22.1 The original protocol
“If it sounds restrictive, that is because it must make clear how it needs to work. The Rules of the Game do not inhibit players once they agree to go on to the field” (BD 1994: 21). Beer wrote this in advance of an objection he expected: that a detailed protocol looks like the opposite of freedom. The icosahedron of chapter 21 had placed thirty people on its edges and twelve topics on its vertices, but a structure alone does not make a meeting. Beer’s second requirement, from the “perfect democracy” passage, was “a procedural protocol” that is “non-hieratic” (BD 1994: 12; see chapter 20), and BD sets one out in detail. The first version, from the 1990 experiments, has three parts: “There are three parts to the Protocol: the Problem Jostle, which is the initial phase, the Topic Auction, which allocates roles, and the syntegrity iterations, called the Outcome Resolve, which follow” (BD 1994: 21). In the Problem Jostle the infoset produces its own agenda, the idea whose origins were traced in chapter 20. The Topic Auction distributes the thirty participants over the twelve topics and the critic roles. The Outcome Resolve is the repeated round of team meetings in which the topics are worked out. Beer states the scale of the undertaking twice. The minimum programme “uses four periods of four hours each, and takes three syntegrity iterations to reach its conclusions, with all players in one location” (BD 1994: 21). The full form is much larger: a “full-scale icosahedral team syntegrity is a big undertaking: it is clear […] that it should last for 5 days, and ideally it needs four facilitators and four support staff” (BD 1994: 88). His comment: “That adds up to a considerable investment: the benefits are, however, overwhelming” (ibid.). The first figure is a minimum for a condensed meeting; the second is what Beer, after the experiments, regards as the proper scale.
22.2 The revised protocol: episodes
Records, participants’ responses and facilitators’ advice led Beer to revisit the protocol in Part Two of BD (“Enhancing Procedures”, ch. 7). There he introduces a new unit: “the eight episodes of a ‘meeting under syntegrity,’ or (as it is now called) a SYNTEGRATION” (BD 1994: 111). Named in the text are:
- Importance Filter: collecting Statements of Importance (SIs), in which participants “let their imaginations soar” (BD 1994: 111). Beer explains the name: “The Importance Filter is the device Syntegration uses to capture raw data of high quality. Thus it is intended to filter out only the low-quality material put forward, and tentatively to begin the process of shaping a variety so high as to be inchoate” (BD 1994: 113).
- Problem Jostle: grouping and polishing the SIs into Aggregated Statements of Importance (ASIs), “the output of the Problem Jostle” (BD 1994: 115).
- Hexadic Reduction: reducing the ASIs to six polar axes, i.e. twelve topics, adapted from George Kelly’s theory of personal constructs (BD 1994: 115–116).
- Topic Auction: allocating the 30 participants to the teams (BD 1994: ch. 8, from p. 137). After the Topic Auction come the Outcome Resolve iterations (BD 1994: 21, 32). Beer does not give episodes five to eight headings of the “Episode One” kind, but he states their content: “Nothing more needs to be said at this stage about the fifth to eighth episodes of a Syntegration: these consist of the three episodes that between them constitute the Outcome Resolve, and the Conclusion session” (BD 1994: 144). Episodes five to seven are therefore the three iterations of the Outcome Resolve, and episode eight is the Conclusion session. The same numbering appears in the timetable, which lists “Outcome Resolve, Episodes One and Two” and “Outcome Resolve Episode Three” (BD 1994: 146). The experiments had shown “two major flaws”, Beer explains. The rules used in the Topic Auction “did not result in an optimal allocation of tasks”, and “the origins of the difficulty lay within the Problem Jostle itself—which was supposed to discover the Consolidated Statements of Importance (CSIs) in one fell swoop. That is to say, exactly 12 items had to emerge from the process” (BD 1994: 113). The revision splits that one step into two. First the Problem Jostle produces an open-ended list of ASIs, whose number “is not predictable” (BD 1994: 115). Then the Hexadic Reduction condenses it to twelve topics “in locations of polar opposition—as determined by the insights of the Infoset itself” (ibid.). The six axes and twelve poles of the icosahedron now enter the procedure directly, and the group itself decides which of its concerns stand opposite each other.
22.3 Outcome Resolve
In the Outcome Resolve the structure does its work. Beer describes it first for the minimum programme: “A Total Session, in this case of 4 hours, consists of six 40 minute Meetings on the part of two teams each time” (BD 1994: 32). With two teams meeting at once, ten players and ten critics are occupied, and so “there will always be ten other players ‘on the move,’ looking at new draft CSIs which they may sign, and noting in particular the activities of their two ‘polar opposite’ teams” (ibid.). No one is idle, and everyone keeps watch on the opposite pole. The critics work under a discipline. “The team works for 30 minutes, in which time the critics must be silent. They then have 10 minutes in which to say something helpful and germane” (BD 1994: 32). Each team aims at a one-page draft Final Statement of Importance (FSI) (ibid.). In the revised timetable Beer tightens this further: 25 minutes of work, 10 minutes of critique and 5 minutes to “metabolize this critique within the draft FSI” (BD 1994: 145). The iterations, Beer insists, must be close together in time. “If the reverberative effect is crucial, then meetings ought not to be spread out in linear fashion like so many episodes of a weekly TV series wherein one cannot remember what is going on. The closure must be evident: a three-dimensional ouroboros (the snake that eats its own tail) in action” (BD 1994: 40). The revised procedure says the same: “The Episodes, as we have come to call them in this chapter, are integral wholes. If an iteration were to be broken in half by a night’s sleep, it would lose its dynamics and therefore also its syntegrity” (BD 1994: 145). The ouroboros is the procedural form of the closure argued for in chapter 20: a night’s sleep would cut the snake in half.
22.4 Facilitation
A protocol this detailed needs facilitators, and facilitators can steer. Beer warns against steering: “If facilitation is to be exerted against a self-organizing Infoset with open-ended mandate, instead of easing it towards acceptable conclusions, then the facilitators are effectively determining agenda, and the Syntegration is spurious thereafter” (BD 1994: 146). The boundary is sharp. Facilitators serve the protocol and leave the outcome alone. If they decide what the group should conclude, the process loses the property that justified it, namely that the agenda and its categories come from the group itself (chapter 20). Beer makes the same point about the Topic Auction in the original protocol. Organizers “are urged to allow the self-organizing propensities of this whole approach to determine the allocation of roles. If not, they might as well sit in their offices and appoint task forces, inter-departmental committees, and so forth, as they have always done—and with the same turgid results” (BD 1994: 32).
22.5 Reverberation as the mechanism
Why should this procedure produce better results than an ordinary conference? Reporting back to one of the experimental groups, Beer wrote: “There seemed to be a lack of appreciation of two key notions in this approach. Symmetric closure is one; reverberation round the system under repeated iterations is the other” (BD 1994: 45). His answer rests on these two properties. Part Three gives the argument a more exact form. Starting from any vertex, “After only one iteration, that information is divided between ten of the remaining vertices via members and critics, and something will immediately be known at the polar opposite vertex, because Aggregated Statements of Importance are then published” (BD 1994: 222). Since the structure is symmetric, “the information will be evenly distributed everywhere—precisely by the process that we have been calling reverberation” (ibid.). Beer calls this an ergodic process: over time, and wherever it starts, it tends to the same limiting distribution (ibid.). The matrix algebra behind this is referred to the collaborators’ Part Five (“see Part Five, Surplus One”); otherwise, Beer writes, “take the value on trust” (BD 1994: 222). Beer reports the result. From it he draws a practical rule: “the scheduling protocol for a syntegration is of vital importance: the capacity of the syntegral graph to resolve the outcome through homogeneous insight must not be compromised by shortages of time or facilitation” (BD 1994: 224). He also asks how many times the snake must go round. His calculation rests on the eigenvalue 1/√5 raised to the power of the number of iterations, which falls from 0.45 after the first iteration to 0.20 after the second (BD 1994: 223; Figure 13.1 there marks 55, 80 and 90 % for one, two and three iterations). In his words, “One episode has only 55% resolving power; two episodes takes us to 80% of the ideal, which is not too bad. With three episodes, the Outcome Resolve is 90% effective” (BD 1994: 224). He adds a reason that lies outside the mathematics: “There is no allowance within the equations for sheer empathy among participants” (ibid.). The experiments suggested “that syntegration is more efficient than the model predicts. Three iterations are enough” (ibid.). The number three thus appears twice, in the first minimum programme (BD 1994: 21) and in the later calculation, which arrives where the earlier practical choice had already stood.
22.6 Beyond one group: hyperinfosets
A single syntegration involves thirty people. The large-scale experiment planned for the World Service Authority (see below) concerned much larger numbers, and for such scales Beer proposes linking infosets. When local groups with a shared purpose recognize each other, he calls the collective “a hyperset of neighbourhood infosets, or simply a hyperinfoset”, whose architecture “is obviously based on a syntegral hypericosahedron” (BD 1994: 173). The rule is recursive: thirty infosets send one member each, and that member is again an edge in a new icosahedron. Each member brings two of the propositions produced by the primary infoset, so the hyperinfoset “begins operations, then, with 60 SIs” (BD 1994: 174). Beer calls this “a second level of recursive embedment” (BD 1994: 173), a term that connects directly with the VSM’s principle of recursion (see chapters 6 and 23). He rules out reading this as a hierarchy: “The purpose of forming hyperinfosets is not to create a hierarchy of warriors or of logical propositions that have advanced status because they have ‘risen to higher levels.’ It is to make consolidated statements available that have been worked over by 900 people, 27 000 people, and so on” (BD 1994: 174). The higher level adds breadth of participation. It does not add authority. A large-scale experiment on this basis was planned for 1993 for the World Service Authority (BD 1994: 175), including “a syntegration manual that could be used by a highly motivated organizer to run a 3-day version of the technique, without prior knowledge, and without professional facilitators” (ibid.). Beer reports that the results were then “being evaluated by the World Service Authority” (BD 1994: 176). BD does not report the outcome.
22.7 The Topic Auction
The Topic Auction, episode four of the revised protocol (22.2), gave facilitators more trouble than any other step, and BD ch. 8, “Vexed Questions of Allocation”, is given over to it. Beer starts from a failure. In the experiments people chose a first team colour and then, once that was settled, a second. This “was unsatisfactory for most people”; since an edge means membership of two teams, “the 30 options should be considered (for the majority) as colour pairs” (BD 1994: 126). The critic roles then follow from the pair. Chapter 21 (21.2) states the Freesman and Adjacency Exclusions (BD 1994: 125–129); Beer’s worked example shows them at work. The Green team’s critical options are Dark Blue, Purple, Red, Black and Orange. For the person Green–Gold, Dark Blue is barred because it is the pole opposite Gold (Freesman Exclusion). Orange and Red are barred because each already shares a member with Green–Gold’s other team, Gold, through Gold–Orange and Gold–Red (Adjacency Exclusion). So “the critical roles of Green–Gold are fully determined as Purple and Black” (BD 1994: 127). The same result follows from the Gold side (BD 1994: 128–129; Tables 8.1 and 8.2). Beer adds a piece of history. His first published rules (written in 1984, published in 1990; BD 1994: 127), with their “handedness” and the return of a critic to the other hand, already implied the Adjacency Exclusion. The insight was lost in the decade before the experiments, which produced scrambles for roles and “anomalous appointments to critical duties”. “And no-one involved perceived what was wrong!” (BD 1994: 127–128). For the auction this means: “What is on offer, 30 times over, is a discrete entity: it is a quadripartite indivisible role” (BD 1994: 129), two memberships and two critic roles bought together. A second constraint concerns who meets whom. Jo Hancock’s simulation of the meeting schedule showed that any player meets four people three times, eight twice, eight once and nine never; five of the nine can still be met in other sessions (BD 1994: 130–131). Joe Truss found that one person is “totally inaccessible”: the player “who holds the edge role that is parallel to a given role, on the other side of the icosahedron”, which Truss called the “polar edge” (BD 1994: 131). Table 8.4 lists the fifteen pairs under the heading “POLAR EDGES EXCLUSION”; for Silver–Green the polar edge is Yellow–Purple (BD 1994: 136). The remedy is informal. Each person should arrange a private meeting with the opposite polar edge, and this is explained at the auction (BD 1994: 131, 137). The auction itself uses two documents (BD 1994: 137–141). The Topic Auction Data Sheet lists, for each of the twelve topics, its five member pairs and the two critic roles each pair entails (Table 8.5, BD 1994: 138–139). Participants are reminded that the pole opposite their own topic “is the topic of their antithetic management” (BD 1994: 137). The questionnaire “Topic Auction of Thirty Quadripartite Lots” asks for up to seven colour pairs in order of preference. Under “PRICING OF BIDS” each colour is ringed from 1 to 7 “to denote degree of enthusiasm to discuss”; anyone who initiated or strongly influenced a topic rings the number twice; topics to avoid are marked M, C or MC, for member, critic or both (BD 1994: 140). Beer frames the form with an ideal. In a “perfect” Infoset “all the questionnaires would be returned blank, and the roles allocated at random” (ibid.). In practice the room is set out as a clock of twelve stations. Holders of a double-ringed 7 go first, then the facilitator works down the price list, and people may change their minds as they watch teams form. Each person allocated receives a badge with two member colours and two critic colours. Facilitators must not force the issue. If the process stalls, they allocate the remaining places from the forms, and “We are left with something akin to a preferential voting system” (BD 1994: 141). Where someone is likely to be disgruntled, a facilitator explains why the decision ought to be acceptable: “This is frankly a sales pitch” (ibid.). Beer treats allocation as an open research problem. Hancock’s algorithm doubled mean satisfaction with the auction compared with a use of Kelly’s approach, yet only up to the median of the scale. David Sutton’s spreadsheet heuristic reached a fit of 71% against 48% for Hancock’s algorithm, and Alan Pearson argues that assigning topics to vertices matters more than the auction (BD 1994: 142). Beer’s verdict on the state of the art: “Meanwhile, it is still perfectly possible to run syntegrations!” (BD 1994: 143).
References
- BD — Beer, Stafford (1994). Beyond Dispute: The Invention of Team Syntegrity. Chichester: John Wiley & Sons.