The process of developing flightdeck procedures and creating the corresponding documentation in the flight crew operating manual can be clarified and improved by using an act-function-phase model of documentation. The model systematically relates the acts of the dialogue among the crew and aircraft to the dialogue between the manual's author and its users, enabling co-evolutionary design. We show how the model can be applied to a variety of elements in the flight crew operating manual.
Keywords: Flight-deck procedures, documentation, dialogue, speech acts, co-evolution
Current approaches to the development of flight-deck systems, procedures and documentation typically follow the waterfall model, in which phases of development are generally not propagated back "upstream" in the process. Even in an iterative- or spiral-process model (e.g., Budde et al., 1992), the direction of propagation is still from system constraints to procedural requirements to documentation. Consequently, flight-deck procedures and their documentation may be sub-optimal because the finite nature of development resources limits the number of possible iterations. A more direct approach would be to enable co-evolution (Fisher et al., 1995) of systems, procedures and documentation, where changes in one part would be readily propagated to the others. Ultimately, it would be desirable to have computer-based systems for design and development that would propagate these relationships automatically so that the consequences of changes would be apparent immediately across all aspects.
To achieve this co-evolution of flight-deck systems, procedures and documentation, it is necessary to have a basis for understanding the relationships among the elements in a fundamental way. As a step in this direction, we propose an act-function-phase model of interaction that represents and relates differences in (1) the dialogues among the crew and aircraft on the flight-deck to (2) the dialogues between the author(s) of the flight crew operating manual (FCOM) and its users. The model relates the dialogue acts in one context to the dialogue acts in the other context. Specifically, the model accounts for the "domain" dialogue acts of doing things and the "meta" dialogue acts of communicating, across the contexts of prescription (i.e., flight-deck procedures) and description (i.e., systems, functions and constraints). We claim that this model leads to useful insights about how to develop procedures and FCOMs-and even the underlying systems-in a co-evolutionary way. In a practical sense, the model serves as a kind of lens through which to view current and proposed FCOMs.
In this paper, then, we review the FCOM, its role and its development method(s), and the key concepts of dialogue acts and meta-acts. We examine issues of development of procedures and FCOMs, explain the need for a theory that links them, and present a hypothesis about the relations among kinds of acts and kinds of users and uses. To address the hypothesis, we develop the act-function-phase model and demonstrate its application to variety of elements in the FCOM. We conclude with a discussion of the results and the consideration of questions raised by the model.
We begin by reviewing two concepts that are key to understanding the proposed model and its application. The first concept is that of the flight crew operating manual, which is the principal aircraft documentation furnished to the flightcrew by the aircraft builder and/or the airline. A subset of the FCOM, called the quick reference handbook, is also provided. The second concept is that of dialogue acts, which are an abstract way of characterizing interaction. Dialogue acts are an extension of the well-known speech-act model of conversation.
The FCOM is written and published by the aircraft manufacturer and sometimes adapted and republished by individual airlines to meet their own guidelines or standards for publications and operations. The FCOM, depending on manufacturer and airline, can be called the aircraft operating manual, pilots operating manual, or other variations along these lines. The FCOM typically contains a set of parameters, such charts for take-off and landing speeds as a function of weight, a set of explanations of the aircraft's systems and controls, such as the hydraulic system and the interface to the navigation computers, a set of procedures, such as for normal, abnormal and emergency conditions, and a set of checklists. The parameters and explanations can be seen as constituting the manual's function of description; the procedures and checklists can be seen as constituting the manual's function of prescription.
At Airbus Industrie and its partners, development of an FCOM begins with specifications of the design of the aircraft and its systems. These technical descriptions are furnished by the engineers to the authors of the aircraft's FCOM who then, in consultation with expert pilots, produce draft text of the manual's descriptions and procedures. In many cases, draft FCOM sections can be based on the manual for an earlier model, so there tends to be a great deal of similarity between drafts and earlier-published manuals. The draft materials are reviewed for technical correctness by the design engineers, and any necessary changes are made in the draft manual. The authors then produce a final version, which is sent to a production team who publish the printed manual.
The FCOM authors face a number of human-factors issues, including how to describe the constraints, systems and interfaces clearly, how to specify safe, effective and understandable procedures, and how to communicate these procedures to the flight crews. The industry's approach to the process connecting the technical descriptions to the manual appears currently to rely on the advanced expertise of the FCOM developers with respect to knowledge of systems, aviation, human factors and technical writing. There does not appear to be a systematic, much less automatic, path from system to description and prescription.
Speech acts, first proposed by Austin (1962) and principally developed by Searle (1969), form a model of human communication that accounts for the (a) relationships between an utterance, its symbolic or abstract meaning as an act, and its intended and actual effects on the hearer, and (b) the logical relationships among the abstract acts. There is a large literature treating speech acts from a variety of perspectives, and in particular speech acts have proved useful for computer systems that are intended to interact using more or less "natural" language.
Classical speech-act approaches have not been immune from criticism. In particular, Levinson (1981) pointed out, among other things, that speech acts do not seem to account for real-life interaction where a single utterance may have multiple intended effects or symbolic acts. De Michelis & Grasso (1994) provide a good summary of the chief problems of speech-act theory and propose a situated variation.
A key extension to speech-act theory has been the development of "meta" models that treat a wider spectrum of communicative acts, particularly those dealing with control of the conversation (Carbonell, 1982; Novick, 1988). These meta-act models use the notion of act to account for communication about communication, such as resolution of reference or turn-taking. Indeed, a conversation can be viewed as a multi-layered composition (Novick, 1988) of conversation acts (Traum & Hinkelman, 1992), that contain both the domain-acts that accomplish things in the world of the parties' nominal goals and the meta-acts that accomplish things in the sphere of the communication itself. Another way of distinguishing domain from meta acts is to distinguish between what the person wants to achieve or how they want to achieve this.
It turns out that the concepts of acts, both domain and meta, can be applied not only to spoken interaction but to written materials as well. It is easy to see that an author, no less than a speaker, will have some set of effects that he or she wishes to produce in the intended recipient. Typically these causes and effects are along the lines of wanting the text to cause the reader to know or to believe something. Possible intended causes and effects could also include wanting the reader to do something in the future, such as adhere to a standard procedure during an activity. Consequently, we use the term dialogue acts to mean a generalization of the notion of speech acts to include the complex of acts associated with communicative action, whether spoken or written.
Written materials can be also formalized within speech act-theory-despite the fact that this theory was mainly used in natural language dialogue studies-by assuming that (1) the act of writing is a kind of simplified dialogue between the author and his or her audience via the document, and (2) the act of reading is an other simplified dialogue between the reader and the author also via the document. While these activities might be seen as hardly constituting dialogues at all, they can be seen also as asymmetrical complements that rely on each other for a "shadow" partner in a temporally disjoint exchange. In acts of both writing and reading there is no immediate reaction from the author or from the reader. Authors construct their document as if they are speaking to their audience (future readers); they assume that the audience could (and would) react in the case of a dialogue, so they create implicit hypotheses about these reactions and write on the basis of the hypotheses. Readers also, in reading, build the knowledge they read as if they were addressed by the author in a dialogue; the act of reading is like a series of queries from the reader to the author. Written "speech" acts, just like oral ones, can cause intended effects in the world. In the case of FCOMs, for example, pilots execute procedures as an intended effect of the acts taken by the authors.
The dialogue-act model can also be applied to interaction with computer interfaces generally. A full treatment of this aspect is beyond the scope of this paper. We note here, however, that a symbolic act can be translated into a set of actions in an interface, regardless of whether those actions are expressed in spoken language or some other language of action suited to, for example, graphical user interfaces.
We now turn to some of the difficulties, both practical and theoretical, that leave aircraft operating manuals and dialogue-act models less useful than they might be. We will show how these two, apparently disparate things, are in fact linked in a helpful way.
Aircraft are extremely complex systems that have substantial safety implications and requirements. Consequently, developers of aircraft operating manuals face the problem of avoiding or minimizing the occurrence of factors that would adversely affect the safe operation of aircraft, such as inconsistencies, unclear or non-optimal procedures, and unclear or non-optimal presentations in the manual.
FCOMs sometimes have aspects of substance and of presentation that can be a contributing factor in incidents and accidents. There are many cases where crews fail to follow procedures, possibly because the procedure or its presentation made it less easy to understand, follow or remember. For example, failure of the flight crew of an MD-82 to comply with checklist procedures to activate an operable pitot/static heat system led to a rejected takeoff with the aircraft coming to rest on a dike and a tidal mud flat (NTSB, 1994a; see also NTSB 1994b). In some cases, the FCOM is unclear or lacks some pertinent information. For example, a Dornier 328-100 veered off the runway when landing because the manual had insufficiently defined information concerning the landing/taxi procedures, particularly with respect to nose-wheel steering (NTSB, 1995). Sometimes the FCOM fails to specify an adequate procedure. For example, the flightcrew of an Embraer 120, flying with total loss of hydraulic pressure, was unable to lower the landing gear because the manufacturer's written procedures in the manual for extension of the landing gear were inadequate in that they did not discriminate between hydraulic and electrical failures (NTSB, 1996).
In the speech-act and meta-act literature, the meta-acts are normally presented as an additional set of acts that, for example, express conversational control acts. Examples of dialogues are explicated in terms of these acts, expressed as additional layers of meaning (see, e.g., Novick, 1988; Traum & Hinkelman, 1992). These papers, though, all tend to assume an independence between domain and meta acts. That is, domain acts could be freely substitutable within a framework meta acts.
In the case of documentation of procedures, however, this assumption of independence can be shown to be incorrect. Documentation of procedures involves presentation of forms of interaction, and is necessarily meta to the underlying domain acts.
There is, in fact, a systematic relationship between the domain acts and the meta acts that present them to the reader (i.e., the pilot-user). This relationship can serve, in turn, as the basis for a systematic approach to developing documentation from the technical specification of both descriptions and prescriptions.
Domain and meta acts have different characteristics when used to produce the operational and referential functions of dialogue. Domain acts have a clear and systematic consistency across both functions. That is, an operational domain act will normally have a recognizable counterpart in the set of referential domain acts. In contrast, meta acts of operational dialogue and of referential dialogue are basically different. Meta acts in operational dialogue can be characterized simply on the basis of the interface through which the crew and aircraft communicate. Meta acts in referential dialogue, however, depend on the tools used by authors and on the type of the reader they are addressing. Thus one of our principal aims is to analyze the relationships of information transmitted in the two phases to develop a method that maps domain and, especially, meta acts across the two functions.
The nature of the relationship between domain and meta acts can be explored through development of a dialogue model that accounts for the FCOM's two principal contexts of interaction.
Viewing the development and use of the FCOM as interactive processes suggests that there are actually two distinct phases, with two corresponding kinds of use:
In the operational dialogue, the acts are generally (but not exclusively) domain-level acts among agents in the flightdeck. We use the term agent in sense that elements of the aircraft (especially its interfaces) and all members of the crew have defined roles, responsibilities, and capabilities (see Boy, in press). Thus the dialogues designed for the operational phase may also be among the users. As a typical example, an airline's FCOM for the Airbus A320 specifies a procedure for Flight Plan Completion in which key elements are carried out solely by the crew rather than through an interface to the aircraft.
In the referential dialogue, the acts are generally (but not exclusively) meta-level acts from the documentation to the users. The design of the understanding dialogue for the FCOM is an especially difficult task because the set of intended users of the manual is not homogeneous. Foreseeable users include not only flightcrew but trainers, future designers and engineers, future authors, and regulators (i.e., for certification). Each of these users has a distinct background, set of goals and mode of use. In this paper, we focus on one type of user in particular: the flightcrew. Similar analyses could extend the model to the remaining classes of users.
It is possible to situate the operational and referential dialogues in a model that accounts for both the kinds of dialogue acts and the kinds of FCOM content. The basic model is presented in Figure 1.
The model distinguishes four contexts that account for the components of the operational and referential phases. The factors characterizing the contexts are dialogue acts (domain, meta) and FCOM functions (description, prescription). As indicated in Figure 1, the four contexts are
While the model described in Figure 1 is expressed in terms of acts and functions of dialogue, it enables us to distinguish the four contexts of communicated information that constitute the two phases of operational and referential dialogues. During the operational dialogue phase between the system and the user, the designer specifies the description of different aspects of the aircraft (such as systems, constraints, and interfaces) so that the user can use the information in a way appropriate to his or her role, such as training pilot or member of the flightcrew. The author also specifies prescriptions (procedures). During the referential dialogue phase between the author and the user, at the domain level one again finds contexts that depend on the aircraft's system and procedures. At the meta-acts level, the contexts involve technical guidelines for writing the manual, as well as stylistic and typographic conventions. These rules facilitate the users' understanding of the manuals-and thus the aircraft's systems and their use.
If the model is correct, we can formulate a hypothesis that suggests a path toward systematizing the development of the FCOM by connecting the operational dialogue and the referential dialogue through the four act-function-phase contexts.
In particular, we claim that there is a relation between the domain-level acts of the operational phase and the meta-level acts of the referential phase.
We test our hypothesis through analysis of the FCOM for the Airbus A340 (Airbus Industrie, 1996). For this aircraft, as for virtually all modern-generation aircraft, we can find both domain and meta-acts represented in the FCOM and its use.
In the procedures context, we can identify a variety of domain acts that involve communication from the crew to the aircraft:
Some actions of the crew can be interpreted as "requests" rather than "commands" because the aircraft interprets and constrains inputs within the permitted flight envelope under most conditions.
Also in the procedures context, we can identify meta-acts that involve communication from the crew to the aircraft:
Conversely, we can identify domain acts in the procedures context that involve communication from the aircraft to the crew:
Finally, we can identify meta acts in the procedures context that involve communication from the aircraft to the crew:
In the referential dialogue phase, all acts are from author to reader, because the FCOM (at least for the time being) is a printed and thus static, non-interactive document.
We can identify two basic domain acts that involve communication from the document to the crew:
We can also identify a number of meta acts that involve communication from the document to the crew. These acts use elements of presentation or style to communicate the function of parts of the text or graphics. The elements include typeface, color, placement and literary style, and the functions include cautions and notes. For example a caution is a warning to the crew about a procedure, and is not a procedure in itself. Thus, it is necessarily a meta-communication. And as a communication, a caution has a number of stylistic elements that have meta-value, such as italicized type, the use of the introductory label "attention," and its placement on the page immediately following the procedure.
This aspect of meta-acts in the referential-phase can be formalized as titling acts (i.e., to entitle things) for elements such as definitions, chapters (and smaller) divisions, enumerations or other list forms, and tables. Some of titling acts that we have found in the Airbus A340 FCOM are presented in Table 1. We note that some meta acts can be expressed through use of a particular layout. Meta acts can also be composed recursively to express more complex document structures.
|Action(Object)Number of Arguments|
Thus the example of the caution message discussed above, whose function is expressed through multiple presentational features, can be represented by the meta act DRAW_ATTENTION(caution_text).
Traditionally, analysis of meta acts is based on the paper ("static") documents that form the current standard for industry use. However, the prospect of support for the development and use of electronic FCOMs through onboard computers involves different, complementary kinds meta-acts. By electronic documents we mean "dynamic" documents that go beyond simply presenting the static pages of printed documents on a screen. In this section, thus, we use the analysis engendered by the act-function-phase model to illuminate the differences that will arise in a digitally documented world.
For static (i.e., printed) documents, writing is performed by a set of actions that create, on one hand, the document itself and, on the other hand, a kind of record of the authors' intentions to communicate. Actions that create a document depend on the tools used to do so, such as grammar checkers and structure editors. For example, if a tool does not permit the user to merge tables, the author will find a more or less wordy approach to create a structure that corresponds to what he or she intended, with a substantial risk that (a) the style of the merged tables will be unsatisfactory or (b) the author will spend relatively too much time on the presentation. The means at the author's disposal enable him or her to express intended communications in more or less complex ways.
We can distinguish two types of action that can be formalized with the notion of meta act: (1) actions that enable expression of cognitive processes such as planning, text generation and revision (Hayes & Flower, 1980) that involve editing activities, and (2) actions that enable formulating actions. These latter actions can, in certain cases, carry meaning; for example, the use of the color red for the text of a warning creates a basis for identifying the text as a warning because of the use of color. In practice, the Airbus A340 FCOM uses, for example, the underlined label "Note" and an italic font to express that a section of text constitute a note, and the underlined label "Caution" and a roman font to express that a section of text is a caution. These uses can be seen as signaling conventions for warnings (cf., Edworthy & Adams, 1996).
Despite the distinction between the actions of writing as expression and the actions of directing behaviors, we can still use a single formalism to represent both of these in the form of meta acts that make clear the distinction between domain- and meta-acts. At a first level of abstraction, meta acts involve the structuring of the document that is, the acts structure the relationships of text (or graphics) units with respect to each other using a logical semantics that corresponds to that of the domain acts. The meta acts coincide with the document's planning process (Hayes & Flower, 1980) and reflect the distinction between descriptive and prescriptive FCOM functions.
Creating text structures can be accomplished with available software tools. For example, using an outliner enables the author to have a customized style for expressing the fact that headings at a certain level are all in the same font at a given size. Similarly, using a document structure standard such as SGML would also ensure this sort of consistency of style. Meta acts can be studied independently of the methods used because we are interested in the meaning of the actions. Tazi (1988) describes a more detailed study of text structures and their representation in an abstract formalism based on speech acts.
When the FCOM is provided and used on-line, it becomes dynamic in the sense that there could be a "real" dialogue between the document and its user. One would have, for example, interactive confirmations and procedures generally. For this dynamic FCOM, there would be three kinds of meta acts: meta-acts of document structure, as discussed with respect to static documents; meta-acts of navigation in the document; and meta-acts of interaction. Navigation meta-acts express the paths the user takes to browse the document via hypertext links; users might consult a stored history of these acts might in case they find themselves "lost in hyperspace." Interaction meta-acts are based on models of the reading task; the document itself could suggest reading paths that follow (a) the structure of procedures, for the prescriptive elements, and (b) the structure of objects, for the descriptive elements. In fact, an interactive FCOM could play an even greater role by keeping track of what the flight crew has (or has not) done, what actions they have taken (or not), and what values they have tested (or not).
True validation of the act-function-phase model would involve demonstration that use of the model enables co-evolutionary development of FCOMs with their associated systems and procedures in actual work practice. So what evidence do we have currently that would make adoption of the model rational, and would make it worthwhile to test the model in a realistic setting?
First, the model produced a set of acts, across all four contexts, that are consistent with both the functions and the phases. This suggests that it enables authors to be explicit in describing the intended effects of their writing actions during development. Second, the model provides a framework for producing authoring guidelines that are systematically consistent with the FCOM's aims. It is interesting to contrast the act-function-phase model with the "model" that underlies current FCOM development across the industry. In point of fact, the current basis of document preparation is not really a model at all, and was simply used rather than validated. Current methods appear to be based on developers' experience-based practices with regard to what an FCOM ought to contain and how it ought to be expressed.
This paper is part of a larger study that aims to produce advanced methods for the design and development of the FCOM. In this context, the act-function-phase model attempts to systematize the process of FCOM development around the human-oriented concept of dialogue acts, with the goal of safer operation of aircraft resulting from clearer interpretation by users of verifiable communications. The current model raises a number of further questions. Is it possible to develop a definitive and exhaustive catalogue of acts in all four contexts? How could the acts be defined formally so that their meanings and effects are known and verifiable? Moreover, the current paper deals principally with a single class of user; how should the model be extended to account for users other than pilots? How can the formalisms, which look toward outputs in terms of intended effects, account for inputs in terms of authoring guidelines?
A systematic approach to the development of flight crew operating manuals holds the promise of remediating problems of substance and presentation that, across the spectrum of aircraft manufacturers, have contributed to incidents and accidents. The act-function-phase model expresses the development and use of FCOMs as a set of complementary dialogues among authors, systems and users. While full application of the model will depend in part on the extensions discussed above, the model can be used, as illustrated in our analysis of FCOM acts, to understand whether the text of a manual actually accomplishes its authors intentions, whether a manual's language meets consistency or quality standards, and whether the manual properly connects its means of expression material to be communicated and learned.
This research was supported by a research contract from Aerospatiale Aeronautique.
Airbus Industrie (1996). A340 Flight Crew Operating Manual, 4, 1996.
Austin, J. (1962). How to do things with words. Cambridge, MA: Harvard University Press.
Boy, G. (in press). Cognitive function analysis. Norwood, NJ: Ablex.
Budde, R., Kautz, K., Kuhlenkamp, K., & Züllighoven, H. (1992). Prototyping: An approach to evolutionary system development. Berlin: Springer-Verlag.
Carbonell, J. G. (1982). Meta-language utterances in purposive discourse, Technical report CMU-CS-82-185, Department of Computer Science, Carnegie-Mellon University.
Edworthy, J., & Adams, A. (1996). Warning design: A research perspective. London: Taylor & Francis.
De Michelis, G., & Grasso, A. (1994). Situating conversations within the language/action perspective: The Milan conversation model. Proceedings of CSCW'94, Chapel Hill, NC, October, 1994, 89-100.
Fischer, G., Redmiles, D., Williams, L., Puhr, G., Aoki, A., & Nakakoji, K. (1995). Beyond object-oriented technology: Where current approaches fall short. Human-Computer Interaction, 10(1), 79-119.
Hayes, J., & Flower, L. (1980). Identifying the organisation of writing processes. In L. Gregg, & E. R. Steinberg (Ed.), Cognitive processes in writing. Hillsdale, NJ: Erlbaum, 17-35.
Levinson, S. (1981). The essential inadequacies of speech act models of dialogue. In H. Parret, M. Sbisa, & J. Verschuren (eds.), Possibilities and limitations of pragmatics: Proceedings of the Conference on Pragmatics at Urbino, July, 1979. Amsterdam: Benjamins, 473-492.
Novick, D. (1988). Control of mixed-initiative discourse through meta-locutionary acts: A computational model. Doctoral dissertation, available as Technical Report CIS-TR-88-18, Department of Computer and Information Science, University of Oregon.
NTSB (1994a). Accident March 2, 94, Flushing, NY, Docket No. DCA94MA038.
NTSB (1994b). Accident February 28, 1994 at Stow, MA, Docket No. BFO94LA053.
NTSB (1995). Accident August 3, 1995 at Portland, OR, Docket No. SEA95FA170.
NTSB (1996). Accident June 23, 1996 at Orlando, FL, Docket No. MIA96FA162.
Searle, J. (1969). Speech acts. Cambridge: Cambridge University Press.
Traum, D., & Hinkelman, E. (1992). Conversation acts in task-oriented spoken dialogue, Computational Intelligence, 8(3), 575-599.
Tazi S. (1988). Parsing textual structures from a typewritten author's work. Proceedings of the Conference on User-Oriented and Content-Based Text and Image Handling, MIT, Cambridge, MA, March, 1988, 925-939.
Tazi, S. (1989). Performing textual structures in documents. In J. André, R. Furuta & V. Quint (eds.), Structured documents, Cambridge: Cambridge University Press, 181-190.
This paper was published as: Novick, D. and Tazi, S. (1998). Flight crew operating manuals as dialogue: The act-function-phase model, Proceedings of HCI-Aero'98, Montreal, May, 1998, 179-184. © Copyright 1998 by ACM, Inc.