The Missing Staircase

How catastrophists skip the hard thinking and call it analysis

There’s a rhetorical move, a powerful yet pernicious sleight of hand, that happens when a certain kind of person wants to address a topic they see as an “existential risk.” A catastrophic scenario will be introduced, the conceivability will be established, but probability of it will be quietly handwaved or just assumed. No real care taken to account for the mechanisms required, the actual ways by which this thing would come to pass. Just a leap straight to the implications of the event itself, because really, that’s all this person wants to discuss.

The premise—that the thing is not merely conceivable but genuinely likely—is treated as established by the act of having imagined it vividly enough. Which is the first mistake. There’s even a Latin term for the fallacy: “possibiliter ergo probabiliter”—possibly, therefore probably. A pattern that’s old enough to have its own Ancient Mediterranean descriptor, and yet still brazen enough to dominate contemporary discourse.

Notice the infrastructure that’s now in place to peddle these odd possibilities (that the peddlers tacitly treat as high probabilities): the catastrophist now has a newsletter, a podcast, a verified badge on social media, a book deal, a “research center” that might just be them and some friends spit-balling. The vague gesture at civilizational collapse has been laundered into prestige. These people go from thinking out-loud about random prospects to soothsayers and “futurists” with their eye on the ball.

I’m here today to challenge the assumptions and remind us of all the necessary steps that were skipped. I call this the Missing Staircase, and its absence is the oft-ignored subtext, the dismissively detached rigor, that underlays this whole enterprise.

---

Lobby to Rooftop

Let’s introduce the building metaphor. We begin with our lobby, which is logical possibility itself. The rooftop is the probability assignment and next steps. What we must contend with is that the catastrophist takes an elevator—maybe even a teleporter—that simply doesn’t exist. They then appear on the roof, all too excited to describe the view. Because it’s only the view that concerns them, not how we get there.

First Steps

Let’s take the stairs. In fact, follow me as we work through all six flights explicitly. Step one is historical & structural precedent. These are the analogies to whatever event is being bandied about and prophesized. Crucially, these are must be load-bearing, entry level pieces to the stairs. They are not decorative, they are no optional. If someone tells you that it’s hot and sunny in California, they can cite precedents as well as present day. If they told you a drought or wildfire were a serious threat in specific areas, they could point to the steps. They’re on steady footing.

But what if they said Los Angeles is due for a snowstorm? Well, the fact that it literally doesn’t snow in most of Southern California a running truism at this point; so much so that there are event songs about it, and people from the coldest regions of the country vacation there in the winter. Is snow impossible? No. It’s happened before, just very rarely and fairly lightly. Yet a blizzard seems well beyond the pale.

Flights Two & Three

Now, what about a California earthquake? Well, the precedents there are much better. But even so, we need more than precedents before we can start talking about probability as if it’s assumed and move on to next steps like buying insurance or leaving the state. That’s where we go to the second step, or the second flight, which is mechanism specificity. This is an actual causal pathway; if you can’t sketch it, you have an atmosphere sans a real argument. What makes earthquakes happen? What makes them more likely in California but less likely in Florida?

Let’s check the precondition inventory, which is step/flight three. How many mechanism conditions are currently satisfied, partially satisfied, or entirely absent? This third flight is where most scenarios wither and die, which is precisely why catastrophists hate engaging with it. We know the preconditions for drought, heatwave, and wildfires are met. We know the ones for heavy snow are not. We would need to investigate about earthquakes, and that very investigation is tedious, even boring. It’d be much more fun to just imagine a big quake, like in the classic disaster miniseries 10.5, or later fictional or hypothetical depictions in media.

Flight Four

But if we never falsify our hypotheses, we’re not really doing analysis, we’re just playing pretend. And as we’ve already pointed out, the catastrophizers don’t want to be Hollywood screenwriters, they want to be serious, respected “researchers” and pundits. Given that, they need to work through the steps. Like flight four, friction accounting. Assess the countervailing forces need to be weighed, measured, and found satisfied or wanting. It’s similar to three, but more about looking at what’s directly obstructing the way rather than what’s already there paving said way.

Back to our beloved Cali, we see that high heat, dry nature, and local fault lines all serve as backdrops to calamity. But we also know that these same conditions forestall competing issues on the cold front, and may even be met with resistance from institutions that exist specifically to prevent the worst disasters from either happening in the first place or getting out of hand. The building codes, construction materials and designs, cooling systems, nature services, fire departments, and other emergency services exist to mitigate all these risks. They don’t work equally well in all instances, but they’re often fit to purpose for the majority. Heatwaves in the United States—including CA—are bad, but manageable. Most earthquakes cost limited damage and are prepared for heavily. Fires are another story, but the overall point holds.

Flight Five

Then we get to flight five, base rate anchoring and the outside view. This is a fancy way of saying we need to stepping outside the potentially biased or preferential view that a juicy narrative provides. Daniel Kahneman, the award-winning psychologist best known for his work on the psychology of judgment and decision-making as well as behavioral economics, urges us to consult the outside view by asking how often situations with this profile have actually produced this outcome. The inside view—which is what catastrophists are always doing—immerses you in the particulars of your case until it feels unique and inevitable. We must resist the trap.

The outside view refuses that immersion and demands a comparison class. The reason this flight is so often skipped is that honest base rate anchoring is almost always immediately devastating to the catastrophist argument. The reference class has to match the precondition profile you established in flight three. So you're asking: how often has California had, say, a major earthquake of 6.4 or higher? What happened when it did? And so on. The answers often make us realize that the sexy scenario, which makes a good screenplay, simply isn’t what we should treat as the probable one. This is why the reference class so rarely gets specified by the catastrophist.

Flight Six

Probability assignment is flight six. This is basically the rooftop exit door, the place many catastrophists want to jump to straight from the lobby. We engage with something falsifiable; something that could, in principle, be wrong. Then we assign a number to it. As we’ll tackle in the sections below, one rather egregious example of this phenomenon, in terms of skipping steps, is AI doom and the talking point of “P(doom) values” or scores. Many people in the AI industry, as well as engaged pundits or onlookers, like to talk about their P-Doom number as if it were a blood type or a zodiac sign—which is to say, an amusing dinner party topic.

Yet these same folks would often blanche, and call incredulous respondants “naive” or “dismissive” if they don’t play along, if they even simply ask what real and verifiable steps went into determining these numbers. That’s the tell. Steps two through five are where the intellectual labor lives—and they’re skipped because completing them almost always returns the same boring answer: “this scenario is unlikely and/or won’t result in what the catastrophists assume it would.” The exercise they want to do is talking about doomsday in the abstract, untethered to reality. But the exercise they really need to do is check the history, the facts, the safeguards, and then be honest.

---

Prior Witnesses

Security Theater and Movie Plot Threats

Bruce Schneier is a cryptographer and security expert. His credibility comes from a career spent doing actual security work. His book Beyond Fear, presented a five-step process for evaluating the value of security countermeasures, and is notable for coining the term "security theater." He coined “worst-case thinking” in a 2010 op-ed, where he described it as an extension of the precautionary principle that involves imagining the worst possible outcome and then acting as if it were a certainty. It substitutes imagination for thinking, speculation for risk analysis, and fear for reason.

He identifies it as bad decision-making on multiple grounds. First, it’s only half of the cost-benefit equation, since every decision has costs and benefits, risks and rewards. By speculating about what can possibly go wrong and then acting as if that is likely to happen, worst-case thinking focuses only on the extreme but improbable risks and does a poor job at assessing outcomes, not to mention properly weighing tradeoffs.

The second charge is harder: it’s based on flawed logic—it begs the question by assuming a proponent’s case. Worst-case thinking is a way of looking at the world that exaggerates the rare and gives the unusual much more credence than it deserves. It isn’t really a principle; it’s a cheap trick to justify what you already believe. It lets lazy or biased people make what seem to be cogent arguments without understanding the whole issue. That last sentence is worth sitting with. He’s not saying worst-case thinkers are confused. He’s saying it’s a trick—a rhetorical mechanism for laundering predetermined conclusions through a veneer of risk-consciousness.

---

Schneier coined “movie-plot threat” to describe how security theater gets promoted and put into place. By pitting the proposed practices against movie-plot scenarios that have the dramatic architecture of a thriller premise, bad ideas get treated as policy-relevant precisely because of that architecture, not because of any probability assessment. Schneier even ran an annual contest inviting readers to submit the most elaborate, implausible-yet-technically-possible terrorist scenarios they could construct. The point of the contest was satirical: the joke is that anyone can do this.

Generating a vivid, technically non-impossible catastrophe scenario requires no analytical skill whatsoever. It requires only imagination and a willingness to dress that imagination up as analysis. This is the direct ancestor of the AI doom argument and others in this genre. Both are movie-plot threats, like Skynet in the Terminator series. They have the right dramatic shape—factions, stakes, a recognizable villain structure—and that shape is doing the work to cover for empty epistemics or missing mechanisms that their proponents would rather you didn’t examine.

Superforecasters

The political psychologist Philip Tetlock teaches us about what he calls “superforecasters.” His 2005 book Expert Political Judgment was the first major indictment. In a study he ran over twenty years, Tetlock asked a group of pundits to rate three possible outcomes for a political or economic event on a scale of zero to ten for likelihood. He collected about 28,000 predictions from 284 experts. After careful analysis, he found the experts' forecasts were only slightly better than luck—and they did worse than simple algorithms.

Crucially, Tetlock proved that the more famous the forecaster, the worse their predictions. I have spoken about something similar with my piece, “Naked Emperor Intellectuals,” specifically what I call intellectual deference cascades. Once someone is famous enough, their say-so is parroted by others of similar or lower status as if it was stress tested and accurate, when it was actually just passed through as received wisdom. What’s most concerning about this expert mode is that it functions like ideological kabuki theater. It’s structured to be memorable and undisprovable rather than accurate and accountable.

Tetlock worked with Dan Gardner to co-author a spiritual sequel book called Superforecasting in 2015, which is the constructive, good news follow-up. They found, through the Good Judgment Project, that tens of thousands of ordinary people could set out to forecast global events. And that, in contrast to putative experts, some anonymous “superforecasters” turned out to be astonishingly good. They beat other benchmarks, competitors, and prediction markets. They even beat the collective judgment of intelligence analysts with access to classified information, roughly 30% of the time. This is for a few key reasons, all of them part of the staircase. Superforecasters anchor to base rates, think in explicit probabilities, update on evidence, keep score; the catastrophist does none of these.

---

Two Flights of Stairs, Left Unclimbed

Now we can address two commonly proffered scenarios where steps were skipped. These are the very reasons, and topics, I drafted this essay with in mind. Each gets thrown about like it’s not only possible but probably, on exactly the shaky basis we established as a pattern above. The catastrophists treat their own sense of excitement or urgency around these hypothetical, alongside their vivid imaginations in presenting them, as the (unearned) grounding for discussing them in the first place. Like they’ve operationalized the infamous Dick Cheney doctrine that if a worst case scenario has even a 1% possibility of being true, we must treat it as an absolute certainty in terms of how we handle countermeasures and public policy.

A Second American Civil War

Let’s run the staircase explicitly…

• Logical possibility: cleared

• Historical & structural precedent: the first Civil War’s actual preconditions were geographic concentration of competing factions, a functioning alternative government within one region, some defectable military units, and material property regimes (the proposed expansion of chattel slavery) under threat. Now we compare these with today, and virtually none have meaningful contemporary analogs. Already, we should rightly stop on the first flight of stairs.

• Mechanism: pathway from “very angry electorate” to “two armies fighting for territory over years” produces silence or a cascade of hypotheticals each requiring its own staircase. The non-existent mechanism problem is so significant that most fictional depictions of a 2nd civil war either leave it blank (as in Alex Garland’s eponymous 2024 film), or they propose ones that beggar belief.

• Precondition inventory: political alignment is not geographic; there is no defectable or competing standing army; no material property regime being defended; anger does not organized, let alone armed, capacity for war.

I’ll stop there, except to add on, from the base rate, that rich and integrated democracies do not do this. If we actually checked the reference class, it would not be encouraging for the catastrophist. A more likely scenario is impeaching a sitting president. Or simply engaging in a high profile, but ultimately ephemeral, public riot in the streets. That the underlying political concerns might be major or legitimate is not enough. The specific catastrophic framing fails nearly every flight above the first.

AI Doomsday

Here we’ll run the full staircase, cuz this is where the framework demonstration does its heaviest lifting…

• Logical possibility: cleared.

• Historical & structural precedent: absolutely none; no reference class whatsoever; the catastrophist is calling for evacuation without even cleanly clearing a few flights of stairs. The lobby is where this lies, not above.

• Mechanism: walk each required step. AGI achieved? This has not happened, the timeline contested, and the goalosts keep moving. Misaligned objectives between AGI/ASI and humanity? Non-existent, both because the tool hasn’t gained this level of autonomy and because general malevolence seems to run counter to current AI industry goals; just saying “it could want this someday” is not the same as explaining how or why. Misalignment concealed? This requires strategic deception that itself presupposes things that haven’t been demonstrated, including the aforementioned malevolence and sentience; the most people point to are AI malfunctioning in ways that can be overly read into like ink blot tests. Instrumental capacity at civilizational scale? Never developed. Countermeasures overcome? Nope. Each step is a separate question requiring separate evidence, none have been satisfied.

• Precondition inventory: step one—AGI itself—is unsatisfied; a scenario whose first mechanistic requirement is unmet is a speculation about what follows from a speculation. Call it spec on spec, for short. The staircase doesn’t work that way. We must not proceed to upstairs until we have the requisite grounding.

• Friction: competing institutional pressures; developing regulatory frameworks; active alignment research; physical and logistical constraints. All these need to be weighed, not waved aside because the endpoint is dramatic enough to make friction seem petty. At present, most AI labs are struggling to keep their coffers full and develop a cohesive business model, to say nothing of the productive partnerships necessary to do the things that the sexy scenario requires. There’s friction all over the place, in short. Far more of that than the opposite.

• Base rate & outside view: zero; no reference class, no stability under the outside view. It’s all inside view, all the time. The calls are coming strictly from inside the house, and they’re functionally crank calls.

• Probability assignment: the serious version (expected value argument) is coherent but demanding—it requires doing the probability work; the casual version parasitizes the credibility of the serious version without earning any of it. In short, the people who wanna tell you their P-Doom number are terrible at explaining the mechanisms, frictions, and counterfactuals they weighed to get there. Just like a guy who took the elevator to the roof can’t tell you what the stairwell really looks like. They want to admire the view, but they haven’t done the work to get there.

This is why I call out the social respectability specifically. Institutional prestige laundering and intellectual deference cascades run rampant on this topic. Demanding the staircase gets climbed now reads as unsophisticated. Epistemic rigor is replaced by fantastical imaginations masquerading as social proof. The argument is treated as “serious” because serious-seeming people said it in serious-seeming rooms. But that is not how probability works, that is not how we make or assses predictions.

---

Why the Stairs Go Unclimbed

Here’s the brutal reality of the pernicious incentives at play: the rooftop observation pays, climbing the staircase doesn’t. When the topic is earthquakes in California, you can write a compelling screenplay or eye-catching insurance sales pitch. You might even convince some people to move. Doing the actuarial or surveying work to establish your own evidence? That’s boring and no one wants to read it.

Same goes for civil war. Cite faction identity, historical grandeur, and moral urgency above all else. Leverage the pleasure of feeling awake to danger. Don’t worry about the sticky questons like removing a president from office, having a competing government with its own military, or figuring out what single, regime-breaking issue would even be worg fighting one’s neighbors over in the first place.

AI apocalypse? Pish-posh. There’s species-level stakes and self-importance so total that it makes all other concerns look parochial. It’s the ultimate high, the ultimate meta issue. Who cares that you can’t get it off the ground floor if you actually work the steps? There are entire enclaves of California, New York, and foreign cities that will pay handsomely just to read a confident prognostication from a purported “expert” on the topic. And with the AI issue in particular, being an “expert” is mostly just a matter of declaring yourself one for long enough that people no longer bother to ask why. Worked for the Cult of Yud. It could wotk for anyone!

The sad fact is that saying the boring truth, that conditions are bad but the structural prerequisites are absent doesn’t generate social media discourse, status or social capital, nor the pleasures of moral seriousness & urgency. It gets written off of consensus opinion (often with a note of derision about being a sheep) at best, or out-of-touch Boomerism at worst. The conclusion that catastrophists want gets selected first for its emotional payoff; conceivability is recruited after the fact as minimum-viable justification. It could happen does the job of it probably will because the audience, primed by narrative richness, doesn’t notice the difference.

---

Conclusion — What Honesty Costs

To be clear about what this framework is not: it is not an argument that nothing terrible can happen. Terrible things happen constantly, some of them without warning, outside any reasonable probability estimate. The staircase is not a guarantee of safe passage. It is a demand for honest passage.

The demand is simple: show the mechanism. Name the preconditions. Account for the friction. Anchor to base rates and be mindful of the outside view. If the probability estimate at the top of those stairs is still high enough to warrant alarm, make the case. Tetlock’s superforecasters would recognize the form. Schneier would recognize it. The argument is available to anyone willing to do the work. But possibiliter ergo probabiliter—the oldest shortcut in the catalog of bad reasoning—is not that argument. Neither is Cheney’s One-Percent doctrine. That lack of substance can’t be elided by dressing it in contemporary urgency or unearned self-importance.

There is a version of real seriousness that involves sitting with the boring answer. That conditions are bad. That the trajectory is concerning in specific, nameable ways. That the probability of the dramatic outcome is, after honest assessment, not high. That version of seriousness simply doesn’t go viral. It doesn’t have the shape of a thriller. It doesn’t offer faction identity or the pleasure of feeling awake at the edge of history. It offers only the considerable advantage of being true. I wish truth were more popular.

The catastrophists will tell you there’s no time for the stairs. The situation is too urgent, the stakes too high, the moment too critical for the pedantry of mechanism and precondition inventory. Notice what this argument does: it uses the urgency generated by the unearned conclusion to pre-empt the scrutiny that would have prevented the conclusion in the first place. The stakes are too high to slow down—but the stakes were established by skipping the stairs. This is circular reasoning, on top of being terribly self-serving. It is the rhetorical equivalent of lighting a fire and then insisting there’s no time to ask who started it.

Let’s stop fires before they start, not with catastrophism, but with care. Let’s climb the stairs. They are not optional in this building. Climb them, show your work at every floor, and whatever view you find at the top, you’ll have earned it. Anything else is just standing on the ground and fantasizing about a rooftop you’ve never actually been to.