The eleven records read here count their seats exactly and leave their clock and their queue open. The executed calculation answers the question most of them ask, how many cohorts, and every recommendation rests on that answer. Its only measure of time stops at the end of the first cohort. Who goes first is the most frequent unknown and is never recommended. And where the records state a ratio, a threshold or a count of cohorts in their own words, those words sometimes part from the figures the run holds, without the record setting the two side by side. All eleven are AEGIS's authored synthetic cases, and these are findings about how the record set frames a decision, not about any training anywhere.
Discussion
The debate on implementation asks whether the carrying out of a decision is better seen from the decision or from the point where it is delivered. These records are written from the decision. They fix the number due, the seats and two periods, and they let an executed rule turn those into a count; the count is exact, it is replayed, and nothing in the records disputes it. Read through Richard E. Matland, Synthesizing the Implementation Literature: The Ambiguity-Conflict Model of Policy Implementation (1995), this is, on this study's reading, the register of administrative implementation, the case of low conflict and low ambiguity. What this study adds to that reading is that the register, in this record set, holds only the part of the decision that the seats determine. The rest is left as needed inputs and unknowns, and they are the facts that a view from the bottom, in the sense of Benny Hjern and David O. Porter, Implementation Structures: A New Unit of Administrative Analysis (1981), would look for in the other parts of the cluster that carries a programme out: the roster of people and their priority, the room actually available, the equipment schedule, the instructors.
The second strand, on discretion and rationing, sharpens one of those gaps. A cohort with fixed seats is a ration in time, and the records say plainly that the rule for who is rationed first is unknown to them. Peter Hupe and Michael Hill, Street-Level Bureaucracy and Public Accountability (2007), describe a policy system as a nested sequence of decisions, and Lars L. G. Tummers, Victor Bekkers, Evelien Vink and Michael Musheno, Coping During Public Service Delivery: A Conceptualization and Systematic Review of the Literature (2015), list rationing among the ways frontline workers cope with workload. This study reads the records as taking the upper decision in such a sequence and leaving the lower one, the order of the queue, where the record cannot see it. Richard C. Larson, OR Forum—Perspectives on Queues: Social Justice and the Psychology of Queueing (1987), suggests why that lower decision is not a detail: a violation of first in, first out is among the factors that may shape attitudes to a queue more than the wait itself. The two alternatives that set an order say as much in their own reasons, which are about those who would wait longer.
The third point concerns time. John D. C. Little, A Proof for the Queuing Formula: L = λW (1961), ties a number waiting to a rate and a time. The records fix the number and the seats; the rate depends on whether cohorts overlap, and the records do not say. Paul A. Sabatier, Top-Down and Bottom-Up Approaches to Implementation Research: A Critical Analysis and Suggested Synthesis (1986), argues that the time-frame used in most implementation research misses many critical features of policy-making. Here the period is set by the rule's third output, and it ends with the first cohort. The two records that look further are the two whose figures make the whole schedule fit exactly. A. F. Forbes, R. W. Morgan and J. A. Rowntree, Manpower Planning Models in Use in the Civil Service Department (1975), describe models of the supply of people as a family of tools, each suited to its problems. The rule in these records is suited to one problem, splitting a list by seats, and the records sometimes ask it a different one, about pace and balance, which it does not answer. Laurence J. O'Toole Jr., Research on Policy Implementation: Assessment and Prospects (2000), argues that questions close to implementation continue in unusual places; a set of authored decision records with executed calculations is one.
Contributions
The paper makes three contributions. First, it gives a close reading of one designed record set in which an exact, replayed calculation and the decision it supports can be compared line by line, and it shows that the calculation settles the count of cohorts and that every recommendation rests on it. Second, it locates the edge of the calculation: the clock that stops at the first cohort, and the queue whose order is named as unknown. It shows that the records speak of the whole schedule where their figures fit and elsewhere speak of balance and of no gap in words the run neither supports nor contradicts. Third, it brings the debate between top-down and bottom-up views of implementation, and the work on rationing and on queues, to bear on decision records that hold a small executed calculation, and states what such a record carries about time and order.
Limits
The cases are AEGIS's authored synthetic settings. Their posts, seats and months are the author's assumptions, and nothing here describes any real training, branch or organisation. The study reads eleven records, all in one programme and all running one rule; the 84 runs of the rule in other programmes were not read closely, and what holds here may not hold there. The figure for the end of the last cohort is this study's arithmetic on one reading, that cohorts run one after another, which one record states and the others neither state nor deny. The records are Korean, and the English is this study's translation; "double" renders 배, which standing alone means twofold, and "the former posts" renders one ambiguous word (전자) in a single record; the frame measure was taken on the Korean, where it is a floor. The readings through the works rest on their published abstracts, not on the whole texts, and are this study's interpretation.
Further research
Three lines of work would test these findings. The first would read the 84 runs of the same rule in AEGIS's domain programmes and ask whether records outside the calculation programme state the end of the last cohort, set an order, or keep their words and figures together. The second would read the other calculation rules of the same programme, which count backlogs, capacities and remaining windows, and ask whether their clocks also stop at the first step. The third would compare such authored records with published records of real training and transition decisions, where those exist in the public domain, to ask whether the edge drawn here, an exact count with an open clock and an open queue, is a feature of this record form or of the decisions themselves.