![[Diagram of resonance causing navigation and articulation]](Figures/Fig4-1.GIF)
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
The problem is the current sum of users' understanding of the display in the light of the context. Users aggregate the problem as their presentation focus resonates on the display and transparent patterns. Problem dimensions are produced in users' emergent articulation, for example, in imagination, conversation, gesturing, writing and drawing (cf. chapter 5–Articulation).
In emergent presentation, the problem arises in a particular setting through a trigger which communicates a trouble which in turn produces presentation needs on different levels of abstraction. In its course, presentation turns the trouble into a problem. It articulates unsaturated problem dimensions such as hypotheses, questions, and queries in order to render the problem manageable. Presentation reaches close-out when the problem is fixed, delegated or when it disappears.
While the problem is explicit in articulated problem dimensions, the accumulating presentationtoryGo toGo to previous chapter 3-ContextGo to previous chapter 3-ContextGo to next chapter 5-Articulation previous chapter 3-Context leaves behind a problem pattern that types and amends existing domain patterns. The substrate of the problem pattern is both in users' minds where it affords the remembering of related facts, negotiations or hypotheses, and in the material context, as problem-specific repertoire of documents. Over time, the problem pattern sinks in to merge with general domain patterns.
Problem and referent of presentation depend on the level of abstraction chosen by an observer. Within a complex overall problem such as ‘managing an ill-understood engine fault’ one can observe dependent problem dimensions down to the level of simple tasks such as looking up a telephone number. The returned number—provided it is correct—saturates the problem dimension. On a higher level, the course of presentation—the presentation history—spans from the initial symptoms to the recursive definition of the problem, to an acceptable— ‘fixing’—solution. On an even higher level, one can detect meta-problems such as ‘keeping the customer happy’ or ‘being seen to be doing something about the problem’. The problem referent is then ‘customer relationship’ or, more abstractly, ‘compliance’. The evolving presentation history may bring about changes to both problem and referent on all levels.
In learning contexts where users may be unfamiliar with the setting and the activities and resources of the substantive domain, presentation must first generate patterns of activity that articulate situational and operational protocols before the substantive problem can emerge (cf. section 3.2 Protocols in chapter 3–Context).
The emerging presentation may be more or less adequate for reaching presentation close-out in a particular defined context. Its adequacy can be assessed according to specific (sometimes competing) validation contexts (cf. section 3.3 Validation context in chapter 3–Context).
The focus [1] is the ‘moving now’ of presentation, i.e., the transiently noticed or articulated part of the display, which includes visible, audible and tactile sensory spaces. That the focus is transient simply means that it cannot be conceived of as a unit or object. It is the locus of aggregation and fading through acts of navigation and articulation.
Focus implies a periphery of not realised resources which may await aggregation through directed search or resonance. The temporal periphery includes remembered and anticipated instances of presentation while the spatial periphery encompasses all available resources in the display.
The focus continually moves on during presentation. Its extension is related to the substrate of the display and determines the temporal tolerances for confluent presentation (cf. chapter 7–Confluence). For example, the substrate of conversation is the shared acoustic periphery, which, for some seconds, makes utterances potentially addressable for anyone in earshot. The substrate of the visible display is the material resource which is more permanent, but not shared in the same sense and only addressable through concurrent articulation.
Resonance mediates the mutual and recursive constitution of focus, problem pattern and display (see figure 4.1). The current focus resonates on similar patterns in the display and on problem and domain patterns [2]. Display resonance activates pattern resonance, which in turn triggers navigation and articulation which advance the focus and change the display. Interferences between display and pattern resonance cause the surfacing [3] of particular problem dimensions which attract the focus of presentation. For example, resonance to a particular displayed fault symptom and similar domain patterns highlights problem dimensions that presentation aims to saturate. A symptom ‘rings a bell’ and induces the engineer to remember and articulate, or navigate to, resources related to prior similar cases. Certain symptoms traditionally associated with hazardous situations may even ‘set alarm bells ringing’.
As presentation aggregates the problem, it differentiates the problem pattern, which in turn slowly changes the domain pattern. The substrate of the problem pattern is both the user's mind [4] and the changing pool of resources. It is both ‘on top of the agenda’ in the users' activities, and expressed in the pattern of documents covering users' desks.
Figure 4.1. The figure shows the mutual constitution of focus, display and problem pattern in resonance.
The advantage of the resonance model over extant models of perception and cognition [5] is greater simplicity and parsimony and the elimination of conceptual fragmentation and the resulting artefacts necessitated by the modular models which permeate cognitive science. In traditional theories, problem and context are categorically separated, for example, as ‘mental model’[6], and ‘memory’, which is seen as a store of representations. In the resonance model, the problem is not treated as an internal representation or model, but as emergent and mutually constitutive relation between the focus on particular problem dimensions and distributed problem and domain patterns.
One can distinguish between local resonance, generic resonance and domain resonance.
Local resonance produces problem dimensions through surface matches. The resonating body is the display and transient ad-hoc articulations interpreting it. Local resonance is typical for novice users without generic or specific domain knowledge which means that resonance on the domain pattern is limited. Instead, resonance concentrates on the display where users look out for local matches between documents (for example, a written question and an identical or similar written answer).
Local resonance was characteristic for the technical communication students in stage 2 of the cinegram evaluation. An example is the reaction to a question such as ‘What is the purpose of the oil system?’ This question clearly matches the beginning of the system description on the opening screen of the cinegram (v.2): ‘The purpose of the oil system is to…’. Users simply copy the answer as soon as they discover the match. The lack of domain knowledge prevents projection (see below) and typical errors related to it. Although answers are often literally correct, they are conceptually ill-understood. General problems which require understanding beyond simple matching cause more problems than well defined questions about detail. Sometimes, there is a weak transfer from contexts which are structurally similar; the oil system overview diagram, for example, was often interpreted in terms of types of flow systems such as electrical wiring diagrams.
The problem produced in local resonance is shallow and preoccupied with situational and operational worries. Uncertainty about the conceptual context of the local match shows in a high degree of confluent articulation (cf. chapter 7–Confluence). Users constantly negotiate the endemic uncertainty they experience in defining the validation context (‘How much do we need to copy?’ ‘Is a more complete answer expected?’) For lack of an external problem referent, the validation context must be defined locally: a question in a question catalogue, for example, is answered and ‘done with’ when a more or less fitting text is inserted in the allotted answer space. Local resonance also explains users' strong preference for text, since textual descriptions correspond to the medium of the question.
Generic resonance produces a problem structured in terms of users' pre-existing generic knowledge which is triggered and illustrated by the documented instance. The focus consists of articulations of hypotheses and explanations which are largely based on the remembering of resonant generic knowledge (cf. chapter 5–Articulation). The display is only used incidentally, or as springboard for elaboration. Generic resonance often causes a strong projection from the remembered generic patterns which ‘fill in’ the context around individual resonating tokens in the display (cf. sections 4.3 Tokens and granularity and 4.6 Projection in this chapter).
Generic resonance dominated in evaluations with engineering students who had appropriate generic engineering knowledge, but no practical experience in the field. Examples of generic resonance demonstrate how appropriate features of the question are hijacked and fed into the problem pattern activated through generic resonance. In response to the question ‘Why are there two filters?’, engineering students produce hypotheses without any prior investigation of the display: ‘It could be because of failure, if one clogs up’; or ‘it might be for added safety’. Generic resonance leads another user to a problem drift away from the original question: ‘…there would be filters before each gearbox because these are most susceptible to damage from debris.’
Domain resonance [7] on a well known and highly differentiated domain pattern activates problems in a different manner. Domain resonance produces either instances or types of problems depending on the specificity of trigger. It may produce the problem instance, as in the case where the mere mention of a technical term that had featured prominently in one engine trouble produced the descriptor (here, the serial number) of the engine concerned: ‘Oh, are you looking at the 13089?’ [FN 12/6/95 P6]. Or it may produce the problem type, as when an given indication of High oil pressure produces a repertoire of typical troubleshooting questions. A common feature of domain resonance is projection from the domain pattern onto the displayed resource.
Resonance turns units of resource into tokens for presentation. A token is the smallest potentially meaningful aggregate or ‘grain’ of presentation on the level of abstraction relevant for a particular presentation. For example, for a user searching for a file in a drawer, this file becomes the relevant target token of navigation. The ‘grains’ of the relevant granular-ity are files - not, for example, documents, pages, or words, or—to move up the scale—drawers, cabinets, or rooms. Once the file is found, the granularity becomes one of documents within the file, and once a document is found, the relevant granularity moves further down to individual pages, paragraphs, or illustrations.
The concept of token as ‘meaningful grain’ has a lower boundary at the level of sentence or part of a drawing, since below this level, it becomes difficult to conceive of tokens as the locus of contextually defined meaning. The sense of a word is critically dependent on its lexical co-text [8], while its meaning is dependent on the sentence level [9]. Similarly, tokenness of a graphic entity requires differentiation within a graphical context or gestalt. An oval may mean an O-ring, but only because it is placed between, say, a chip detector and its housing.
As a unit is invested with problem-specific meaning, it becomes a token
and thereby a dependent part of a problem aggregate. To a lay person, a
unit in a technical flow diagram like 
may be noticed as a describable graphical entity, while being perfectly meaningless. In a next step, presentation tokenises the unit by incorporating it as part of some tentative aggregate. If, for ex-ample, both
and ![[symbol of closed valve]](Figures/clValve.GIF){kind=small}
occur in the same diagram, they may be aggregated as the same type of thing
in different states, or as two things belonging to the same class; when other tokens advance the hypothesis that the aggregate is a flow system,
and
may eventually be identified as valves.
It is important to note that the concept of token does not imply atomistic properties [10]; the only requirement is that local resonance produces it as a distinct element, and begins to relate it to other tokens. In generic resonance, the token may be recognised as one of a particular kind, at which point one can speak of an icon, label, schematic drawing, heading, or paragraph without committing to any particular interpretation of content. Finally, in domain resonance, the token is recognised as this particular element with a well defined meaning.
Resonance activates patterns in the display or mind similar to the current focus. These patterns are layered and nested and produce tokens on different levels of abstraction. Resonance selects the relevant display granularity. It may ‘latch on’ to a particular overall gestalt (such as the shape of an engine cross-section), or to a particular unit. An unfamiliar gestalt might contain singular resonating tokens, while a resonating gestalt might contain unfamiliar tokens (cf. section 6.3 Noticing in chapter 6–Navigation).
The aggregation of the problem is reflected in the trail of resources recording the presentation history. On the case level, these resources are engine history databases and engine event files, on the level of one situated presentation, resources are the trail recording the on-line interaction history, the minutes of a meeting, or notes recording a conversation. The availability of this history means that the problem pattern can be ‘refreshed’ in presentation to resist the gravity of the domain pattern.
The trigger [11] is the link between presentation context and problem. It is the message or request as it enters a particular setting. The beginning of presentation is therefore bound to the particular level of abstraction of the setting, and can be identified differently when this level is scaled up and down. A presentation in the setting of service engineering department, for example, is defined by the setting's boundaries and the points of exchange. But it can also be treated as a dependent moment in airlines' maintenance process which contracts out particular emergent tasks [12].
On the level of any particular conceptualised presentation, an emergent trigger originates outside the presentation boundary, as input or irritation. Whether a phone call or a delivered fax, the trigger replaces, interrupts, or irritates [13] a user's prior activity. Depending on the chosen boundary of description, a trigger may be a message from another organisation consuming a service agreement; a request from others within the same organisation; a question from a colleague across the desk; or the next topic in training session.
The emergent trigger can be traced back to some trouble. Presentation will then move from effect to cause in order to unfold and understand the conditions that caused the trouble. A trouble may appear as both cause and effect of other troubles (depending on the temporal anchor of observation which conceptualises it as event) [14]. A trouble may show directly, as when a fault prevents engine start-up or a fault indication is displayed at the flight deck; or it may be revealed through maintenance activities like checks or test procedures. A value may be out of spec, or oil stains indicate a leak, or the use of a troubleshooting chart reveals a typing error, or the check of a training course module reveals that a step has been missed out.
The intruding trigger calls for the user's attention and is the resource around which the problem begins to aggregate. In emergent contexts, the trigger presents the trouble as effect of a cause which presentation aims to understand. In scheduled contexts, the trigger is the next scheduled topic, which must be tokenised and contextualised in users' response presentation.
If the trigger is a message, its author implies certain intentions [15], usually formulated as a request or order. The trigger predicts presentation to various degrees. At one end of the scale, the predictive trigger initiates a well-defined activity; at the other end, the open trigger initiates an emergent presentation the result of which is still unknown. The predictive trigger contains its close-out protocols like a mould (such as a request for a certain part number which closes out when the number is retrieved and transmitted); the open trigger provides a few patchy and ill-structured indications of trouble, which means that the mould—the close-out protocols for an acceptable response—must be defined and negotiated through presentation. Predictive triggers include schedule triggers which invoke a finite sequence of displays or the performance of a module in a curriculum.
The ‘understanding’ of the trigger depends on the validation context of the target setting. For example, it is impossible to decide from the outset if a question can be understood without access to the validation context which renders it meaningful [16]. The resonance of the trigger generates a provisional problem pattern which puts in relief relevant resources in the mind, at hand, or second-hand: I know that so-and-so knows, or that the answer must be in this collection. The validation context includes the hypothetical validation of predictions: it affords the imagination of situations in which the decision based on presentation will turn out to be right or wrong.
Even in a stable validation context, the trigger never fully specifies the . Follow-up questions will unearth further layers of context— ‘common understandings’ are inexhaustible [17]. For example, without being situated in a conversation that constrains its semantics, a question like ‘Where is the pressure filter?’ can be answered in numerous legitimate ways, for example, by naming the assembly of which it is a part, locating it in relation to gearbox or whole engine, quoting the IPC reference, describing what cowls one has to lift to access it, and so on. All these answers are correct—they merely address different views of the same referent.
As presentation advances the understanding of the problem, it produces new problem dimensions, such as hypotheses or questions. The ‘information need’ is activated through resonance of the currently focused problem dimension on both patterns and display. A difference in resonance causes an interference pattern which cues dimensions as incomplete or contradictory. This in turn triggers the question or navigation aimed at saturating the problem dimension and filling in the display [18]. For example, resonance of a problem hypothesis of ‘low oil pressure’ on the domain pattern activates the engine's history of oil consumption as a relevant problem dimension that will not match display resonance if facts saturating this dimension are missing in the report. The resulting interference pattern activates the ‘need to know’ and triggers the consequent question or query, which will in turn trigger retrievals or measurements further down the line. Interference may also appear when a displayed hypothesis fails to correspond to a hypothesis produced through pattern resonance. For example, an airline's hypothesis that a particular pump is the source of debris blocking a bearing chamber outlet interferes with service engineers' pattern resonance producing a different ‘most likely case’ hypothesis. This interference triggers a visual inspection of a systems diagram in order to reveal evidence eliminating the displayed hypothesis, leading to the discovery of a filter between pump and bearing chamber —a fact that makes the articulated hypothesis rather unlikely. This shift of focus also produces the filter as a new relevant problem dimension to be saturated (-was the filter perhaps blocked and bypassed?) on the way towards validation of the hypothesis.
The movement from trouble to problem and from problem to fix involves transformations on different levels [19] (cf. section 5.2 Transformation in chapter 5–Articulation). The ill-understood trouble (first presented in the trigger) is transformed into problem hypotheses which are transformed into questions. These questions are transformed into measurements of values and retrievals of attributes that are transformed into facts that work towards elimination, confirmation or change of hypotheses. At the end there is an explanation that suggests a problem —which may turn out to work or may cause new trouble.
Presentation moves not only from unsaturated towards saturated problem dimensions: on all levels, results of saturation can change protocols for saturation. Unexpected values can inform different measurements. Surprising facts can inform different questions. A failing explanation can inform the recasting of a problem. An equipment fix can produce new unexpected trouble.
When the problem is fixed, delegated or has disappeared, presentation has reached close-out. The term close-out appears in vivo in the service engineering context: cases are closed—or problems ‘put to bed’—with a close-out report. The validation context defines protocols (cf. section 3.3 Validation context in chapter 3–Context). The problem referent may change during presentation and is then passed on to another setting. For example, an operational problem initially attributed to the engine may eventually turn out to be caused by another aircraft system and is delegated to the airframe manufacturer. A problem may also disappear because the underlying trouble disappears: a pipe blocked by carbon, for example, may become unblocked through oil pressure, heat or engine vibration.
In domains not grounded in referent processes, close-out depends on validation contexts negotiated during presentation. In the dialogical evaluations this negotiation appears as a sequence of offers and refusals up to the point at which evaluator and user tacitly agree that a question is ‘done with’ (cf. appendix V–Evaluation). Validation protocols for close-out are situational rather than operational: they emerge through the mutual interplay of tact and estimations of the threshold of effort the evaluator can demand, rather than on measurements of the ‘correctness’ of given answers. Regularities emerge which are protocol-like but not strictly quantifiable. This can be shown in situational modifications to the basic primitive of evaluation, the question.
In a dialogical evaluation, the evaluator is likely to restate the question if the user fails to answer, but only for a limited number of times and under constant modification of the question. A precise repetition would be mechanical and insulting in that it would appear to be ignoring users' efforts.
Instead, restating undergoes recursive changes which reflect the adaptive learning process of both user and evaluator. The evaluator may restate through actual or implied question repetition, e.g., by pruning digressions and problem drift; or restate with relaxed question specificity or changed emphasis (e.g., ‘how much can you tell from just looking at it?’); or restate on the level of validation context by insisting on some answer (‘Imagine you had to come up with an answer’); or restate on the level of operational protocol by advising actions that may lead to an answer (‘it is an interactive system…so what would you do to find out’); or, finally, restate by offering a re-negotiation of validation context, e.g., downgrade saturating condition from ‘answer’ to ‘guess’ (‘Just have a guess’).
Emergent presentation and schedules develop the problem in different ways. While users' aggregation of the in emergent presentation is highly recursive, the schedule accumulates its topics in linear progression and thereby constrains users' response presentation. The scheduled topic is only appropriated by users to the extent that in their response presentation they succeed in realising it as a solution to an implicit problem. The topic of ‘oil system’, for example, is understood to the extent that users realise the that a particular dimension of the system has solved. The general problem of heat and friction developing in the bearings is solved by a supply of oil; the particular problem of getting oil to locations deep inside of rotating assemblies is solved by a central oil tube, and so forth.
The success of aggregation in response presentation depends on the demands of the pace and topic difficulty of the schedule relative to users' experience and degree of resonance. Another aspect is the availability of articulation media. In settings that discourage interruptions, the only available medium for articulation is users' imagination—a transient and vulnerable medium overlaid by the ongoing schedule. The possibility of re-contextualisation in emergent presentation, even the availability of pen and paper to make notes, increases the chance of effective aggregation [20].
One consequence of paced schedules is that topics have no time to become understood as problem. They do not sink in, but only leave anecdotal surface marks in users' memory. It is telling to see how in a scheduled setting such as an engine familiarisation course, the topic suddenly turns into a problem when one user questions a particular design feature. The feature itself, such as the position of an cooling air outlet, may be irrelevant, but it acts as a catalyst for aggregation. Emergent presentation now aggregates the original larger problem which motivated the design of the detail, and renders the chosen design as just one solution which may be worth improving. Design appears as ever expanding space of design alternatives, and learners turn into designers to the extent that they begin to think of, and articulate, these alternatives.
As the trigger sets off aggregation, users build the core of the problem pattern according to emergent probability assessments before the growing repertoire of problem-specific resources becomes available for recourse. The problem pattern is therefore first of all an instantiation of the domain pattern, caused by the initial set of symptoms communicated in the trigger. The pattern substrate is both the resonating problem-relevant repertoire of material resources, and situated rememberings of problem-relevant dimensions from other cases.
The problem pattern settles in the problem-specific display—the pattern of documents spread out on users' desks, filed in daily files, or pinned up against the wall. It also settles in annotations such as the hand-written cluster of telephone numbers pencilled onto the top of the file which point at people concerned with the problem, and in problem-specific repertoires of documents, such as memos, data logs, reports, or photographs. Filing on personal, group, section and organisational levels aggregates problem patterns as a permanent document system—which, unlike the mental substrate, forces discrete decisions, such as selections (‘Throw away, or file?’) and orders (‘File under category X or Y?).
Users' remembering is the other substrate of the problem pattern. Resonance affords users' association [21] of other problem-relevant instances which extends the temporal and spatial scope of presentation. The current problem of this–engine, for example, associates the history of problems of this–engine, or similar problems of the–engine (the engine type) across several this-engine instances. At the same time every new problem pattern differentiates and further develops the domain pattern.
The problem pattern may develop through resonance on a myth that has become part of the domain pattern. For example, a service engineer related that ‘in response to High Oil Consumption, there is a tendency with certain operators to immediately think of changing the oil pump. In fact oil pumps on the RB211 have virtually never caused HOC.’ (Laffan 13/11/95) [22].
The term problem pattern is paradoxical: it ‘blackboxes’ particular problem dimensions within a boundless body of resonance that only takes shape in occasioned resonance, while at the same maintaining a distinction between problem context and domain context. The black-box nature of the problem pattern is indicated by the fact that users' presentation often turns on the problem in a way which does not suppose reflection and explicit reasoning. The term problem pattern also serves to de-individualise the problem. It expresses the importance of both similarity and difference in aggregation. This means that the problem develops its particularity simultaneously in association to and distinction from prior problem and domain patterns.
The weakness of the term problem pattern is that as a black box, it must be taken on trust. It stands against the production of explanatory artefacts through modular conceptualisation. But by the same token, it sanctions the metaphysics of ‘it just works’. Delving into the neurophysiological mechanism of pattern adds little to its understanding [23]: its complex, recursive, and massively parallel overall activity only emphasises what can already be gathered in observation, namely, the high degree of mediation where every new problem dimension instantly creates a field of relevant similarities and differences. Presentation of this field reflects, produces, and modifies problems at the same time.
The problem affords projection from the problem pattern onto the display. The strength of projection depends on the amount of users' relevant generic and domain resonance. The stronger the resonance, the stronger the projection caused by resonating problem and domain patterns [24].
A common feature of projection is the phenomenon of binding and hiding of display tokens. Projection strongly anticipates saturation of a problem dimension by a display token. The first resonating token is bound to the dimension. This binding hides other tokens. It can obliterate the material display to the extent that users see non-existent features and fail to see seemingly obvious ones.
For example, a service engineer intimately familiar with the Trent oil system displayed a cinegram animation segment called ‘cold start-up’ which shows what happens in the Trent Pressure Pump and Filter component during start-up. Since in all similar oil systems a pump relief valve opens during the cold start-up procedure, projection bound the expected relief valve to the first thing in the display that moved after the pump had begun to rotate—a filter anti-leak valve which sits at a different place, looks quite different, and was well known to the user.
Projection can bind the problem dimension to a display token discovered prior to a more fitting token if resonance produces it as related to the anticipated target. In a cinegram evaluation, one service engineer clicked at the main gearbox in the overview diagram in order to navigate the pressure pump which is clearly shown at a different location of the same overview. Domain resonance had afforded a projection that instantiated a relation between gearbox and pump since in all similar systems these components are adjacent. This binding of the —invisible—pump simultaneously hid the visible pump icon on the same overview diagram.
Projection may be very strong and persistently hide the presented resource. An erroneous binding of a projected problem dimension to a particular displayed gestalt may aggregate a distinct concept which from then on constrains aggregation. For example, a user looking at a particular drawing may at some point conceptualise the oil flow through a relief valve the wrong way round (cf. figure 4.2) and from now on consistently fail to notice the actually displayed flow direction.
![[3 view of a component showing how projection wrongly conceptualises fölow direction]](Figures/Fig4-2.gif)
(a) (b) (c)
Figure 4.2. A diagram of the pressure pump and filter component (a). In the case of filter blockage, a user conceptualised the oil flowing from left to right, thus partly bypassing the pump (b). The correct event however is a recycling of oil back to the pump inlet (c).
Projection also bridges the temporal and semantic gap between the initial trigger or information need and the eventual saturation. This bridging works in two directions. Presentation may produce a question which with hindsight renders a past focus as a meaningful answer. Or a focus may realise some token as possibly relevant in anticipation of future questions. During evaluations of a maintenance manual, users came across tokens that felt relevant, so that they cued them as potential answers through remarks, marks, earmarking or note-taking; later, when an appropriate question appeared, they re-articulated the token as a fitting answer(and often navigated to the cued resource to ascertain the validity of their remembering).
The problem is not defined once for all at the beginning like a scheduled topic to be merely elaborated. Aggregation can change it substantially—it may even replace the problem referent. There can be gradual changes such as problem drift, or sudden flip-over changes. In troubleshooting, aggregation looks for a clear favourite among possible problem referents, such as the set of components which may have potentially caused a fault. In evaluations, problem drift is often a result of a projection which replaces the initial intentions expressed in the trigger. Problem flip-over may occur when users reach a point of breakdown, e.g., when they discover that they have misunderstood a question.
The likelihood of problem change during presentation is determined by a range of factors: the status of problem referent; the pertinence of the display (the trigger, for example, may be a verbal or printed question); problem complexity; or the stability of the setting.
Scheduled settings and professional settings define the problem referent in different ways. In scheduled settings such as evaluations or training courses, topics or questions spell out an example problem referent, whether users understand it or not. The example may be subject to different understandings and projections, but it is static, i.e., not grounded in the dynamics of processes in the referent domain. In professional settings, the trigger affords hypotheses which are recursively changed and eliminated until a likely problem referent is found. Problem changes are not (or not primarily) caused by the contingencies of presentation, but by measurable changes of the referent domain, for example, the equipment. These changes can be gradual or ‘flip over’. A possible problem hypothesis may posit that oil leakage at a bearing seal causes odour in the aircraft cabin. An equipment modification planned to counter this problem will be ‘put on ice’ if it turns out that other measures have gradually improved the problem. Also, it may eventually turn out that the cause of the cabin odour was wrongly attributed to the engine. This flip-over of problem referent will cause presentation close-out for those concerned with engine problems (and delegate presentation to others concerned with troubleshooting the airframe).
The more pertinent (the less transient) the display, the less likely is problem drift. The confluent availability of the displayed resource provides a source for recourse and calibration while presentation proceeds towards close-out. For example, in evaluations using question catalogues, the printed question affords a comparison between the offered close-out and the initial pertinent trigger: users have to insert their answer right underneath the question. In contrast, users in dialogical evaluation settings have no pertinent display to keep their presentation on course. Instead, the problem and validation protocols are articulated in a transient medium, conversation, and are frequently amended through dialogical feedback.
Complex problems set off by open triggers are more likely to change than those set off by predictive triggers. Complex problems may involve many factors and change many times. There are cases where the problem is never safely established because of conflicting evidence, difficulties of predicting changes of the problem referent, or the disappearing of the trouble. Problem complexity makes the prediction of successful close-out difficult. A service engineer estimated that only about a half of all engine modifications (based on an ‘understood’ problem) lead to lasting improvements.
Presentation develops and recursively defines the problem in transient and material articulation. Articulation qualifies confluent resources and transforms evidence into new resources. It is the subject of the next chapter.
[1] The concept of focus corresponds to Shanon's (1993 pp283) concept of crystallisation as ‘a cognitive system's generation of concrete, articulated expressions.’ Shanon includes thought sequences and mental images which he describes as crystallisation in the medium of consciousness, imagination.
[2] A similar concept of resonance has been suggested in ecological psychology. Goldfield (1995 p66) relates that Gibson, ‘influenced by the ideas of Lashley, proposed that during information pickup, the nervous system "resonates"…to information.’ Michaels & Carello (1981 p64) illustrate Gibson's concept of resonance by likening it to radio reception and tuning to one of the incoming signals. Information, in this view, is the modulation or interference of the radiation. This leads the authors to the following illustration: ‘A pencil lying on the table broadcasts its structure by modulating the light in special ways. If we imagine that a perceiver is looking for something with which to write, he or she is tuned in or prepared to resonate to information specifying the affordance of writing. When the pencil comes into view, the perceptual system resonates to the information—which is to say, the affordance is detected.’ (op cit p65)
[3] A related expression is [visual] ‘pop-out’ (Treisman, quoted after Rogers 1995 p482). The acoustic equivalent is the moment when we suddenly attend to a peripheral conversation when one of the speakers utters a familiar name.
[4] Winograd & Flores (1986 p76) express a similar idea in describing the mutual orientations of users within a ‘consensual domain of interlinked patterns of activity’. It is important to note that while such consensus may exist on the level of language, it is often lacking on the level of domain interest. In the troubleshooting context, for example, a consensus as to who or what was responsible for an engine event includes the assignment of costs between engine manufacturer and operator.
[5] See, for example, Norman's (1986) model of interaction, Hegarty's (1988 p52) flowchart model of text-and-diagram comprehension, Elgot-Drapkin's et al (1991 p84) memory model, Rogers' (1995 p489) model of visual interaction, or Narayanan's et al (1995 p521) model of diagrammatic reasoning.
[6] The theory of ‘mental models’ deserves a longish footnote. That people informally talk about having mental models does not prove the existence of mental models as mechanisms of cognition. Critics have noted that ‘the concept lacks clear definition’ (Rowe et al 1992 p1195) and that mental models are ‘…incomplete, …unstable, …[without] firm boundaries, …unscientific, [and] … parsimonious’ (Norman, quoted after Driscoll 1994 p152). A central question is whether mental models are epiphenomenal presentations or underlying mechanisms of human thought. Johnson-Laird (1983 p402) advocates the latter. He claims that ‘all our knowledge of the world depends on our ability to construct models of it’. If it is granted that a model needs boundaries to function as a model at all, how does the model model its relation to what it excludes? Any increase in scope or granularity leads to a proliferation of apparatus and ends in the tautology of the world being its own model.
Another problem for the hypothesis of mental models is human consciousness. If the mind is composed of models, it requires ‘a mental model of a device that makes mental models’ (Johnson-Laird 1983 p470). This leads to the paradox of self-awareness, also-called homunculus problem: if the mind contains mental models which can be used or ‘run’ by the human agent (cf. Newman & Lamming 1995), then one needs to draw a line between the embodiment - technically, implementation - of the agent (or meta-model) and the models acted upon. Since the substrate of both models are the neuronal processes in a single brain, there can be no fundamental ‘hardwired’ difference in the type of embodiment, which means that it is necessary to introduce a second-order agent regulating the functional separation between models and first-order agent. Johnson-Laird offers a recursive solution of the paradox of self-awareness. But the essential indexicality that provides the contextual inferences that ‘flesh out’ mental computation (op cit p452) now evaporates in an infinite regress. The map which includes itself leads to recursive drawings of ever smaller maps, down to ‘a point at where the physical realization of the map would become too small to be perceived or to be physically drawn, but the values of the variables that were computed could, in principle, go on diminishing perpetually’ (op cit p471—italics added). The abstraction squeezes any awareness of physical referents out of the system; the flesh of the world boils down to formula and principle.
[7] My distinction between generic and domain resonance resembles that of Smyth et al (1994) who distinguish between algorithmic and memory retrieval. The authors say that general-purpose algorithms build up ‘instances in which a particular response is made in a particular situation…The more instances there are in memory, the faster retrieval becomes’(op cit p150). However, it is not the singular instance or some generic logic or algorithm, but the pattern of instances that activates the dominant response (cf. section 4.6 Projection in this chapter).
[8] Eco (1987) has introduced the distinction between co-text as the textual periphery of a lexical unit, and context as its semantic dimension, where the latter is inseparable from the reader's situated and historical interpretation.
[9] This should not be read to imply that the meaning
can be fully determined on the level of sentences; it depends on the occurrence
of the sentence within an articulated or designed context.
[10] Cf. Lindsay
(1994 chapter 1) who defines a primitive as ‘entity…proper for
that analysis in that situation.’
[11] A similar use of the term ‘trigger’ is made by Winograd & Flores (1986 p168) who describe triggers of conversations in and between organisations, i.e., when a customer request triggers a stock request.
[12] The purpose of this contracting out is to hide the complexity of certain operations, i.e., to treat a contingent and complex network of operations as a single budgetable object. The movement towards integrated service contracts which sell ‘power by the hour’ (cf. Skapinker 1995) will further reduce complexity for the airline.
[13] One may liken the trigger to Maturana's concept of perturbation irritating the internal activity of the system (cf. Winograd & Flores 1986); however, the permanence of a material trigger, such as a fax, means that it turns from irritant into a patterned (filed, annotated, etc.) retrievable external resource. This means that it is not sufficient to describe the perturbation and structural coupling on a perceptual level; the perceiver is simultaneously an actor recursively changing the status of the irritating percept. Retrievability over time constitutes ‘external’ reality. Whether one assumes, with Gibson, that the outside world is an invariant environment affording perceptions or, with Maturana, that it is the product of structural couplings of the nervous system makes little practical difference. It is telling that despite ontological differences in their construction of perception, both Maturana and Gibson (1979) are mainly concerned with individual behaviour such as orientation, not with social action leading to qualitative changes of the environment. This is reminiscent of Marx' (1970/ 1857-58 p188) critique of bourgeois economists (such as Ricardo or Smith) who select the isolated hunter and fisherman, not the context of social production, as starting point of their economic analyses.
[14] Cf. Ladkin's (1996) paper ‘Reasons and causes’ and the concept of causal hypergraphs in his analysis of the Lufthansa A320 Warsaw accident in ‘The X-31 and A320 Warsaw Crashes: Whodunnit?’
[15] The analytical theory of action which debated whether there are causal or logical links between intentions and actions (cf. Kelle 1994 pp65) seems largely irrelevant here, since it focuses (1) on intentions and actions of individuals, not actions defined by job descriptions, situated in settings, and tied into social and economic processes; (2) on explicit and rational intentions and actions suitable for syllogistic formulation (cf. Davidson 1980/1967 p158), thus neglecting automatic or patterned aspects of activity.
[16] Stubbs (1983) discusses the complexity of question and answer in a conversational, that is, meta-linguistic and social context. He quotes Fries' (1952 p165) comment that ‘…the question itself is part of the frame in which the answer as an utterance operates.’
[17] Any recursive backtracking pushes back further the ‘accompanying "fringe" of determinations that are open with respect to internal relationships, relationships to other events, and relationships to retrospective and prospective possibilities’ (Garfinkel 1967 p41).
[18] A concept similar to that of unsaturated problem is the ASK (anomalous state of knowledge) model of Belkin, Oddy & Brooks (quoted after Schamber 1991 p6).
[19] A similar more specialised model for process control tasks in the cognitivist tradition is suggested by Rasmussen, quoted after Bergan (1995).
[20] Driscoll (1994 p96-7) reports that retrieval (an indicator of successful problem aggregation) is enhanced by note-taking activity. Rubens (1988 p169) states that ‘active, constructed responses have a significantly greater effect on the participant's learning of the instructional material than passive, multiple-choice responses.’
[21] In the terminology of dynamic systems theory, this association is explained through dynamic attractors that suck perceptions into existing trajectories crossing a dynamic landscape (cf. Thelen & Smith 1994 p56 122; Goldfield 1995 pp168). The metaphor itself explains little—its emphasis on the non-discrete fluidity of mental processes mainly explains why little is explained through metaphors. It has the advantage that it avoids the logical bias of explanations that posit discrete mental models (cf. Johnson-Laird 1983).
[22] My hypothesis is that the pertinence of such myths is due to the lack of a confluent presentation which would afford an understanding of the trouble. This lack is caused by the dispersed nature of maintenance activities (1), and the lack of an integrated case history (2).
[23] Cf. Rutkowska (1992). The posited ‘operational closure’ of the brain is of no consequence for higher level operations. Whether perception is described as the entering of stimuli from a reality ‘out there’ as in positivism and the dominant strand of cognitive science, or as perturbation of an autopoietic system (as, for example, in Maturana 1987 or Varela et al 1991) has no bearing on the operational mediation of subjects, the carriers of minds, on the level of society and its aggregates, such as organisations. Even if operational closure would exist in a narrow biological sense—a wider sense would at least describe body and mind as an action system (cf. Goldfield 1995)—it is illegitimate to assume such closure on the social level. The lack of mapping between communication structure and organisational structure indicates that institutions and other organisational units fail as substrate of systematic closure. The ‘system’ remains the wish (or unconscious assignment) that hallucinates a closure which is always a closure of symbolic expression. Kierkegaard (1968/1846) has demonstrated this for Hegel's system, Adorno (1973) for Heidegger's.
[24] Projection is related to the term transfer. It encompasses both automatic and conceptual shades of transfer. Automatic transfer was investigated by Schneider & Shiffrin (1977). As negative transfer, it explains slumps in performance when established patterns change abruptly, e.g., when the known position of two buttons is inversed. The conceptual shade includes the transfer of intellectual skills, such as defining concepts and applying rules, across domains (cf. Gagné 1985; Driscoll 1994, p160/336). Emergent and productive aspects seem not included in extant concepts of transfer. The term projection includes the recursive change of the source of transfer through the emergence of an articulated problem which binds and hides tokens as it progresses.