The Single Arc

I'm partway through Annie Jacobsen's new book, Biological War: A Scenario, her second, after Nuclear War: A Scenario in 2024 (which I reviewed here). Somewhere in between I read AI 2027, the online scenario project a handful of AI researchers built around what happens once a lab makes something smarter than its makers, and Europe 2031, the policy narrative a group of researchers wrote in June about Europe losing the AI race. Three projects in three years, all calling themselves a scenario. What's going on?

"Scenario" is maybe one of the most recognized terms to come out of the field of foresight (or what many people lazily call "futurism"), and the least understood. Familiarity gets substituted for understanding, and that substitution grants permission: treat a scenario set as a menu, pick the one that matches your priors or your fear, instead of using it to surface what's vulnerable. I see this constantly, and not just in books and advocacy projects — in meeting rooms and government ministries, wherever a scenario set gets commissioned.

So it's worth knowing where the word came from, because the origin explains what it's doing on these projects. As I've written about elsewhere, Herman Kahn appropriated it from the screenwriter Leo Rosten at RAND in the late 1940s, converting weapons counts and casualty rates into a story a room full of generals could feel. "Screenplay" was too Hollywood, Rosten told him; "scenario" better fit these narrative set-pieces. The complexity — the driving forces and assumptions, the uncertainties, the plurality — was foresight's addition on top, giving the appropriation some body and difference. And for decades the term and tool were used in places where big futures were analyzed and decided: RAND, the Pentagon, Shell, strategic planning groups in corporations and governments. Check LinkedIn and you'll see scenarios roll off the production lines of big bodies like the World Economic Forum or OECD all the time.

Properly understood, a scenario is an input, not an end product. Using various methods, you build several futures, usually divergent, and test a plan or strategy against them. This forces an accounting of what you don't know and how much it matters. You don't get to pick a winner like a selection from a wine list. You build a decision sturdy enough to survive whichever future shows up.

So, we've established that most scenarios come in sets: plural. Why this one-arc story form, and why now? Because the uncertainty — in AI capability, in biosecurity, in what happens to Europe's tech dependence — is genuinely high right now, and high stakes create demand for something that looks like it's grappling with that seriously. What gets rewarded in a moment like this is the thing that tells an audience what to do — brace, invest, regulate, panic, relax, pat yourself on the back. A single dramatized arc, wrapped in the credibility of a more complex discipline, delivers exactly that: resolution. A genuine scenario set doesn't resolve the discomfort of not knowing. It stays with it, on purpose, to keep you honest, which is often more honest and less useful to anyone trying to move people fast. A resolution, on the other hand, is what people expect in their Netflix dramas, beach reads, and studies of the future alike. The only thing better is a sequel, or cinematic universe.

Jacobsen's approach in her books collapses complexity the hardest. Her new book, like her last, is one arc, one ending, and reads more like watching the operation of a mechanism that, once started, doesn't stop until it reaches an (unhappy) end. Her work here is structured from journalistic shoe-leather and uses institutional grammar to indicate it is a product of research and plausibility rather than fictional fancy. Her own framing: the book "is based on facts and projections sourced from exclusive interviews with presidential advisors, Defense Department officials, medical doctors, microbiologists, epidemiologists, diplomats, congresspeople, C.I.A. officers ... and first responders." That's the legitimacy claim. "Scenario" does the same work "based on a true story" does in a trailer — tells you not to file this under invention.

AI 2027 splits the binary at least once: the authors wrote two endings, a "slowdown" and a "race," rather than claim they knew which was coming. Two branches off one axis: how fast people get scared and hit the brakes. And the credentialing does more legitimacy work than the branching: the project describes itself as informed by "more than a dozen tabletop exercises with hundreds of different people," built on "over 30 iterations" of the exercise, sentences built to generate trust. This is "expert" work, so the priors of those experts are largely ushered through.

Europe 2031 admits the game up front. Its authors call it plainly "an attempt to explain what Silicon Valley is seeing and Europe is not," written to shake European decisionmakers into action rather than forecast anything. Like Jacobsen's books, it's a scary story whose ending you want to avoid if you're in that class. The policy analyst Max von Thun's critique of the piece gets at a weakness inherent in this one-arc approach: the narrative's "emotional tone primes readers to accept at face value the authors' premises, many of which are highly debatable." The selection logics of plural scenario models determine some of the arcs in that approach. Here, it's down to the storyteller and their aims.

I'm not the only one who noticed. Gideon Lewis-Kraus's New Yorker review of Jacobsen's new book does exactly what it's diagnosing everyone else for: it files her disaster nonfiction, AI 2027's branching forecast, and Europe 2031's policy narrative as the same thing, because they share an ending, not a method. He circles the real problem anyway, faulting these works for feeling "argued backward from foreordained conclusions," and saying they might land harder if they were "more inventively open-ended." His fix is imagination, a "collective failure of the imagination," cured by having the nerve to... look "silly"? But open-endedness isn't an outcome you reach by trying harder to be more wild. It's an informed selection of methodology shaped by an intended use, not a mood.

None of this is really about any author cutting corners or mis-packaging their work, though. It's what happens whenever the objective is a response instead of a critical discussion, whether that response is action, piqued curiosity, fear or just a feeling of entertainment. A single arc gets a reaction faster than a technocratic array does. And a response is the only measurable outcome. A scenario can't be wrong: if that particular future doesn't arrive, no harm, no foul. If it does, genius!

Either way, the lay reader walks away thinking that this is what a scenario is, that the work is just telling a scary or a cool story well enough, sourced well enough. We've seen this happen around another conflation: prediction and forecast, a bet on one outcome versus a statement of odds, now used interchangeably until the difference stopped mattering. What a scenario is good for is harder and less satisfying than storytelling: sitting inside real uncertainty long enough to prepare for more than one future, instead of bracing for whichever one someone else needed you to feel. So expect more of these, better sourced and more cinematic each time. The conditions aren't going anywhere. When the next one lands, the question to ask isn't whether it's plausible. It's what it wanted from you, and what you'll do in response.