitioners in particular; diplomats and translators; poets, who bend and fold language to express thoughts and feelings that are otherwise inaccessible; and indigenous knowledge-keepers. Each, in their own way, takes apart our preconceptions and provides ambiguous alternatives. The common problem they share is that abeyance requires a particular psychological and institutional profile that none completely satisfy. Our candidate ambiguist needs to be:

• Genuinely comfortable with unresolved incomprehension over long timescales—not as a temporary state but as a permanent one.

• Highly precise and analytical, able to distinguish the merely unsettled from the truly ambiguous.

• A genuine generalist, interested in everything and knowledgeable enough to be comfortable with trends analysis, modern physics, sociology and anthropology.

• Resistant to the social reward of providing answers, which is enormously strong in almost every knowledge profession (and every religion).

• Capable of articulating why something resists categorization without accidentally categorizing it in the process.

• Institutionally protected from the pressure to deliver actionable conclusions.

In the case of the last qualifier, SF writers have it down. Nobody’s paying us to reach conclusions. (Generally speaking nobody’s paying us at all, which helps.) But arguably, it’s not science fiction writers either. Remember, we utterly failed to anticipate Large Language Models; arguably, it’s our fault that the technology has been so easily labeled as Artificial Intelligence, because it so neatly fits into a set of cliche definitions that largely, we generated.

What else has SF failed to anticipate? What else are we unable to imagine because narrative closure demands that we fit every new thing into some kind of story? This question matters, because we are informally but genuinely taken seriously as the mad prophets of technology. Yet maybe, in today’s world, we’re no longer up to the job.

The ambiguist is not an artist; but an ambiguist is like an artist. In free societies, the arts fall into a special category that’s exempt from government pressure. We are what Shakespeare called “an allowed fool.” The ambiguist needs to be an allowed fool: in a recognized institutional position that grants license to declare what is productively unresolved, without that being treated as failure.

The ambiguist, in short, presents the world as an unfinished symphony, drawing a clear line around the missing movement.

The Department

Abeyance at the institutional level would be a persistent procedural constraint on downstream decision-making. Not a conservative attachment to existing systems and power structures for their own sake, but the recognition that we are not yet in a position to decide what particular things are. In technology policy, abeyance manifests as limitations on how universal technological commercialization is allowed to be. Currently, we impose such restrictions only after the fact, when we discover that a technology is not what we initially labeled it as. The growing international movement to ban social media accounts for kids under 16 years of age is your example here. Wouldn’t it have been better to start by restricting it to adults, to see how it actually affects people, and study its impact on the adolescent mind before putting it in the hands of every confused teenager on the planet? We didn’t do that because the label—’social media’—tricked us as effectively as the companies pushing these products. We decided too soon that we knew what we were dealing with, and we’ve paid the price.

In governance, any legislation or major policy touching a designated ‘registered but unidentified’ signal should require a formal abeyance review—essentially an audit asking ‘are we Procrustizing this?’ The strangemaking function here isn’t constructive (producing new scenarios) but defensive, slowing the capture of the genuinely new by existing categories.

Strategic Foresight and Abeyance

Defamiliarization techniques in foresight research have identifiable sub-types, and each maps onto a function the Department of Abeyance would need:

• Ostranenie (Shklovsky’s making-strange): perceiving the familiar as foreign. Applied institutionally, this becomes a practice of deliberate re-alienation of things that have been domesticated too quickly—revisiting “settled” interpretations of anomalous signals.

• Scenario inversion: asking what the world would have to be like for the strange signal to be entirely obvious and expected. This is a way of holding abeyance while still doing analytical work.

• Causal layered analysis (CLA): Inayatullah’s method explicitly holds multiple mutually inconsistent “myth and metaphor” layers alongside the surface data without resolving them. The CLA level that maps onto abeyance is the myth/metaphor layer — the level at which narratives are held in suspension before one wins out.

The hardest practical problem is this: democratic institutions exist to decide, and they need to be accountable for their decisions. Abeyance creates a new category of political object—the ‘officially incomprehensible for now’—that could be misused as a way to avoid accountability. A government practicing abeyance would need to make binding decisions anyway, but there's a follow-on design problem: what disciplines the declaration of abeyance itself?

One possibility, which connects defamiliarization directly to institutional design is that an abeyance declaration requires showing your work—documenting specifically what is strange and why it resists existing categories. This is analogous to how judges must write reasoned opinions even when ruling for deferral. The act of articulating the strangeness in language creates a kind of inverse Procrustean test: if you can describe exactly why it doesn't fit, you've done something beyond dismissal. This documentation then becomes the input for future sense-making once the categories do develop. And this, in turn, highlights the importance of words in both premature closure, and abeyance.

If Facebook is not really “social media,” then what do we call it? If ChatGPT and Claude aren’t AI then what, exactly, are they?

One of the jobs of the ambiguist is to coin the right word for a new thing; this creates appropriate closure. But the ambiguist may also create new words that point at a blank area of our map without saying “here be dragons,” or indeed saying anything. I did this quite deliberately in my first novel, Ventus, with a new word that’s vitally important to the resolution of that story: thalience. Thalience isn’t the answer to a question, it is the question. (In this case, about “the Pinocchio problem:” when does AI cease to be a kind of ventriloquist’s puppet—merely spouting our thoughts and worldview back at us—and start having its own ideas about the world? If sentience is thought as we understand it, thalience is thought exactly and precisely not as we understand it.)

Invented terms are abeyance tools and can cut both ways, to closure or deferral. Asking whether LLMs are thalient doesn’t resolve the question of what LLMs are; but it does reframe the debate more productively. As another example, in my 2019 novel Stealing Worlds, I created the term deodand to refer to AIs that think they are some natural system, like a river or forest, and act in “their own” best interests. The term deodand raises a big question mark over the priorities AI will have. We imagine they’d either identify with us, or distinguish themselves as a separate people. The common assumption—that they will identify as people—might be wrong.

Thalience points out that we don’t yet know what kind of relationship we have to reasoning machines, but we should not assume it’s derivative of our own reason. Deodands suggest that we don’t yet know what moral and legal category applies to nonhuman agency (including both AI and ecosystems), and we should not assume it’s mere property. Both terms are engineered incomprehension of exactly the kind a Department of Abeyance would be designed to formalize.

The Department therefore treats neologism as institutional technology. The coinage of genuinely new terms—ones that resist immediate domestication into existing frameworks—might be one of the primary tools available to an abeyance practice. Not jargon, but terminology that carries the strangeness in its form to resist the lazy synonym. The fact that “thalience” sounds unfamiliar is a feature, not a bug. It prevents the reader from substituting a familiar concept and moving on.

There is danger in becoming complacent with such neologisms. Terms coined to hold conceptual space open tend to get captured anyway. “Cyberspace,” for example, started as estrangement and ended as branding. How do we maintain the abeyant quality of a term over time? Maybe the Department of Abeyance would need to be in the business of retiring terms as well as coining them: formally declaring when a word has been captured and can no longer perform its function.

The Department’s First Motion

Imagine that your government has just minted its Department of Abeyance, and the bureaucracy has cranked out its first recommendation. What would that recommendation be?

My vote is that they would say, “We vote to retire the term ‘social media’ because it does not accurately describe the technology that Facebook, X, TikTok and other companies are promoting. We are opening a public discussion on what other term is more appropriate, or whether we need an entirely new word that avoids mistaking the technology’s function as inherently pro-social.”

Obviously “AI” is another term that might draw the ire of the Department. But what new words do you think it might invent? Where are the blank spots on the map that should be pointed to? Have you got any suggestions?

Let me know. And meanwhile, ponder the figure of the ambiguist, strolling away in some rain-soaked, black and white Noir shot. How do they tilt their hat? Do they slouch like a detective, or glide like a poet? Who is this person? Are they a disillusioned science fiction writer? A former foresight practitioner?

Is it you?

—K

---