A record of the past does not tell anyone what to do now. A war history may report that supply lines were cut; an old manual may set out a maintenance procedure; a memoir may recall a decision at a decisive moment. Before any of these can count in a present decision, someone has to say how much of it passes across: the fact it reports, the lesson someone drew from that fact, the rule that lesson might become, or nothing at all. A decision record that begins from such a source has to take a position on that passage, whether it says so or not.
This paper asks what position one designed set of decision records takes. AEGIS, the safety and protection studio of the RISCON family, publishes 6,540 authored cases, each a decision record with a question, three choices, an author's marked recommendation and a list of what must be known before deciding. One small programme among them, a test set of 18 cases in nine new domains, is built around sources of exactly this kind. In 15 of its cases the situation opens by naming a public or declassified record: a war history, an old manual, two classic texts of strategy, a manufacturer's published test results, declassified operational and intelligence documents, a body of public practice, a public wargame, the minutes of a security meeting, a posture assessment, a memoir, a life-extension report, a combat manual and a public debate on verification. Those fifteen are the subject of this paper.
They are authored synthetic training cases, not records of real events. AEGIS types each one realistic synthetic and calls each an authored question seed whose judgment mark is the author's record, not a factual basis from public records. More than that, AEGIS says of the whole programme that every figure, question and actor in it is the author's assumption and refers to no public record. The war history, the manual and the memoir are therefore part of the authored setting. What can be learned from these cases is how a designed record set frames the use of a record, not what any history, manual or document says.
The research question is this: when an authored decision case starts from a public record of the past, what does it say may be carried from that record into a present decision, on what conditions, and what does it rule out?
The argument in brief
The fifteen cases agree on one thing before anything else. Each separates the record from what is drawn from it. The first case puts it in so many words: the record is a description of fact, the lesson belongs to a layer of interpretation, and whether the lesson applies depends on whether today's conditions are those the record reports. In every case the author's recommended option carries the record over only on stated conditions: after a comparison, within a stated scope, in a pilot unit, item by item, by order of risk.
The cases disagree, in a patterned way, on what to rule out. By a rule on the Korean words of the choices, set out in section 1.1, the option the author excludes is a way of holding the record back in 11 of the fifteen: leaving it as reference, preserving it only, never quoting it, deferring it to a later cycle. In only four is the excluded option the literal transfer of the record. Where neglect is excluded, the literal transfer usually survives as the author's alternative, and the reasons given for that alternative soften it, adding the filters and marks that make it look like the recommendation. Read as a set, the cases treat ignoring a record as the graver error and copying it as a fault that can be repaired.
Finally, every case except one lists the record itself as the first thing needed before deciding: the original text, the translation, the released version, the declassification list. None of these is in the record set. No situation gives the record a date or a place; no case carries a source link; and the programme says it refers to no public record. The cases recommend a comparison with an original that they assume but do not hold.
The debate this paper enters
Whether the past and the elsewhere can teach is an old question in the study of policy and war, and the answers divide.
One side holds that records of other times and places can be used, if they are used with care. Richard Rose, What is Lesson-Drawing? (1991), frames the question as one of circumstance and degree: under what circumstances and to what extent a programme that is effective in one place can transfer to another. Lessons can be sought across time as well as across space, and the process ends with a prospective evaluation of what would happen if a programme in effect elsewhere were transferred here. David P. Dolowitz and David Marsh, Learning from Abroad: The Role of Policy Transfer in Contemporary Policy-Making (2000), place lesson-drawing within a wider family of transfer, in which knowledge about policies and institutions in one setting, past or present, is used in another, and they take up the relationship between transfer and policy failure. Barbara Levitt and James G. March, Organizational Learning (1988), describe organisations as learning by encoding inferences from history into routines, and they discuss how organisational memory keeps the lessons of history despite turnover and the passage of time. Michael Howard, The Use and Abuse of Military History (1993), a reprint of a lecture of 1961, grants the tendency to mythologise and the difficulties of evidence but argues that finding out what really happened is worth those doubts, and proposes width, depth and context as the criteria for studying it.
The other side doubts that the passage is made well. Yuen Foong Khong, Analogies at War (1992), argues that leaders use historical analogies not only to justify policies but to perform cognitive tasks in deciding, and draws on cognitive social psychology to explain why policymakers often use analogies poorly. Jack S. Levy, Learning and Foreign Policy: Sweeping a Conceptual Minefield (1994), observes that decision makers seem always to be avoiding the failures of the past and generals always fighting the last war, and that inferences from experience, with the myths that accompany them, often have far more effect on policy than standard rules of evidence warrant. James G. March, Lee S. Sproull and Michal Tamuz, Learning from Samples of One or Fewer (1991), ask how organisations turn infrequent events into interpretations of history, and how they weigh the need to agree on an interpretation against the need to get it right.
A third body of work moves the question from history to evidence. Nancy Cartwright and Jeremy Hardie, Evidence-Based Policy: A Practical Guide to Doing It Better (2012), hold that predicting a policy's outcome is betting that the policy can play the right causal role in a given situation and that the situation has the support factors it needs; one of their chapters warns against fidelity, the injunction to implement a policy exactly as it was done where it worked. Nancy Cartwright, Presidential Address: Will This Policy Work for You? Predicting Effectiveness Better: How Philosophy Helps (2012), states the general problem as using evidence of efficacy from good studies to predict effectiveness here.
The positions do not divide into those for and against using the past. They divide on where the risk lies. For Rose the question is under what circumstances and to what extent a lesson transfers, and he holds that searching for fresh knowledge is not normal; for Howard the benefit of finding out what really happened outweighs the doubts. For Khong and Levy the risk is in the inference itself, which runs ahead of the evidence. For Cartwright and Hardie it lies in carrying a result over without the conditions that made it work. A designed set of decision records has to place its risk somewhere, and it does so in the options it recommends, allows and rules out.
Contribution
This paper adds to that debate a reading of how a record set, rather than a decision maker, encodes the passage from a record to a decision. It makes three claims. First, the fifteen cases build in the separation of fact from lesson and the conditions of transfer that the cautious literature asks for: in this study's reading, the recommended option in each case is close to what Cartwright and Hardie call looking for the support factors, and close to Howard's demand for context. Second, the cases place the greater risk in failing to use a record rather than in using it too readily: they rule out neglect more often than they rule out the literal copy. In this study's reading, which goes beyond what the works say, that sits nearer to Rose's remark that searching is not normal and to Howard's weighing of benefit against doubt than to the concern of Khong and Levy with inference that outruns evidence. Third, the comparison with the original on which every recommendation rests cannot be made inside the record set, because the original is not there. The set states a discipline of source criticism that it does not, and as written cannot, carry out. That gap is the paper's main finding about what a template can and cannot do.