A theory of presentation and its implications for the design of online technical documentation
©1997 Detlev Fischer, Coventry University, VIDe (Visual and Information Design) centre
Confluence is the interaction of several concurrent material or articulated resources in presentation. Confluence is spatially delimited by the available display, and temporally delimited by temporal tolerances that determine what navigated units of display can at any time become related to the problem through things remembered, articulated and projected.
The ubiquity of confluence is indicated by the fact that meaning is often not in the resource, but between resources [1]. Engineers discuss details while comparing dimensions on two schematics. While talking to the customer on the phone, they take notes, generate to-do lists, or compare and check the customer's oral account against a written report. They phone to announce faxes which they then orally comment and qualify. They generate formal and informal messages in parallel, each tailored for different addressees or the same addressee in different functions.
One problem is to define the substrate confluence. While its social and interactive nature suggests that confluence cannot be treated as a psychological category, it is equally not a shared good in the sense of intersubjectivity [2] or consensual domain [3] because each participant in confluent presentation may focus on different aspects of the display and harbour different notions as to its semantic implications.
A solution is to conceptualise confluence as a constant recursive negotiation of displayed evidence in the light of the problem. It is individual in that users always have different foci on the problem, and social in that these foci constantly interact with each other and with the independently paced referent domain.
It is useful, then, to turn to the display as the most tangible substrate of confluence. However, confluence is not simply the sum of singular resource availabilities since these interfere and mutually affect each other. Two computer applications, each available on its own, may not run at the same time. A large schematic buries other documents. A controversial argument hinders the thorough study of problem details.
There are other interactions between resources in presentation which must be distinguished from confluence. These are interactions between different presentations (multitasking); between presentation and decision; and between presentation and action.
Multitasking describes the concurrent availability of resources which belong to different and independent presentations. It describes a situation of switching between repertoires of resources associated with different problems which users prioritise, interleave, or try to present at the same time [4]. In multitasking, problems are not confluent but parallel, and can be mutually obstructive if a second task interferes before the completion of the first task [5].
Decision leaves the scope of presentation, delegating presentation or action to other settings. Its outcomes may often re-enter presentation at a later time in the form of resources articulated down the line.
The interaction between presentation and action is an oscillation between two processes, for example, the installation of a component and the reading of the documentation scheduling its installation. Both processes are confluent insofar as the equipment is itself a resource for presentation. For example, the width of a bolt directly specifies the width of the spanner needed to tighten it (but not the right spanner load). Interaction between action and presentation can validate resources, e.g., an error in a FIM troubleshooting tree is only detected in a situation of use [6].
Oscillation may include intermediate resources linking root and derivative resource, e.g., part number etchings or labels on parts to be installed that are matched by labels in documents, or tools that may look as if they were part of the equipment and are therefore labelled ‘tool must be removed after use’ (RB211 Troubleshooting 1980).
Confluence suggests an extension to the concept of resonance introduced in chapter 4–Problem. The confluence of resonating resources may be consonant or dissonant. Confluence is consonant when both resonances complement each other so that they indicate the same pattern (and thereby seem to support the same interpretation). They suggest a validity which is only local and may be disproved by other resources. Confluence is dissonant when individual resonances produce an interference pattern indicating discepancies that call all confluent units of resource into question.
The confluent display of resources produces local validation through consonance or dissonance. This works within and between resources. Two yes-arcs leaving a box in a FIM (Fault Isolation Manual) decision tree clearly indicate an error by dissonating to the protocol prescribing just one yes-arc and just one no-arc. Contradicting illustrations of the same mechanism indicate that at least some must be wrong.
Ted pointed out a problem of finding out how the oil should be shown travelling through the matrix. According to Raja it goes through vertical slots, a Lucas drawing however mentions oil pipes—two arrows point at them. A drawing of the Japanese manufacturer shows the matrix as a few thin vertical lines without giving any detail. An oil system ETG for the V2500 shows a very similar cooler type with a detail section drawn in 3D next to it. [FN 28/9/94]
Dissonance between articulated hypotheses and navigated resource may also affect the validation context beyond presentation. It can undermine the status of ‘facts’ and trigger an investigation of the protocol of measurement that produced these facts. For example, in investigating ‘High Oil Consumption’, dissonance between different accounts may lead service engineers to establish if a topping-up procedure filled the oil tank to the bottom or the top of the tank neck—a difference of 4 US quarts crucial for any valid hypothesis accounting for the observed oil loss.
Dissonant resources have particular salience: they routinely become occasions for repair and receive extra elaboration. An example is a Rolls Royce training course instructor asking all trainees to change a detail in a diagram: the supplied cooling air for a bearing has changed from HP3 to HP8. This sparks an confluent discussion about the reasons for this change, in which the level of instruction moves from descriptive to analytical. The result that the dissonant detail has rendered its referent more memorable.
The described effects of display confluence must be seen against the background of pattern resonance and the projection it affords. Such projection may be strong enough to hide local dissonance or to type individual resonances to projected problem dimensions.
An example from the cinegram evaluations is the projection of causality onto two display events which are co-variant, i.e., caused by the same preceding conditions, but not causally related [7].
Figure 7.1. The figure shows a small section of a cinegram animation segment. An oil pressure rise triggers a signal in the ÆP (differential pressure) switch, symbolically shown through a spark. Shortly afterwards, the further rise in oil pressure forces open a relief valve. Although both events are merely co-variant, their confluence led several users to construct a causal relationship between them.
A rise in oil pressure triggers the first event, a delta-P switch signal, symbolically cued through a spark (cf. figure 7.1). Further rise in oil pressure triggers the second co-variant but independent event, the opening of a pressure relief valve. Both events are confluent through spatial and temporal proximity: they happen in the same assembly view and follow each other immediately.
This confluence of events induced several users to project a causal connection into the display. They assumed that the electrical signal triggered a mechanism opening the valve. This projection is explained with generic resonance, since it occurred only with engineering students who had extensive generic engineering knowledge of systems with electrically-operated valves [8], but no knowledge of the specific domain. Generic resonance was strong enough to hide a local dissonance: the display shows no electrical connection to the pressure relief valve which would allow it to be electrically operated.
Confluence often literally depends on managing the availability of a combination of resources at one time and place: being able to bring together equipment (or derivative evidence), documents, and experienced users. Combination implies that resource utility cannot be established on the level of the individual resource because any singular utility hinges on the availability of all resources needed in a presentation.
Examples of resource combination are Rolls Royce field reps stationed at the airline's home base, or systems such as Lufthansa's BISAM system that afford remote joint viewing of damaged equipment, or the combination of users in meetings where each user may in turn bring in documents.
For the confluence of documents, portability is crucial. Service engineering once considered changing the container type for engine event filing and subject filing from small and flexible manila folders to voluminous lever arch files. In spite of the advantages of lever arch files (such as increased volume and view permanence—the manila folder must be held open) the change was not implemented. Users routinely take documents to meetings which may be quite remote, and carrying more than a couple of wedge-shaped, rigid, and slippery lever arch files is no pleasure.
Confluent gestures such as pointing or hand gesturing qualify presented resources. Qualification cues the resource or presents further dimensions. Gestures cue display tokens, or make such tokens the reference point for articulations which present further dimensions, such as change over time [9].
Confluent presentation of change is grounded in the display and simultaneously articulated verbally. In hand modelling, both hands act as a tool that dynamically presents relevant features of a part, component, or process. Hand modelling often starts with display tokens such as diagrammatic illustrations of components; led by the momentum of articulation, users then seem to pull off such tokens from the display into the air where they manipulate the hand model. While listening, other participants watch the hand model and occasionally glance at the referent token in the display; the presenter may also respond to any signs of users' hesitation by strengthening the reference relationship between hand model and referent token through explicitly pointing out features on the display which may surpass the granularity or dimensions of the hand model. Tools may also enter such confluent presentation of change: in discussing a bearing chamber cross section, an engineer may indicate the changing oil level with a horizontally held ball pen.
A necessary condition for confluence is the availability of resources within temporal tolerances which depend on domain patterns and presentation weight. For example, meetings may impose an absolute deadline for resource availability if they have a fixed slot (such as 10.00 the first Monday each month), are urgent and important, or involve too many people to be re-scheduled. In other cases tolerances may be softer and easier to re-negotiate. This study was supposed to be finished in September 95 and was in fact finished one year later.
Temporal tolerances also appear as lead times and expiry dates. Lead times force the projection of needs that trigger the advance scheduling of auxiliary presentations so that they are likely to close out before the time of the confluent event. For example, lead times of non-routine maintenance activities are cut through advance preparations such as getting the needed documentation, parts and tools, and expert opinion. Lead times must also be managed in the design and implementation of engine modifications. To afford confluence of all resources needed at the moment of decision—the clearing of the scheme in the official scheme review meeting—much work such as detail design and ordering of raw material is informally commissioned and carried out long before it is officially sanctioned: it would simply take too long to go through the process in linear fashion according to protocol. One implication of advance preparation to manage temporal tolerances is that often presentation will operate with virtual resources, or pointers to future resources.
Temporal tolerances are built into formal documetns as expiry dates such as ‘Do not use after…’-stamps on print-offs from schematics or maintenance manuals. Maintenance work cards can only be used once since a revision might be introduced the next day. Outdated resources are used nevertheless, but usually they are modified through confluent remembering or corrective annotation of changes.
Externally paced processes in the referent domain impose situated conditions for confluent presentation which shape the temporal tolerances for resource use and thereby, resource utility. For example, the utility of the DRUID system cannot be evaluated based on the merits of its singular availability, but must acknowledge actual and possible conditions for confluence. Giving service engineers mobile phones or putting DRUID on every desk would boost its utility because this would afford confluence with the telephone. Also, the navigation distance is dependent on more than the system's architecture: it depends on states such as starting conditions and load of the infrastructure. For example, it increases when users have to remove the CD the previous user has left in the drive, or when network traffic slows down transmission times.
Presentation is dependent on maintaining relevance for externally paced or emergent referent processes. Temporal tolerances mean that the pace forced onto presentation by the referent domain decides if likely navigation gains are larger than delay costs through navigation distance [10].
When the domain forces tightening of temporal tolerances, presentation suffers. An instructive example is the failure of the Airbus A320 G-KMAM at Gatwick on 26 August 1993 which was to a large extent attributable to maintenance work impaired by an unrealistic deadline which meant that crucial resources—support collars, flags, appropriate documentation—were unavailable and caused the invisibility of the faulty equipment state which led to the incident [11].
Design defines a discreteness of views that confluent presentation defies. Bird's-eye views and classifications present the system as observer-independent architecture. But this architecture always appears in crossing view boundaries in presentation. The abrupt difference of switching reveals the boundary, but confluence of before and after in a temporally extended focus simultaneously qualifies the prior and the future state. This can appear as loss of situated meaning or as transparent meaningful transition [12] depending on the users' projection, anticipatory and retrospective, between the view they have just left and the newly displayed target view. For example, noticing the target view may correct projection and at the same time retrospectively re-interpret the already hidden pointer token.
Confluence bridges the navigation gap between mutually exclusive views set apart by container type characteristics. For example, it is not possible to display two pages filed at different places within a manila folder without removing (and having to replace) the pages; or to display two cinegram nodes at the same time; or, in the virtual screen display mode used by Lufthansa's BISAM system, to display at the same time windows belonging to different co-operating applications.
Confluent articulation can bridge such navigation gaps by extending the focus through remembering afforded by the problem pattern, and cueing afforded by auxiliary tools such as hands or bookmarks. Remembering is imagined or uttered articulation which extends the focus to a near-confluent presentation of discrete displays. Cueing extends the focus through pointers such as bookmarks or trail functions—or simply fingers stuck into different places in a document—which link the cueing pointer to the view and afford quick switching, for example, when users toggle between two pages (cf. figure 7.2).
Figure 7.2. A schematic view of confluence bridging navigation gaps when displaying discrete printed or electronic pages.
Confluence improves the decision base in complex presentation, for example, in situations where no appropriate resource exists. Best-fit resources are then transformed into articulations scaffolded by the confluence of other relevant resources. For example, when a component removal procedure only exists for the overhaul domain but may be needed in the line service domain, a service engineer may study the overhaul removal procedure and then walk down to the shop floor to imagine the needed transformation based on the visibility of the equipment and confluent assessments of people with hands-on experience.
The situated validation context of such a complex presentation is equally complex. It will include time estimates for the removal task, estimates concerning available skills at the target setting, and protocols regarding instructability, since the protocols of expert activity must be transformed into protocols of instructed action [13].
Confluence increases visibility of more variables of the domain which often implies a higher span of control or influence on the operational level. For example, the fact that Service Engineering is at the link between the external operator and many internal sections of the organisation such as Design, Repair, Materiel, Procurement, and Production & Testing, affords confluence of all these perspectives which leads to considerable insight into the genesis and causes of problems across these divisions.
The operational level of control is mediated with the decisional level of management. The more operational processes are understood and indeed controlled by employees, the more threadbare managerial prerogatives must appear, and the more likely that operational control turns into a powerful instrument in industrial disputes [14].
[1] Schamber (1991 p39) has criticised that ‘the majority of studies in information science have not considered…multiple-source environments; rather, they have focused on one type of source or system at a time, such as interpersonal resources…or automated retrieval systems.’
[2] Luhmann's (1992 p19) constructivist critique of the concept of intersubjectivity ignores the informing intrusion of a grounded and externally paced referent domain into ‘the recursive network of communication’ (op cit p20). Even if one is willing to accept the notion of information as differentiation of possibilities for recursive structural coupling in communication (op cit 321), this would simply concern the substrate of the operation of reference, not its form. The form of reference recurs on external—i.e., grounded—changes of the referent problem.
[3] Cf. Winograd & Flores (1986 pp48).
[4] A diagrammatic view of multitasking and toggling between presentations can be found in Cypher (1986 p262).
[5] The phenomenon is known as ‘refractory period’ in psychological research (cf. Smyth et al 1994 p132). There is evidence that interference lessens if simultaneous tasks involve different response systems, e.g., haptic movement and oral articulation. There is also evidence that reduced transformation effort between available display and required action reduces the refractory period: ‘it is easier…to move a response key to the right if an arrow on the computer screen points to the right than it will if the word elephant appears and you have to remember a rule which says "elephant means to the right, giraffe means to the left"’ (op cit p135).
[6] ‘During the troubleshooting of a recent problem in Singapore, a typing error has been noticed in the FIM procedure for ATA 79-00-00’ (Excerpt from a request of a Rolls Royce field representative in Sydney)
[7] Norman (1988 p11) has discussed this false causality problem in some detail.
[8] This is the appropriate extract from the evaluator's report: ‘He revealed his thinking when he told me that "the bang, the flash on the controller [delta P switch] shows a signal to the Electric controller and that opens up the bypass valve"’ [LF 3.2 P37].
[9] In her concept of mental animation, Hegarty (1988, 1995) assumes an articulation-independent mental process that adds movement to static displays. In the field, however, mental animation is frequently presented through displays, accounts and hand modelling.
[10] Cf. Lamb (1996) who quotes Linder's argument that decisions are usually made on the basis of incomplete information, because getting informed consumes time needed for production.
[11] For details of this very instructive case study, cf. Mellor (1995).
[12] ‘We have observed that users are not at all surprised by the fact that they can directly access the problem source document from the report document. Not only are they comfortable with the fact that the associated information appears from this new location, they expect it. They seem to think in terms of this structure as a unit. The information "belongs" together. Users do not seem to think about navigating or moving somewhere when using this structure. Their focus is on viewing the information they need’ (de Young 1990 p245).
[13] PTF (production and test facilities) can in certain situations bend pipes to remove a component —‘they know what they are doing’—but ‘you can't instruct that in a manual’.
[14] Cf. Beirne (1990) who discusses such political issues in his analysis of user participation in computer redesign, and Lindsay (1994 chapter 8).