This paper asked of twelve AEGIS records what ties the record of one decision to the record of another, what the records refuse to tie, and who is left to decide the thing two decisions share. The twelve are authored synthetic training cases, not records of real events, and the answers describe how one designed record set frames the tie between decisions. They describe nothing about how any real body has divided a budget or a stock of hours.
Discussion
The set gives a clear answer to the first part of the question and a precise silence in reply to the last. What ties two decisions is one named resource: a stock of aircraft hours, a sum for joint production and cooperation, a budget for minerals or for parts and consumables, a negotiating channel, a stock of operating time at sea, a system for sharing information. The records state that resource as an item of the original they were corrected from, give it a figure in two pairs, and in the strictest of them say that a choice in one case changes the allocation of the same funds in the other. Then they refuse every other tie, in a sentence that every one of them repeats: only the stated resource grounds the link, and what the original did not confirm is not linked. The ten other cases of their programme apply the same discipline from the other side and say why they link nothing.
This is a definite position. In the terms of Malone and Crowston, The Interdisciplinary Study of Coordination (1994), the records name a dependency of one kind, the shared resource, and deny every link an order of precedence or cause. Against the picture in Cohen, March and Olsen, A Garbage Can Model of Organizational Choice (1972), they join nothing that is merely present together, and nothing that merely resembles.
The silence is about the allocator. The two records of every pair have different owners and different deadlines. Each places its own decision before an allocation of the shared resource, or asks for the allocation table as an input, or says the allocation will be approved jointly, without naming the partner. One record gives the approval of its pool to its own council, after a check of the effect on the other claims, while its partner says the approval is joint. Five records carry the shared resource nowhere beyond the link. In the language of von Hagen and Harden, Budget Processes and Commitment to Fiscal Discipline (1995), the records write down the particularistic view of each claimant with care, and do not write down the comprehensive view their model asks for. In the language of Scharpf, Games Real Actors Could Play (1994), this study reads the closest the set comes to coordination is negative: check the effect on the other claims, do not set the amounts arbitrarily. In the language of Kettl, Managing Boundaries in American Administration (2006), the records draw the boundary of resources across cases and keep the boundary of responsibility within each.
Two further readings are this study's own. First, the refusal of any link that the original did not confirm can be read beside Pollitt, Joined-up Government: A Survey (2003), and Christensen and Lægreid, The Whole-of-Government Approach to Public Sector Reform (2007): joining decisions up has costs and risks, and a record set that joins only what it can ground keeps those costs small at the price of leaving most of the joining undone. Second, the misfit between the link's own content and the template around it shows how a studio's frame can outlive the use it was written for. The template, written for likenesses of reasoning, says that a comparison between cases with different owners and deadlines does not hold; a shared resource ties exactly such cases. A reader who took the frame at its word would discard the one kind of link the twelve exist to record.
Contributions
The paper adds three things to the debate on coordinating interdependent decisions. It describes, with counts from the whole population of 6,540 AEGIS synthetic cases, a rare form of link between decision records: grounded in one named shared resource, symmetric, confined to a closed set, and joined to an explicit refusal of every other tie. It separates the link's own content from the studio's frame and shows that the two were written for different kinds of tie. And it finds that the records name what decisions share without naming who decides it, the point the literatures on common pools, negotiated coordination and joined-up government would each put at the centre.
Limits
The cases are synthetic. AEGIS marks every one of them realistic synthetic, names every deciding body fictional, and says that every figure is the original's or a training assumption, so nothing here is evidence about any real allocation, institution or alliance. The set is small, twelve records in one programme, and it was selected by a single field that only this programme fills; the findings describe that programme's way of writing links, not the studio's in general, except where a count from the whole population is given. The links are symmetric in every pair, which makes the pairs easy to read and says nothing about how asymmetric dependencies would be written. The Korean records were read in this study's English, and the classifications of wording are rules on the Korean words that may miss paraphrases. The frame measure is a floor on Korean text. The readings through the works named are this study's interpretation of short notes on each work, not of the works in full, and the records were not compared with the original text from which the programme was corrected, which this study did not read.
Further research
Three lines of work follow. The first is to read the 212 links elsewhere in AEGIS whose reasons say two cases share an allocation, budget, hours or staff, and to ask whether any of them names who decides the shared resource; the comparison would show whether the silence found here belongs to one programme or to the studio. The second is to read the ten cases of the same programme that decline to link, and to set their stated reasons beside the twelve that do, since the two halves were written under the same rule. The third is to ask how a record set could write the allocator in without inventing one: for instance, by a field that names the body that holds the shared resource, or by a statement that no such body has been named. Any of these would be a study of how decision records are written, and like this one it would leave the decisions themselves to the people who must make them.