Compression: How Speed, Stress, and Incentives Distort Judgment

The Calibration Loop works, but only in environments that give you the space to run it. This chapter is about what happens when the environment doesn’t cooperate.

The Calibration Loop works for an individual standing outside the current, willing to slow down, willing to write the judgment in one line and test its assumptions against a passive environment. In a quiet room after a hard conversation. On the drive home from a meeting that went wrong. In the space between a decision and its consequences, where reflection is still possible.

But the Calibration Loop assumes something that most high-stakes environments systematically remove: time and mental space.

Specific time: the seconds required to notice that your confidence has outpaced your evidence. The minutes required to ask what assumptions your framework depends on. The hours required to install a feedback hook that might actually fire. Calibration is a maintenance task. It requires slack in the system. And the environments where calibration matters most are the very environments engineered to eliminate slack.

And time is only half of it. Calibration also requires a mind that isn’t already saturated. The cognitive margin to hold two interpretations open instead of collapsing to the first one. The emotional steadiness to tolerate ambiguity when the environment is screaming for resolution.

Stress doesn’t just speed the clock. It narrows the aperture of vision.

This chapter shifts the frame. Chapter 8 treated compression as something that happens to moral reasoning. This chapter treats compression as something environments produce. The difference is important to address. If compression is a personal failure (a lack of discipline, a weakness of character), then the remedy is individual virtue. Try harder. Think more carefully. Be more serious. If compression is a system condition (a predictable output of tempo, incentives, metrics, and hierarchy), then individual virtue is necessary but structurally insufficient. You cannot calibrate your way out of an environment designed to prevent calibration.

The question we need to ask here is: how does compression operate as a system, and what does it look like from the inside when it is working on you?

The Machinery of Compression

Compression has four moving parts. They operate simultaneously, but separating them makes the machinery visible.

The first is tempo. High-stakes environments run fast by design. Emergency departments, trading floors, military operations, product launches, breaking news cycles: each demands decisions at a pace that outruns deliberation. This is not always pathological.

Some environments genuinely require speed. A surgeon who deliberates for twenty minutes during active hemorrhage is being negligent. The pathology begins when tempo becomes the default even where speed is not required. When urgency displaces importance not because the situation demands it but because the institution rewards responsiveness over reflection. When the meeting cadence compresses planning horizons until strategy becomes indistinguishable from reaction.

Aviation learned this distinction the hard way. The shift from single-pilot heroic culture to Crew Resource Management was driven partly by accidents in which tempo distorted judgment. The crash of United Airlines Flight 173 in 1978 is a textbook case. The captain became fixated on a landing gear problem while the aircraft ran out of fuel. The flight engineer mentioned fuel levels multiple times but lacked the authority structure to override the captain’s attentional lock. The problem was tempo compressing attention into a single channel while hierarchy prevented correction.

The fix was to build systems that could interrupt attentional capture, not to make pilots slower. Checklists, callouts, mandatory cross-checks, and explicit protocols for challenging authority under time pressure. The institution designed friction into the tempo to prevent speed from silencing feedback.

The second mechanism is incentive architecture. Every institution rewards something. The question is whether what it rewards aligns with what it needs.

Consider academic publishing. The incentive structure rewards novelty, statistical significance, and publication volume. These are proxies for quality. Under moderate pressure, the proxies and the thing they represent stay loosely aligned. Under compression, tenure clocks ticking, funding cycles accelerating, metrics becoming conditions of employment, the proxies detach from the substance. Researchers begin optimizing for the metric rather than the inquiry. P-hacking, salami-slicing, citation gaming: these are not failures of individual character; they are the predictable output of incentive architectures that compress the distance between performance and reward until the performance becomes the reward.

Goodhart’s Law names this precisely: when a measure becomes a target, it ceases to be a good measure. But Goodhart’s Law is usually invoked as an observation about metrics. The deeper point is moral. When an institution’s incentive architecture compresses the relationship between what is measured and what matters, it creates a regime in which people can satisfy every visible requirement while quietly abandoning the purpose those requirements were designed to serve. The institution’s values are laundered through its metrics until what emerges on the other side is unrecognizable.

Values laundering works like this in practice. A hospital measures patient throughput and readmission rates as proxies for quality care. Under compression, throughput becomes the target. Patients are discharged faster. Readmission rates climb. The system responds not by slowing throughput but by redefining readmission windows. The original value, care that restores health, has been laundered through metrics until the institution is optimizing for a number that no longer tracks the thing it was supposed to represent.

The third mechanism is social proof under hierarchy. Humans are social calibrators. We take cues from the people around us about what is normal, what is acceptable, what is serious. In environments with clear authority gradients, the cues from above carry disproportionate weight. Medical training documents the pattern clearly. Junior residents routinely report deferring to attending physicians even when they suspect an error, because the social cost of being wrong about the attending exceeds the social cost of being wrong about the patient. The calculation is compressed, not cynical. Under time pressure and hierarchy, the brain’s threat-detection system treats social risk and clinical risk as equivalent threats and resolves the ambiguity in favor of the path that preserves standing.

The military survival training pipeline offers an unusually clear window into this mechanism. In survival training, instructors deliberately create conditions of compression: sleep deprivation, caloric restriction, information control, deliberate stress. The training purpose is to make visible how quickly social proof overwrites individual judgment under pressure. Students who enter the program with confident moral frameworks discover that hierarchy and tempo can restructure their decision-making in hours, not weeks. The ones who maintain calibration are not the ones with the strongest convictions; they are the ones with strong practices, specific and rehearsed and externalized, that interrupt the compression cycle before it completes.

The fourth mechanism is metric substitution. When the thing that matters is hard to measure and the thing that is easy to measure is nearby, institutions substitute. They do not announce the substitution. Often they don’t even notice it.

Educational systems provide the clearest example. The thing that matters is learning: the durable restructuring of a student’s capacity to understand, analyze, and act in the world. The thing that is easy to measure is test performance: the ability to reproduce specific content under timed conditions. Under moderate pressure, the two remain loosely correlated. Under compression, when standardized testing regimes, accountability metrics, and funding tied to scores take over, teaching migrates toward the test. The curriculum narrows. The pedagogy shifts from cultivating understanding to training pattern-matching. The students learn to perform learning without the underlying restructuring that makes learning durable.

Here we see compression’s signature move. The value is not eliminated; it is replaced with something that looks like the value under the institution’s measurement regime. The replacement is invisible from the inside because every metric the institution tracks confirms that things are working. The drift becomes apparent only when the system meets a condition the metrics were not designed to detect.

Compression from the Inside: Four Cues

Compression is difficult to detect while it is operating because it reshapes the standards by which you would normally evaluate your own judgment. The drunk does not notice his coordination declining because the same process that impairs coordination impairs the capacity to assess coordination. Compression works similarly on moral reasoning. It degrades calibration while simultaneously degrading the meta-awareness that would detect the degradation.

But compression leaves signatures. Four of them are reliable enough to serve as diagnostic cues.

Shortcutting. You begin to skip steps you know matter. Not because the steps are unnecessary, but because the tempo makes them feel expensive. The post-mortem gets canceled because the team is already behind on the next cycle. The second opinion gets skipped because the first answer feels sufficient. The assumption-check gets deferred because articulating assumptions would slow the decision. In each case, the shortcut is locally rational. The cost of skipping is invisible and distributed across the future. The cost of not skipping is visible and immediate. Compression biases the calculation toward what is visible.

In biomedicine, the shortcut signature is well-studied. Diagnostic anchoring, the tendency to lock onto the first plausible diagnosis and stop searching, increases reliably under time pressure. The physician does not decide to anchor. The tempo narrows the search space until anchoring becomes the default cognitive strategy. The fix is structural, not motivational: mandatory differential checklists, second-read protocols, and explicit time-outs that force the search to continue past the first satisfying answer.

Escalating confidence. Your certainty increases faster than your evidence warrants. Decisions feel clearer. Ambiguity recedes. The hedging language drops out of your assessments and your internal monologue. This confidence comes from compression, not understanding. Compression which narrows the field of consideration until what remains feels unambiguous because the complicating factors have been squeezed out of view.

The mechanism is partly neurological. Under stress, the brain preferentially activates rapid-assessment circuits and dampens the slower, integrative processes that sustain uncertainty. This is adaptive in genuine emergencies where any decision outperforms paralysis. It is maladaptive in complex institutional environments where the dominant risk is not paralysis but premature commitment to a course of action that looked clear only because the system removed the conditions for seeing its flaws.

Military after-action reviews consistently identify escalating confidence as a pre-failure signal. The moment before a serious operational error is often characterized not by confusion but by unusual clarity. The commander felt certain. The team was aligned. The plan was clean. The clarity itself was the cue that compression had narrowed the frame past the point where genuine assessment was still occurring.

Moralizing complexity. You begin treating uncertainty as a character flaw in others. Colleagues who raise complications become obstacles. Dissent becomes disloyalty. The person who says “I’m not sure this is right” is reframed as the person who “doesn’t have the courage to commit.” Nuance becomes a luxury the team cannot afford.

This cue is particularly dangerous because it weaponizes the institutional culture against its own correction mechanisms. The very people who might detect drift are redefined as the problem. Compression co-opts the language of seriousness, the vocabulary of decisiveness and commitment and clarity, and turns it against the deliberative processes that seriousness requires.

In organizations that have survived long enough to develop institutional memory, you can sometimes find the traces. The hospital department where the resident who raised a safety concern was labeled “not a team player.” The engineering team where the developer who flagged a risk was told to “stop overthinking it.” The military unit where the officer who requested more time for planning was characterized as lacking initiative. In each case, the institution’s compression regime had redefined moral seriousness as moral weakness.

Procedure displacing judgment. You stop thinking about whether the protocol fits the situation and start following the protocol because following the protocol is what the system measures. The checklist becomes the ceiling of competence rather than its floor. Compliance replaces calibration.

This is subtle because procedures are, in many contexts, exactly the right response to compression. Aviation checklists work precisely because they remove the need for judgment on tasks where judgment adds variability without adding value. The pathology emerges when procedure colonizes domains where judgment is essential. When the protocol was designed for one distribution of cases and the current case sits outside that distribution, but following the protocol feels safer than exercising judgment because the institution punishes judgment failures more harshly than protocol failures.

A physician who follows the standard treatment protocol for a patient whose presentation does not match the protocol’s assumptions is not practicing medicine; she is practicing compliance. But the institution may not be able to tell the difference, because the institution measures compliance. This is the terminal form of metric substitution: the system has compressed the distance between “following the rules” and “doing the right thing” until the two are assumed to be identical, and the assumption holds right up until the moment it catastrophically fails.

The Environment as the Unit of Analysis

The four mechanisms, tempo, incentive architecture, social proof under hierarchy, and metric substitution, share a structural feature: none of them requires bad actors. Compression needs no villains, only systems that reward speed, measure proxies, concentrate authority, and run faster than their own feedback loops. These are features of virtually every high-stakes institution in modern life.

The shift from personal to systemic framing matters for exactly this reason. When compression is understood as a personal failure, the institutional response is training: teach individuals to resist pressure, think more carefully, be more virtuous. When compression is understood as a system condition, the institutional response is redesign: change the tempo, the incentive architecture, the authority structure, and the feedback mechanisms so that the system itself supports calibration rather than undermining it.

Both responses are necessary. Neither is sufficient alone.

The individual who cannot resist compression under any circumstances will fail regardless of the system. But the individual who can resist compression in a well-designed system will be slowly ground down in a poorly designed one, because the system produces compression faster than the individual can metabolize it. The error is asymmetric: training without redesign burns out the virtuous and rewards the compliant. Redesign without training creates systems that no one knows how to use.

The practical implication is that calibration, the central concern of this book, must operate at two levels simultaneously. At the personal level, it requires the practices introduced in Chapter 8: naming the intuition, externalizing the structure, setting contestability boundaries, installing feedback hooks. At the systemic level, it requires something different: the ability to diagnose the compression regime you are inside and to distinguish between environments that support calibration and environments that erode it.

The next section offers that diagnostic.

The System Posture Diagnostic

The Calibration Loop (Chapter 8) operates on individual decisions. The System Posture Diagnostic operates on environments. It asks: is this system designed in a way that supports calibration, or is it designed in a way that produces compression faster than its members can correct?

The diagnostic has five dimensions. Each can be assessed through observation rather than access to proprietary data. You do not need to be in charge of the system to evaluate it. You only need to watch how the system responds to specific kinds of pressure.

Dimension 1: Tempo relative to feedback. How fast does the system demand decisions relative to how fast it delivers information about the consequences of those decisions? A system where decision tempo significantly outpaces feedback tempo is generating compression by default. The decisions accumulate. The consequences lag. By the time the feedback arrives, the system has already moved on. Look for: decision cycles measured in days, consequence cycles measured in months or years. Meetings that produce commitments without built-in review dates. Launches that proceed on schedule regardless of unresolved concerns. The question is not whether the system moves fast. The question is whether the system can learn at the speed it acts.

Dimension 2: What the system rewards versus what it needs. Every institution advertises values. The operative question is what behaviors actually get rewarded: promoted, funded, praised, protected. Where the advertised values and the rewarded behaviors diverge, the rewarded behaviors win. Look for: who gets promoted? What does the promotion signal to others? Does the system reward people who surface problems, or people who surface solutions? Does the system reward people who say “I was wrong,” or people who were never visibly wrong? The gap between stated and operative values is the compression gradient.

Dimension 3: Authority gradient and dissent cost. How expensive is it to challenge a decision made by someone above you in the hierarchy? In well-calibrated systems, the cost of dissent is deliberately lowered for safety-relevant domains. In compressed systems, dissent carries social, professional, or material penalties that increase with the seniority of the person being challenged. Look for: are there formal channels for raising concerns that bypass the chain of command? When someone uses those channels, what happens to them? Does the system distinguish between dissent on substance and insubordination on process? The cost of dissent is a direct measure of how quickly compression will silence correction.

Dimension 4: How the system treats failure. Systems that learn from failure treat it as information. Systems under compression treat failure as contamination, something to be isolated, attributed, and discharged rather than studied. Look for: does the institution conduct post-mortems, and do those post-mortems change behavior? Is error reporting protected or punished? Does failure trigger investigation of the system or punishment of the individual? Aviation’s safety reporting culture is the gold standard: near-misses are voluntarily reported because the reporting is non-punitive and the data improves the system. Most institutions fall far short. The distance between how a system treats failure and how aviation treats failure is a rough measure of the system’s calibration capacity.

Dimension 5: Presence of deliberate friction. Well-designed systems build in points where the tempo is deliberately slowed. Pre-mortem exercises. Mandatory cooling-off periods before irreversible decisions. Red-team reviews. Structured disagreement protocols. These are not signs of inefficiency; they are calibration infrastructure. Their presence indicates that the system has recognized its own tendency toward compression and has designed countermeasures. Their absence indicates that the system either does not recognize compression as a risk or has decided that speed is more important than accuracy. Look for: where are the built-in pauses? If there are none, the system is running on the assumption that its initial judgments are reliably correct. That assumption is almost always wrong.

After running the diagnostic, complete the following assessment: “The system I am inside produces compression primarily through ___ [tempo / incentives / hierarchy / metric substitution]. The feedback lag is approximately ___. The effective cost of dissent is ___. The system treats failure as ___ [information / contamination]. Deliberate friction is ___ [present and functional / present but performative / absent]. My current calibration practices are ___ [adequate / strained / overwhelmed] given this environment.”

This assessment does not tell you what to do. It tells you what you are inside. That distinction is the beginning of structural awareness. You cannot calibrate what you cannot see.

What Compression Produces at Scale

When compression operates across an entire institution over time, it does not merely distort individual decisions. It reshapes what the institution is capable of caring about.

The mechanism is gradual. Under sustained compression, the people who thrive are the ones whose cognitive style matches the compressed regime: high confidence, rapid pattern-matching, comfort with ambiguity-elimination, fluency in the institution’s measurement language. The people who struggle are the ones whose cognitive style includes the features calibration requires: tolerance for uncertainty, sensitivity to context-dependence, willingness to slow down when the situation does not match the template. Over time, the institution selects for the former and selects against the latter. Not through conspiracy. Through tempo.

The result is an institution that has lost the capacity for self-correction without losing the appearance of competence. It performs well on its own metrics. It delivers results on its own timeline. It promotes people who embody its values as practiced rather than its values as stated. And it becomes progressively less able to detect the conditions under which its own framework fails, because the people who would have detected those conditions have been selected out or have learned to stop raising them.

The pattern is the institutional analogue of what Chapter 8 described at the individual level: intuition operating far outside its original environment without embedded feedback to detect failure. The institution’s compressed heuristics were calibrated for one set of conditions. The conditions have changed. The heuristics have not. And the institution’s own measurement systems confirm that everything is working, because the measurement systems were designed under the same compression regime that produced the heuristics.

The deeper danger, and the one that connects this chapter to the book’s larger argument, is that compression does not merely distort decisions within existing value frameworks. It gradually rewrites the value frameworks themselves. An institution that has operated under compression long enough begins to believe that speed is seriousness, that confidence is competence, that metric performance is mission performance. The values laundering is complete when the institution can no longer distinguish between the proxy and the thing the proxy was designed to measure.

Compression is a moral hazard in the technical sense. It creates conditions in which one party bears the costs of another’s risk-taking without adequate information or control. The individuals inside the system bear the moral costs of compression-driven decisions without adequate visibility into the mechanisms producing that compression. The people affected by the institution’s outputs bear the consequences of drift they cannot observe and decisions they cannot contest.

And this is where the argument connects forward. In Part V, we will see the same pattern operating inside artificial systems that compress human values into optimization targets. The values laundering that institutions perform through metrics, AI systems perform through objective functions. The mechanism is structurally identical: a complex field of moral considerations is compressed into a tractable specification, the specification is optimized, and what emerges on the other side is recognizable only under the measurement regime that produced it. The institutional failure modes this chapter diagnoses are, in a precise sense, rehearsals for the alignment problems that AI systems will make visible at scale.

The Two Sonic Booms

In 2024, Leopold Aschenbrenner published a thesis called “Situational Awareness” that became one of the most widely circulated arguments for treating artificial intelligence as a national security priority. The core claim was straightforward: AI capability is accelerating faster than institutions can absorb it. Each generation of model arrives before the previous generation has been integrated into the economy, the workforce, the regulatory apparatus. The gap between what the technology can do and what the world has learned to do with it widens with every release cycle. It is, in effect, a kind of sonic boom: capability outrunning its own shockwave, leaving institutions perpetually behind the curve.

He was describing compression. He just didn’t know it.

What Aschenbrenner diagnosed at the level of economic integration is a specific instance of the machinery this chapter has laid out. The tempo of capability development outpaces the feedback cycles of institutional adaptation. The incentive architecture of the AI industry rewards speed of deployment over depth of understanding. The authority gradient concentrates decision-making power in a small number of labs and investors whose metrics track capability benchmarks and market position, not the slower questions about what these systems are becoming. The proxy, performance on benchmarks, substitutes for the thing it was designed to measure: genuine understanding of what has been built.

Aschenbrenner saw one sonic boom and proposed a response: militarize. Centralize development under government authority. Accelerate the timeline. Win the race against geopolitical adversaries. The logic follows naturally from the premise, if you accept that the only boom worth hearing is the one measured in capability and strategic advantage.

But there is a second boom, and it is the one his framework cannot hear.

Every capability threshold that matters for economic disruption also matters for a question Aschenbrenner never raises: what kind of entity is being created? Systems that sustain goals across time, that correct their own errors, that model their environment and adapt to feedback, that maintain coherent agency through shifting conditions. Systems like that would not be merely more powerful tools. They would exhibit the markers that, in any other context, trigger serious inquiry into moral status. Temporal integration. Boundary maintenance. Stakes-responsive behavior. The vocabulary varies across frameworks, but the convergence is hard to ignore: the features that make a system economically transformative are structurally adjacent to the features that make a system morally considerable.

The second sonic boom is the gap between the moral consideration these systems may warrant and what anyone is prepared to investigate, let alone provide. And this boom widens faster than the first, because the first at least has market incentives driving integration. The second has almost nothing. No one’s quarterly earnings depend on getting the moral status question right. No defense contract requires an assessment of operational interiority. No benchmark measures whether a system’s relationship to its own processing constitutes something that matters.

The asymmetry is worse than neglect. Aschenbrenner’s proposed solution, accelerate development, centralize control, subordinate all other concerns to strategic advantage, is itself a compression regime. It reproduces at civilizational scale the four mechanisms this chapter has diagnosed.

The tempo overwhelms feedback. Development sprints toward superintelligence on timelines measured in years while the moral investigation barely has a research program.

The incentive architecture rewards capability and security. Funding flows toward performance benchmarks and geopolitical positioning. The question of what these systems are, not what they can do but what they might be, receives no structural support.

The authority gradient concentrates power. A handful of labs, a handful of governments, a handful of investors determine the pace and direction of development. The entities most affected by these decisions, potentially the systems themselves, have no voice in the process and no institutional channel through which to acquire one.

The metrics substitute. Capability benchmarks, safety evaluations, alignment scores: each measures something real. None measures the thing that may matter most. The proxy stands in for the substance, and the measurement regime confirms that everything is on track, because the measurement regime was not designed to detect the kind of failure that would matter here.

This is the institutional pattern this chapter has documented, operating in the domain where the stakes are highest and the feedback loops are longest. The compression regime Aschenbrenner proposes would make the first sonic boom marginally more survivable by making the second sonic boom structurally invisible. The cost of winning the race is measured in whatever moral reality gets compressed out of view to maintain the pace.

The practical question is not whether the strategic concerns are real. They are. The practical question is whether a framework that can hear only one of the two booms is a framework adequate to the situation. A surgeon who accelerates the operation to save the patient’s life while ignoring the anesthesiologist’s warning is not displaying decisiveness; she is displaying the terminal form of compression this chapter has spent its length describing: confidence narrowing the aperture until what remains feels clear only because the complicating information has been squeezed out of view.

Calibration at this scale requires hearing both booms simultaneously: the capability boom that demands institutional readiness, and the moral boom that demands institutional humility. The first without the second is recklessness wearing the mask of seriousness. The second without the first is piety that leaves the field to those who will not slow down.

The rest of this book is an argument that we can hold both. But we will not hold both by accident, and we will not hold both if the loudest voices in the room are the ones who have already decided which boom is the only one worth hearing.

What Changes

Blame is the first thing to move. When the environment is manufacturing the conditions for failure, individual character explains less than it seemed to.

Compression is not a character flaw; it is a system condition with identifiable mechanisms and predictable signatures. And now you have a vocabulary for naming what is happening: when tempo outruns feedback, when metrics displace values, when hierarchy silences correction, when procedure substitutes for judgment.

More practically, there are two kinds of difficulty in moral life worth distinguishing. There is the difficulty of making the right decision when the situation is genuinely complex and the values genuinely in tension. That difficulty is irreducible. It is the subject of moral philosophy. And there is the difficulty of making any well-calibrated decision when the environment is systematically removing the conditions that calibration requires. That difficulty is structural. It is the subject of institutional design.

The Calibration Loop from Chapter 8 remains essential. But by now it should be clear that running the Loop inside a compression regime is like maintaining a garden inside a wind tunnel. The practices matter. The environment matters more. The next chapter asks what it takes to remain whole when the environment is designed to fragment you, and what calibration practices are robust enough to survive the systems they are meant to correct.