1.1Why a literature review in a book of questions
A book made from decision records still owes its reader an account of what has already been said about the subject those records concern. Meridian's fifty defence questions are not the only writing on war, peace and the decisions taken at their edges. Four large literatures have long argued about how wars begin and end, how warnings are read and missed, who directs armed forces and on what grounds, and what is owed after mass violence. None of them was written to read a collection of decision questions. Their subjects are wars, states, institutions, doctrines and ideas; their evidence is history, archives, theory and cases examined after the fact. This chapter sets out what each of the four says, in the terms of its own reading notes, and marks where Meridian's records can bear on the argument and where they cannot.
A review in this position is a piece of navigation rather than a survey. The chapter does not weigh the four literatures against one another, does not decide which of them is right about any episode of war, and recommends no option in any of the fifty cases. It establishes three things for the chapters that follow. First, the questions each literature asks, so that a reader can see why a decision record answers some of them and is silent on others. Second, the vocabulary this study borrows — the tract of time, friction, the commitment problem, meaningful human control — each term taken up at length in the chapter that needs it rather than here. Third, the narrow gap the study occupies. What fills the gap is a reading of how one question is written down before its answer is known.
The four literatures do not share a method, and the difference matters for what a decision record can contribute to each. The first works from history and from theory about time: Thucydides reports an assembly and an island, Hobbes and Clausewitz define conditions and purposes, and the modern studies of war termination compare wars that have already ended. The second works from cases of warning and surprise, most of them examined after the attack, and from theories of how people and groups handle information they would rather not have. The third works from institutions: constitutions, officer corps, oversight bodies, and the daily traffic between elected officials and military organisations. The fourth works from the aftermath of mass violence — trials, commissions, reparations, records — and from arguments about what one generation may bind another to. Meridian's records share none of those bodies of evidence. They hold a question before its answer, with a mandate, a deadline, named options and a list of what is missing. Each of the chapters that follow therefore uses one literature as a lens on what a question writes down, and says where the lens fits badly. A literature can tell a reader what to look for in a question; it cannot tell the reader what the question should have said.
The four literatures appear here in brief and are taken up at length in the chapters that own them. The first, on war as time and on its ending, is treated in Chapter 3 (Thucydides on the reversed decision and Kant on a peace that hides the next war; Iklé, Goemans, Fearon and Walter on ending; Fortna on ceasefires) and Chapter 5 (Hobbes and Clausewitz on war as a tract of time and as policy carried on by other means). The second, on warning, perception and surprise, is Chapter 4's. The third, on civilian control and on what armed organisations do when left alone, runs through Chapter 5 and Chapter 6. The fourth, on what comes after war and on promises that outlast their makers, is Chapter 7's and Chapter 8's. What follows is therefore a map, not a verdict: each literature reported in the terms of its reading note, with a closing paragraph saying what the records can and cannot show against it.
1.2First literature, war as time and its ending
The oldest of the four concerns war as a period of time rather than a single act, and the act of ending it.
Two features of that period shape the literature that follows. A war has a beginning and an end, each decided somewhere, and between them lies a stretch of time in which the parties adjust to one another and to their own costs. Studies of how wars end therefore face a question that studies of how they begin do not: what the decision to stop does to the people who make it. That question is where Iklé, Goemans and Fearon part company with one another. It is also where a decision record has something of its own to offer, because the record is written by the people the question is about, before the stopping has cost them anything.
Thucydides, in the History of the Peloponnesian War, gives both sides of that period. In Book III the Athenian assembly votes to put the men of Mytilene to death, and then reconsiders the next day after Cleon and Diodotus have spoken, whereupon the motion of Diodotus carries by a near-even show of hands. A second galley leaves a day and a night behind the first, its rowers eating as they row, and puts into port just as the commander is preparing to carry out the first decree, and prevents the massacre. The leaders of the revolt, rather more than a thousand, are put to death on Cleon's motion. This study reads the record as being about the time of a decision as much as about its content: a decree made and unmade within the time a ship needs to cross a sea. In Book V the same historian stages the Melian dialogue, in which the Athenian envoys argue from power and interest while the Melians argue from justice and hope, and in which the argument is settled by siege and by killing rather than by either side's reasoning. Chapter 3 takes up both passages, the Mytilenean debate at greater length.
Hobbes, in Leviathan, Chapter XIII, gives a definition, and this study reads the Mytilenean record as living inside it: without a common power, men are in a condition of war, and war lies not in battle only but in a tract of time in which the will to contend by battle is sufficiently known; all other time is peace. The definition carries a consequence that matters for every record in this study: peace is whatever the tract of war is not, and what separates the two conditions is the will to contend being known or not known. This study reads the rest as its own inference: no decision makes that will true or false. A decision about the beginning or the end of the tract settles which period a society is recorded to be in, on the evidence of that will; it does not settle the condition itself. Hobbes also answers the question of what holds a common power together, by the covenant of subjects with one another to make the sovereign, in Chapters XVII and XVIII. Chapter 5 returns to both.
The covenant point matters to a collection of questions as much as the definition does. A decision question names a body that is to decide — a committee, a council, an authority — and behind the name sits the question Hobbes answers: what holds such a body together. His answer is an agreement among those who will be subject to it. This study reads several questions here as writing the terms of such an agreement rather than its fact: who sits at the table, what a withdrawal costs, what a generation that did not vote inherits. The record can show the terms being offered and the body offering them; it cannot show anyone consenting.
Clausewitz, in On War, Book I, Chapter 1, holds that war is a continuation of policy by other means. This study reads that as making war answerable to a purpose outside itself, and an ending, when it comes, a political act as well. Book I, Chapter 7 names friction, the condition in which the simplest thing in war becomes difficult; Book I, Chapter 3 holds that much of what action in war rests on is hidden in uncertainty. Friction and uncertainty are the two terms this study will use most often when a record lists what a room did not know. Chapter 5 treats On War alongside Hobbes.
Clausewitz's sentence is often read as a claim about why wars are fought; this study reads it also as a claim about who is owed an account of them. If war carries on policy by other means, then the authority behind the policy can be asked what the means are for, and asked again when they stop working. Friction and uncertainty set the limit on that questioning from the other side: the purpose is settled outside the fighting, but what happens inside the fighting is only partly knowable in advance. Every question in this study stands at that junction. Each gives a mandate and a deadline to a named body and lists what the body does not know, which is one way of writing down both halves of the picture: what the decision is for, and what the room cannot see.
Kant, in Perpetual Peace: A Philosophical Essay, writes the terms on which peace might be kept. The first of the six preliminary articles is that no treaty of peace is valid if made with the secret reservation of material for a future war; the sixth bars the use of hostilities that would make mutual confidence in a later peace impossible. The three definitive articles are a republican civil constitution, a federation of free states, and a cosmopolitan right limited to universal hospitality. Chapter 3 reads the first preliminary article against the ceasefire records; Chapter 7 reads the federation against the alliance records.
Against those old texts stand three modern arguments about how wars actually end. Fred Charles Iklé, in Every War Must End (1971), argues that wars are ended by decisions, and that governments often fail to think those decisions through when they begin their wars. H. E. Goemans, in War and Punishment (2000), argues that whether and when leaders end a war depends on what ending it will cost them at home, and that leaders who expect to be punished for a lost war may fight on in the hope of turning it round, as in the First World War. James D. Fearon, in "Rationalist Explanations for War" (1995), argues that because war is costly, rational states should prefer some negotiated bargain to fighting, and that war nevertheless arises from private information with incentives to misrepresent it, from commitment problems, and from issues that cannot be divided. The three are not the same argument: Iklé is about the decision to end, Goemans about the decision-maker's domestic cost, Fearon about the bargaining position that makes war rational to avoid and difficult to avoid.
Iklé's short book makes a demand that a decision record can answer in part: think about the ending before the beginning. His subject is a government that starts a war without deciding how it will stop. Goemans adds a variable that no question in this collection carries: the leader's own fate at home after a lost war. The fictional settings have no electorate, no party and no punishment waiting for a leader, so they cannot show what he describes; the nearest thing they hold is the account of who waits and pays while a decision is pending, and even that names roles rather than political futures. Fearon's account supplies the vocabulary this study uses most for why a war can start when both sides would rather have a bargain: private information with incentives to misrepresent it, commitment problems, and issues that cannot be divided. A question written before its answer often contains all three at once, in the unverified account a party gives of its own strength, in a promise that would outlast the party making it, and in a subject that admits no shares.
Two further works ask why some agreements hold and others do not. Virginia Page Fortna, in Peace Time: Cease-Fire Agreements and the Durability of Peace (2004), finds that ceasefires backed by stronger agreements — buffer zones, peacekeepers, joint commissions, verification — tend to last longer. Barbara F. Walter, in Committing to Peace (2002), finds that settlements of civil wars fail mostly at the point where the sides must disarm and can no longer protect themselves, which is a credible commitment problem, and that third-party security guarantees and power-sharing arrangements make a settlement more likely to hold.
Walter's finding and Fortna's stand near each other, and keeping them apart clarifies what each can do for a question. Fortna asks what makes a ceasefire last once it is signed, and answers with the machinery placed around it: buffer zones, peacekeepers, joint commissions, verification. Walter asks why a civil-war settlement fails, and answers at the point of disarmament, where a party that gives up its weapons cannot protect itself and has no reason to trust the other side to keep its part. Both therefore point at the same ground: what a party is asked to give up, and who stands behind the giving. Neither work tells a decider which option to take; each names what an option has to survive — a party left unable to protect itself, or an agreement with no machinery to hold it. A question written in advance cannot supply that machinery, but it can ask whether any is proposed, and by whom.
What the records can bear: the ceasefire questions in this study name who is at the table and who is left outside it, what would reopen an agreement, who keeps observation records, and by when the answer must come. Fortna's list of what makes a ceasefire stronger — verification among them — is the kind of thing a question can write into its options, and Chapter 2 sets out what the question form guarantees. What the records cannot bear: they contain no war, so they cannot show why a war ends, what it costs a leader at home, or whether a bargain was rational. These questions are held before their answer, not after their history.
1.3Second literature, warning, perception and surprise
The second literature asks why decisions about war are taken on bad information, or on good information that nobody believed.
The question has two halves and the literature divides accordingly. The first half is about information: whether the warning was there to be found, and what else was there to be found with it. The second is about belief and action: whether the warning, once found, was credited and used in time. A decision record sits awkwardly between the two, because it can show what was written down at the moment of deciding and cannot show what anyone made of it. That awkwardness is why this literature is the hardest of the four to read against a question held open before its outcome.
Roberta Wohlstetter, in Pearl Harbor: Warning and Decision (1962), finds that the signals of the attack on Pearl Harbor were present beforehand but buried among irrelevant and misleading ones, which she calls noise. The failure, in her account, was in telling signal from noise and acting on it, not in the absence of information. Richard K. Betts, in Surprise Attack: Lessons for Defense Planning (1982), finds that surprise attacks have usually succeeded not for want of warning but because warnings were not believed or not acted on in time, and that the obstacles are political and organisational as much as a lack of information. The two accounts differ in emphasis more than in the place of failure: Wohlstetter's runs from telling signal from noise through acting on it, while Betts puts the weight on warnings being believed and acted on in time. Chapter 4 reads the debate in full.
Both books are written after the event they explain, which is the position a decision record never occupies. Wohlstetter can line up the signals that were present before the attack against the noise that hid them because the file can be read from end to end. Betts can compare attacks that succeeded with warnings that existed because the comparison is retrospective. A question posed before the outcome has no such file. What it has instead is the room's own account of what it holds and what it lacks, written while the outcome is still open, and that is what makes the second literature worth carrying into the chapters that follow.
Robert Jervis, in Perception and Misperception in International Politics (1976), supplies an explanation for the second point: decision-makers fit new information to what they already believe, and misread the intentions of others.
Jervis's claim is about fit: a decision-maker reads the new through the old, which is efficient when the old is right and costly when it is not. The intention of the other side is the hardest case, because it is never in the file; it is inferred, and the inference is drawn from what the reader already expects. This study reads a question that lists what is missing as a partial defence against that, because it names the unverified before the reading begins. The record will show the list, and not the reading. Irving L. Janis, in Victims of Groupthink (1972), describes a failure of a different kind, inside the group: cohesive policy groups under pressure come to seek agreement over a realistic appraisal of alternatives, suppressing doubts, as in the Bay of Pigs decision. Janis proposes remedies of the same practical kind, such as assigning a critical evaluator and inviting outside experts. Graham T. Allison, in Essence of Decision: Explaining the Cuban Missile Crisis (1971), reads one crisis through three models: the rational actor, organisational process, and governmental politics. This study reads Allison's three models as three rival explanations of the one crisis rather than as three levels of decision, and reads the choice among them as a choice about where an explanation for a decision is thought to reside.
Thomas C. Schelling writes from the other side of the table. In The Strategy of Conflict (1960) he describes focal points that let parties coordinate without talking, and commitments that bind the one who makes them; in Arms and Influence (1966) he separates deterring from compelling, and describes threats that leave something to chance. This study reads a difference of emphasis here: Wohlstetter, Betts, Jervis and Janis describe how a situation is appraised, while Schelling's focal points, commitments and threats that leave something to chance describe what one party's move does to another's choices. The two directions meet in any crisis record: a warning is both something received and something published.
Schelling's terms are tools for reading an option list as much as for reading a crisis. A focal point is a place two parties can arrive at without talking; a commitment is a move that binds the party that makes it; a threat that leaves something to chance is one whose outcome the speaker does not fully control. A question that offers its decider a deadline, a signature or a public promise is asking for thought in exactly those terms. The record will show which option was taken and what grounds were written down for it, and nothing about how the option was read across the table.
What the records can bear: every question in the collection lists what is missing, and the ledger counts 148 such items across the fifty questions. Several questions carry a verification step, an independent signal, or a named role whose consent or objection would settle an unknown — the kind of structure Betts's account of organisational obstacles and Janis's remedies both point toward. What the records cannot bear: a decision record writes what a room knew and wrote down, not how the room read it, and it cannot show whether a signal was noise to that room or whether a group suppressed a doubt it had. The limit is on verifying behaviour rather than on representing it: a form can record an expressed belief, an organisational arrangement, an objection and a monitoring responsibility, and still cannot show what the room did with any of them. Chapter 4 reads four crisis-line questions against these works and says where the comparison stops.
1.4Third literature, civilian control and military organisations
The third literature asks who directs armed force, and what organisations do when the answer is ambiguous.
The ambiguity has a history in the writing itself. Some of these authors argue about a design: how the two spheres should be arranged so that force serves policy without politics entering the profession. Others argue about a relationship: what elected officials and military organisations actually do to one another from day to day, and how a scholar can tell whether a design is working. A collection of questions cannot settle either argument, but it can show which parts of an arrangement a body is willing to write down — who gives an order, who may stop it, who funds the stopping, and who reviews it afterwards.
Samuel P. Huntington, in The Soldier and the State (1957), argues for objective civilian control through a professional officer corps that stays out of politics. This study reads that as keeping the two spheres apart, with control achieved by the military's professional competence and political neutrality rather than by civilian involvement in military matters. Morris Janowitz, in The Professional Soldier (1960), argues from a different observation: that the military profession was coming closer to civilian society in its skills and outlook. Janowitz therefore looks to civilian control through the officer corps' integration with society and to a constabulary conception of force, against Huntington's separation. The two agree that control matters and disagree about what kind of relationship makes it work.
This study reads the two designs as each carrying a cost the other answers. A professional corps kept out of politics is also kept at a distance from the conversation about what force is for; a corps integrated with society carries civilian habits into military judgment and risks importing politics by another door. Neither author's account says which relationship any given state has; the two are arguments about arrangements, and the arrangements show up in a record wherever a body is told who may direct it, who may stop it, and who may later call it to account. This study reads the emergency-rule and oversight questions as writing such an arrangement in advance, in the plain terms of a mandate and a stop authority, and leaves the argument between the two authors where it stands.
Eliot A. Cohen, in Supreme Command: Soldiers, Statesmen, and Leadership in Wartime (2002), attacks the normal theory that civilians set the aims and leave operations to soldiers. Cohen finds that Lincoln, Clemenceau, Churchill and Ben-Gurion questioned and intervened in military matters, and describes what they did as an unequal dialogue. Peter D. Feaver, in Armed Servants: Agency, Oversight, and Civil-Military Relations (2003), treats the same relationship as a problem of agency: civilian principals monitor, and military agents may work or shirk. Where Huntington and Janowitz argue about the design of the relationship, Cohen and Feaver are concerned with what happens inside it — intervention at one end, monitoring and non-compliance at the other. Chapter 5 takes this debate up with the emergency-rule questions.
Barry R. Posen, in The Sources of Military Doctrine (1984), compares two explanations of why military organisations adopt the doctrine they do: explanation from organisation theory and explanation from the balance of power. Posen finds that military organisations left alone tend to prefer offensive, stable doctrines, and that civilians intervene to change doctrine when the international threat is high. Stephen Van Evera, in The Cult of the Offensive and the Origins of the First World War (1984), describes the belief itself as a cause: before 1914 European militaries and governments believed that the offence had the advantage, and that belief made mobilisation hasty, made a first move seem urgent, and made war more likely. Posen describes the incentive to prefer a doctrine; Van Evera describes what such a preference can do to a decision to move first.
The two therefore differ about where a doctrine comes from. On Posen's side it grows out of how an armed organisation is built and what it is left alone to prefer; on Van Evera's side it grows out of what a country believes about the weapons and the world around it. The difference matters for who is asked to change a doctrine, and for whether change is expected at all. This study reads a question that offers its decider the option of waiting, or of stopping a system, as writing a way to act against the pressure both describe: the pressure to move first, and to treat delay as danger.
Scott D. Sagan, in The Limits of Safety (1993), sets normal-accident theory and high-reliability theory against the record of nuclear command, and finds that near-accidents occurred despite safeguards. This study reads the finding as a warning against designing for perfect reliability and assuming that safeguards hold.
This study reads the two theories as promises about organisations rather than findings in themselves: one treats serious accidents as a property of certain kinds of system however careful their design, the other treats very high reliability as something organisations can reach and study. The outcome Sagan reports sits awkwardly between them: near-accidents occurred despite safeguards. A question that offers a safeguard as one of its options is written in the shadow of that finding. It can ask for a safeguard, name who holds the stop authority, and record the approval that led to it; the record will show all of this without showing that the safeguard would have held. On autonomous weapons, Paul Scharre, in Army of None (2018), describes the pressures of speed and accountability and the question of where the human sits — in the loop, on the loop, or out of it. Peter Asaro, in On Banning Autonomous Weapon Systems (2012), argues that lethal decisions require human judgment that a machine cannot supply, and that a ban on autonomous weapon systems and a requirement of meaningful human control over the use of force follow from that. The two works share a subject and handle it differently: Asaro's draws a conclusion in favour of a ban, and Scharre's supplies the vocabulary and the pressures. The autonomy questions of 2026, 2034, 2042 and 2192 belong to this subject and do not settle it.
What the records can bear: the questions write mandates, funding structures, and — in the autonomy series — who holds stop authority and what record an approval leaves behind. The ledger's coders settle the human place as human-decides in 46 of the 50 cases and find a human role specified in all 50. Several autonomy questions ask the same thing in different words: what would let a later reviewer prove that a person approved, and what would let that person stop a system in time. This study reads those questions as touching the subject Scharre and Asaro treat without being the same questions: the autonomy records concern approval management, oversight and the preservation of records, not lethal decisions made by a machine. What the records cannot bear: they cannot show whether an oversight body would actually monitor, whether a doctrine would change when the threat changed, or whether a safeguard would hold — because none of those has happened in a synthetic setting, and the record holds the decision before it.
1.5Fourth literature, after war
The fourth literature concerns what a society owes after mass violence, and what a promise is worth to people who did not make it.
The first half of that subject is about accounting: what is established, who is blamed, what is repaired, and by which procedure. The second half is about distance: the people owed something are separated from those who owe it by time, by death, or by the simple fact that they were not in the room. The works in this literature divide along that line, and the records in this collection sit mostly on the second side of it, because they are written for bodies that must answer now on behalf of people who will live with the answer later.
Ruti G. Teitel, in Transitional Justice (2000), sets out the forms justice takes after political transition: criminal, historical, reparatory, administrative and constitutional. This study reads the variety as the point of the work: no single procedure answers for every transition. Martha Minow, in Between Vengeance and Forgiveness (1998), examines trials, truth commissions and reparations as responses to mass violence. Priscilla B. Hayner, in Unspeakable Truths (2001), studies truth commissions comparatively: what they can establish and acknowledge, how they differ from trials, and the choices they face about naming perpetrators, reparations and access to their records. The three works are on the same subject at different levels — the forms, the options, and the practice of one of them.
This study reads Teitel's five forms as doing different work: criminal proceedings attribute responsibility to persons, historical accounts establish what happened, reparations transfer something to those who were harmed, administrative measures change who holds a post, and constitutional change rewrites the rules themselves. Minow's three responses are choices within that range, and Hayner's comparison stays with one of them, the truth commission, to show that its design is itself a set of political decisions: whether to name perpetrators, whether to recommend reparations, and who may see what the commission gathers. A question that writes who may see a record, and by when, is writing one of those decisions before any commission exists.
Two works put limits on what either kind of procedure can be made to do. Gary Jonathan Bass, in Stay the Hand of Vengeance (2000), finds that liberal states support war crimes tribunals because their leaders hold legalist beliefs, but that such support is limited: those states are reluctant to risk their own soldiers to bring the accused to trial. Jon Elster, in Closing the Books (2004), compares transitional justice from Athens in 411 and 403 BC to the twentieth century, and asks who is tried, purged or compensated after a change of regime, and what motives — emotions, interests, reasons — and what constraints shape those decisions. Where Teitel lists the forms and Hayner the choices inside one form, Bass describes a limit on the support given to one of them, and Elster sets out who is tried, purged or compensated and what shapes those decisions.
Two further works sit beside the after-war questions: Snyder for the dilemma alliance parties are still inside, and Jonas for obligations that outlast the parties who make them. Glenn H. Snyder, in Alliance Politics (1997), describes the alliance security dilemma: the fear of abandonment against the fear of entrapment. Hans Jonas, in The Imperative of Responsibility (1984), argues that technological power reaches so far into the future that ethics must take responsibility for future generations, and proposes giving the bad prognosis priority over the good — a heuristics of fear — where the stakes are irreversible. Snyder's two fears describe a relationship that is never settled once; Jonas's imperative asks what the present may bind the future to. Chapter 7 reads both against the alliance records of 2026, 2030, 2047 and 2164.
Snyder's dilemma gives a way to read an alliance option list: every option answers one fear at the cost of the other, staying close answering abandonment and adding entrapment, pulling away reversing the exchange. The records of this collection hold alliances in the middle of that exchange, at the moment when a member has given notice, or when a compact survives the peace that once justified it. This study reads those questions as writing the exchange down in terms a body can settle by a deadline: who continues to pay, for how long after notice, and what protection continues meanwhile.
What the records can bear: this study reads the aftermath questions as writing out some of the choices Hayner names — access to the records and the naming of perpetrators — and adds others her study does not reach, such as who may object to a correction and what is done when no witness is left. This study also reads the alliance questions' exit settlements, transition periods and inherited burdens as mechanisms of the kind Jonas's imperative asks for, where the stakes cannot be undone; Snyder's account describes the dilemma of abandonment and entrapment; the fit of his work here is partial. The horizon row of the ledger shows why the second group matters: a few of these questions are written to reach a generation that has not yet voted. What the records cannot bear: they contain no truth commission, no tribunal and no generation that has accepted or refused a covenant, so they cannot show what any of these mechanisms achieves. They can only show how a question writes down who may object and by when.
1.6The gap
The four literatures study cases, states, institutions and ideas. Each offers an account of how wars behave, how organisations fail, how doctrine is chosen, how justice is done after a regime change. None of them reads a collection of decision questions on the same axes for what those questions write about the deadline, the bearer of waiting, what can be undone, and what would reopen the decision. That is this study's narrower place.
The gap is narrower than a new theory and wider than a catalogue. Each of the four literatures explains something that has happened: a war began or ended, a warning was believed or missed, an organisation worked or shirked, a transition chose a procedure. The evidence for each explanation comes after the thing explained, and that is the one thing a decision record cannot have. Meridian's questions are written before their answers, and they are written for a body that must answer by a stated time on a stated mandate. Reading them on the same set of axes allows one question to be compared with the next, which the four literatures do not attempt, because their units are wars, crises, institutions and regimes rather than decisions still open. This chapter has described the ingredients for that comparison; the rest of the study makes it, one axis at a time.
The shape of that place is set out in the ledger, whose axes, cells and two coders Chapter 2 describes. Fifty cases, all synthetic, are coded on the same axes, and every case carries a deadline, a list of what is missing, and conditions for revisiting the judgment.
Set beside that, the eight hypotheses the ledger tests ask questions that none of the four literatures poses in this form: whether the cost of waiting falls on whoever decides; whether the conditions written down for reopening a judgment could be observed by anyone outside the room; whether the cases in which the judgments diverge are the cases whose questions list what is missing; whether the harder a decision is to undo, the further apart are those who decide and those who bear the cost; whether a restriction outlives the memory of its reason; whether the longer the horizon, the less is decided now; whether a decision left to an automatic rule leaves the human's place empty; and whether the options that cut the immediate loss most often give up evolvability and resilience. Chapter 2 sets out the hypotheses and the verdict rule; Chapter 15 reports the verdicts. One verdict is worth naming here only because it defines what the gap is: on the question of whether the conditions for reopening a decision could be observed, the ledger's answer is against the hypothesis in all fifty cases; separately, the ledger counts 798 conditions for revisiting in all, 629 of them stated with a number, a threshold or a time.
What this study does not add is as plain as what it adds. It offers no new account of war, no explanation of any real conflict, and no evidence about real states, forces or people: the fifty settings are synthetic, in fictional polities, and the thirteen judgments published with each are AI-generated. The readings of the classic texts stay within the reading notes the study was given for them; where this study reads a work further than its note, it says so as its own reading. Nor does it settle any of the four literatures' disputes: whether surprise is a failure of information or of belief, whether the separation or the integration of military and civilian spheres is the better design, whether truth or justice should come first — Chapter 4 takes up the first, Chapter 5 the second and Chapter 8 the third, and the records here bear on them only in what they write down. What the collection can show is narrower and, this study argues, worth having: how one question holds the time between being asked and being answered, in whose hands the answer sits, how long the waiting runs, what cannot be undone once made, and what anyone outside the room would have to see before the decision is reopened.
Those five things are what the rest of the study is about, and each of them is a question a reader can ask of any decision, inside war or outside it. The four literatures answer versions of such questions about events that have already happened. These records answer them about events that have not, which is a smaller claim on different evidence, and the chapters that follow keep the difference in view.
Chapter 2 sets out the method.