8
From Real Situations to Training Situations: Conservation of Functionalities
Simulation is often presented as the core of training and competences assessment in Dynamic Environment Management. The question of why and how to modify or conserve real task functionalities is a crucial one. Organizational settings, human-human interactions, resources and constraints are task components which may be affected in the transposition of real situations into simulation situations. The key components of professional expertise are presented in a model of operational knowledge as a framework for analyzing competences and training. The central point is then developed: How intractable new situations may become manageable by trainees. Confronting task analysis and operator’s model of operational knowledge serves to identify crucial variables for managing task functionalities for training goals and draw implications for training situations.
Simulation is often presented as the core of training in Dynamic Environment Management or at least as the main training tool, although evaluation of training on performance in real situations is less frequent: This raises the issue of the ecological validity of simulated situations for training. Moreover, numerous studies have been run using simulators to analyze operators’ competences and strategies, for instance on Nuclear Power Plant Control. One main and obvious reason is the scarcity of real problem solving situations being not pure routine operations which can be observed in research settings (the exception of traffic control with buses or planes and crisis management is due to the role of operative communication). Although simulation is virtually the only way to analyze and improve competence in Dynamic Environment Management, the question of why and how to modify or to conserve1 real task functionalities is a crucial one, if one wants to avoid circular reasoning about training efficiency.
After an era where tasks were defined mainly at the level of one man-machine interaction, the existence and effects of several tasks components led to the main conclusion that simulated situations could not be not reduced to technical simulators (Leplat, 1989). In vitro (simulation and/or experimental settings) and in vivo (on real DEM) observations and data show the effects of the environment in which a goal is defined for a task: namely human-human interactions, organizational settings, resources and constraints. The need to conserve functionalities in training situations touches on all these tasks components and are discussed in part one.
A theory of operational knowledge acquisition is also required. A model is briefly outlined in part two, in which professional competence is attested by but not reduced to the tasks an individual is able to tackle and solve efficiently, and includes symbolic representation (mental models) as well as experience. This model was developed to reflect components of expertise in academic or occupational domains, and to suggest some mechanisms for the development of competence. It aims at expressing training needs in terms of skills, knowledge, and abilities in task performing. Another dimension is the articulation of perceptual-motor, representational (cognitive) and psycho-sociological levels in the carrying out of real and simulated tasks: Needs for training “beyond technical knowledge” have to be stressed.
Last, but not the least, the key point is developed: How should intractable new situations become manageable by trainees? By modifying task complexity components, making all external conditions perfect, or decomposing complex tasks? Designing training tasks with real tasks functionalities in order to ensure efficiency on these real tasks is not the only goal: Designing unusual tasks to train operators to manage unforeseen situations and/or develop knowledge on crucial points is also needed. However, even if task functionalities are all, and clearly, identified, is it really possible to avoid all simulation biases?
In conclusion, some implications for training design in dynamic environment management are underlined, based on the notion of “generativeness” of competences and on the role collective reflexivity should play in individual and collective learning and in competence development.
TASK-ORIENTED TRAINING AND SIMULATION
Task oriented training requires cognitive task analysis in order to identify what is cognitively required for performing the task in an efficient and reliable way, and the consequences of conserving or modifying task functionalities in the design of training situations. These involve components of task complexity and dimensions of the organizational setting.
Conserving and modifying task functionalities
The term “task functionalities” refers, on the one hand, to properties of the deep structure of the task, involving cognitive requirements and, on the other hand, to properties of what is often seen as “context,” which are in fact closely tied to the task and have strong effects on operators’ behavior when performing a task. Designing training situations results in a trade-off between conserving and modifying task functionalities in order to manage efficiency in current situations as well as behavior accuracy when faced with the unforeseen (routine vs. crisis, see Norros, chapter 9, this volume) and to ensure or at least facilitate transfer of competences to future professional situations such as using new systems or facing new tasks due to change in organizational setting.
The discrepancy between operators’ efficiency after training in simulated situations and observations made “in vivo” on errors in performing real tasks has been clearly demonstrated in situations such as flight studies (Amalberti, 1993; Green, 1990). This led to the decision to conserve task components such as human-human interaction (pilots are trained on collective Cockpit Resources Management), and integration of specific tasks in a larger one (Line Oriented Flight Training). Our own studies on training in Emergency Management (Samurçay & Rogalski, 1993) show that competences acquired individually on a method taught as a support for decision-making in crises were insufficient to explain differences in collective use of this method as in real settings. There were major interactions between efficiency, task distribution, and effective use of the acquired method. Moreover, the organizational setting dimension was of main importance in the implement of the task (Rogalski, 1989). Regardless of domain, all these studies highlight the need to include human-interactions as specific task functionalities to be conserved in training tasks
However, modifying task components for training purposes is sometimes needed in order to control the level of task difficulty in training, to cover a wider range of tasks than the current ones, or to train subjects on specific difficulties. Controlling the level of difficulty does not always involve maintaining a constant level of difficulty for “adaptive learning,” or staying in the subject’s “zone of proximal development” (Vygotsky). For instance, teaching methods for a strategic approach to dynamic environment control may require dealing with the entire cognitive task complexity, to make the method meaningful, and long-term temporal organization of training is then used to handle complexity. Because the task changes with growing expertise, as it was shown for medical diagnosis (Boreham, Chapter 6, this volume) and through learning, changes are required, in and during training, to confront learners with different task demands.
Whatever the goal, conserving or modifying task components, there is a need for deep analysis of what can be transposed form real to simulated tasks, that is what the so-called “didactical variables” are in the cognitive, organizational, intentional and emotional dimensions of the task. Leplat’s definition of a task, “a goal to achieve in specific conditions: technical conditions of the controlled system, physical and organizational conditions and the operator” (Leplat, 1989), needs to be somewhat more detailed in order to specify possible variables between real and simulated tasks.
On the one hand, a task links a system of prescribers and a system of operators, with some explicit and/or implicit contract, which is an occupational contract in real situations, a didactical one in training. The “prescribed task” (Leplat) is seen from the point of view of the prescribers, the “effective task” from the actors’ perspective. On the other hand, a task may be defined by a set of three components. First, a task has a specific goal (a target state for the dynamic environment to be managed); this goal is embedded in an organized system of general purposes, goals and sub-goals, and may interact with the goals of other tasks (tasks interdependence). Second, criteria are used for evaluating if and how the target-state of the dynamic environment has been reached; these criteria may be hierarchized and are embedded in a system of more general values (personal and cultural values). Finally, a task has to be performed through a system of resources and constraints (time allotted, predefined rules or procedures, operative tools such as displays, support systems, etc.).
Realistic situations: Components of task complexity
Conserving or modifying tasks functionalities in training situations touches on the components of task complexity. We refer to a typology of process control complexity developed in chapter 1, this volume, and an analysis of components of complexity in open dynamic environment management such as emergency and crisis management, aircraft piloting; (Baron et al., 1989). Some of the variables which can be conserved or modified when transposing a real situation into a training one are examined below.
Temporal variables are probably the most crucial ones. De Keyser and her colleagues have stressed the cognitive difficulty in dealing with time (de Keyser; 1990; van Daele & de Keyser, 1991). In fact, several temporal variables are involved in an interactive way. “Tempo” may be defined by the relevant time unit to evaluate situation evolution (very high for flight, low in blast-furnace control): It introduces task constraints per se, such as time pressure in decision-making or stress. Continuity and duration of events in dynamic environments also affect task complexity (they are involved in time evaluation, memorization and planning). Latencies are also time parameters: Delays in information gathering affect subjects’ performance (Brehmer & Allard, 1991), system response latencies to operators’ actions may modify strategies, while delayed feedback about actions makes control and learning more difficult. Coordination of moments and durations for collective synchronization, that is succession and simultaneity, are parameters of tasks interdependence, and factors of complexity.
Control and command proximity are also key points as components of task complexity. They involve several dimensions: causality, “intermediate entities,” and mediatization through an automatized device or through human actors. The parameters are, respectively: the length of causal chains between “physical” measured variables, the existence of “intermediate entities” (expressing global properties of the process to be controlled at a symbolic level), which are “constructs” between physically measured and/or controlled variables (Samurçay, chapter 7, this volume), the “cognitive transparency/opacity” of the automatisms, and the “deepness” of human mediatization (complexity of the communicative chain in command implementation of action or in information gathering). Most often control and command proximity is intrinsic to the task: modifying them induces substantial change in the task itself. However external representations can be played with to manipulate cognitive transparency as a variable in training tasks.
Controlling the system of evaluation criteria and values is another important point in ecological validity of training: There are instances where trainees set an “operational approach”, in which a decision was evaluated by raw results, against a methodical approach evaluated in terms of its conformity to a “handbook” method: They did so although this method was designed to be and had the properties of an aid for operational decision-making. Moreover, it is not always possible to transpose the embedding of goals and values in training situations; rather the possible impact of the resulting modifications although difficult, needs to be analyzed.
The other task component, resources and constraints, is more flexible. The key is that procedures and operative rules are both resources, which enable rule-based instead of knowledge based activity, and constraints, which exclude other solutions which could be best adapted to the actual situation, due to circumstances. In her analysis of responsibility and risk control as elements of process operators’ professional expertise, Norros underlines that “facing and solving the basic conflict of doing wrong in order to do right… constitutes the essential learning process and results in the increase of expertise through which a stepwise integration of the traditional user orientation to the designer orientation is taking place” (Norros, 1989). Interaction of individual cognitive activity (and more general behavior) with properties of organizational settings is in fact the crucial factor.
Organizational settings
Organizational settings have effects which impinge on the transformations introduced in designing training situations in this multidimensional space. Defining dynamic environment management at the most general level (the abstract functional level in Rasmussen’s terms, Rasmussen; 1986) leads to the notion of virtual operational device, which involves modeling task organization in a “vertical” (hierarchical) dimension and a “horizontal” dimension of cooperative work (Rogalski, 1991). In real organizational settings, when a given control task is performed, operators cooperate in a real operational device, with rules for task sharing, information, and command flows. They are involved in complex prescribers/actors relationships (from the Management to the roundsmen in a plant for instance, or between a chief and his/her on-line staff).
Comparison of the communication flow in a real and in a simulated open dynamic environment management (large forest fire) reveals differences which can be explained by the transposition of the structure of the operational device (artifactual devices and human settings) to a training situation, whereas some other differences were due to a lower level of trainees’ competences (Rogalski, 1989). Training goals concerned with a complete achievement of cooperative tasks requires conserving operators’ organizational place and role in the “operational device,” whereas making the “rules of game” clear can be sufficient when aiming at a more specific technical purpose.
Studies in the military field have shown the need to take into account the entire context, as is the case in leaders’ training for battle as well as for garrison command (Hunt & Phillips, 1991). Taking into account the operational device embedding a given set of tasks implies overcoming the dominant and overly narrow meaning of team (as a permanent unit at a given hierarchical level) to incorporate the “vertical” dimension in collective activities and training. A recent book on team performance and training (Swezey & Salas, 1992) underlines the unresolved problems in this field.
Designing training situations for cooperative work requires taking the “secondary task” due to interaction into account. On the one hand, operators’ tasks are inserted in a network, from a logical and organizational point of view. Operators have to learn how to deal with this task interdependence and to manage the required communication. On the other hand, real tasks involve individual (and collective) representations of goals with potential conflicts, individual criteria, and systems of values which may differ; they also involve various knowledge and mental models about the situation itself. These two perspectives can result in different training situation properties. The first perspective centers on conservation of task functionalities (in terms of organizational settings), as observed in comparing situations for training officers in operational command posts for emergency management. The second may be achieved via meta-cognitive activities, as indicated by a study on collective operators training (nuclear power plant) for the management of unforeseen events where the main training goal was “to construct a common mental model and/or models of others cognitive functioning” (Jansens et al., 1989). A survey of this specific issue of collective training is presented elsewhere (Rogalski, in press).
MODELING ACQUISITION OF OPERATIONAL KNOWLEDGE
In operational knowledge, knowledge (K), experience (E), operative cognitive tools (O) and problem solving (PS) are dynamically linked in a model — KEOPS — developed for analyzing professional competence. Beyond technical operational knowledge, metaknowledge and psycho-social representations intervene in individual and collective activities in professional settings. All of these are affected by training.
Developing all the components of operational knowledge
Modifying or conserving real task functionalities in training situations depends on training purposes. The first main distinction differentiates training mainly concerned with increasing the automaticity of behavior and training concerned with developing the knowledge for devising a new working method (Bainbridge; 1989). Looking at professional competence for actors at various hierarchical levels in dynamic environment management should also involve consideration of the set of tasks actors are able to tackle and their capability to use cognitive aids, from making use of an abacus to calling for human expert specialists.
A diagramatic model of integrated operational knowledge is in figure 1 below. The four components (knowledge, problem solving, experience, operative cognitive tools) on which general cognitive “functions” operate, such as hypothesizing, inferring and deducing, planning, and controlling, are embedded in metaknowledge about tasks and about our own knowledge and know-how. In KEOPS terms, knowledge-based, rule-based, and skill-based behaviors are respectively connecting a given situation to knowledge structure, problem solving, or in an automatized, direct way to experience.
Training can be oriented toward all these components. New open situations develop Experience, structure Problem-Solving, and may require new Knowledge to be taught. They can also be used for structuring previous Knowledge in a new way: introducing “intermediate entities,” constructing new levels in the knowledge architecture. Developing Experience may form units from pieces of knowledge giving rise to “encapsulated notions” or from a set of isolated procedures to “precompiled strategies.” Available as units, these new entities in Knowledge structure contribute to decreasing the cognitive work load in Problem solving. This process also develops a form of automatization (in cognitive activity) which is constructed from explicit knowledge, from a skill-based behavior. It enables individuals to retrieve rule-based or knowledge-based behavior if required by incidents. A comparison of training methods for diagnosis of plant failures, based on one of three facets: a) knowledge ranging from faults to consequences; b) rules (after presentation of the plant), which in our model are defined as operative modes for the subclass of diagnosis PS in plant control; or c) only training on specific cases (experience) shows similar efficiency for trained faults, but better results for new faults for the method (b) based on rules for PS (Shepherd et al., 1977). Note however that the trainees performed less well on the post test as compared to the pretest on new faults, which could indicate that the individuals did not assimilate the knowledge on plant presented at the beginning of training.
Depending on the process to be controlled, training may have to cope with a transition in both directions from the physical level (in the real situation) to symbolic representations (involved in KEOPS): Controlling the properties of experience acquired through training situations with respect to what is needed in real situations may appear less complex in highly automatized processes where interaction with a process is nevertheless accomplished through symbolic representations. In fact, collective work on complex cases requires controlled interactions across various processing levels and raises the issue of developing or maintaining ability in relating symbolic representations (structured Knowledge and Problem-Solving) to perceptual-motor level (Experience). Collective reflexive activities on previous situations may also develop shared Experience (Baerentsen, 1991) and thus enable learning through others’ Problem solving.
FIGURE 1. KEOPS, a model for operational knowledge.
Knowledge, Experience, Operative cognitive tools and Problem solving are dynamically linked in KEOPS: In a given occupational domain, Knowledge structures Experience, and manages the use of Operative cognitive tools; Experience encapsulates Knowledge and automatizes Problem solving; Operative cognitive tools support Problem solving; Problem solving constructs and attests Knowledge, and produces and structures Experience; Metaknowledge and metacognition regulate cognitive activities.
Let us turn to the fourth component: “Operative cognitive tools” (OCT). Operative cognitive tools are external aids for problem solving. They take charge of some part of the cognitive load because knowledge is implemented in their design; they may be numerical aids such as tables of values on several parameters, graphical tools such as an abacus representing functions (depending on parameters), artifacts such as a pocket calculator or professional software. Integrating OCTs in the personal problem solving process is a component of professional competence. As underlined by Hutchins as regards technology in team navigation, they do not amplify the cognitive activities of the team members, but instead transform what are normally difficult tasks into easy ones. They need training to be used and “assimilated” in the Piagetian sense. In fact, operative cognitive tools may also play a role in modifying operators tasks and (internal) representations of the situation and hence the Knowledge component itself (Rogalski, 1993; Rogalski & Samurçay, 1993; Samurçay, chapter 7, this volume).
Learning how to manage task complexity is a part of methodological training which deals with Problem solving structuration (classification, identification of common operative modes, change of representative frames, strategy and tactics adapted to a class of problems). Managing task complexity is an important point in competence in handling new, unforeseen situations. From this point of view, Operative cognitive tools, which can be seen as interfaces between the individual and the task, are often an underestimated component in operational competence. Conserving or modifying the place of OCTs in task achievement may have striking consequences on what is really acquired during training.
Analyzsing operational knowledge into interacting components indicates that different training goals may be defined, leading to various constraints on appropriate training situations, and simulation situations are not the only ones dealing with problem solving and experience. Moreover the “target object” in dynamic environment management is a physical one (a chemical process, an aircraft in physical space, etc.) whereas simulations mainly deal with interfaces, at a more functional and symbolic level: They do not conserve a main task functionality.
Beyond “technical” knowledge
KEOPS schematizes relationships between the four main components of “technical” knowledge related to the content of the dynamic environment (what process has to be managed). Three other points are worth stressing. The first is metaknowledge, which is related to individual activity; one dimension of metaknowledge is reflexivity, linked to expertise (Olsen & Rasmussen, 1989). The two others involve psycho-social representations and abilities to manage social implementation of action (giving/receiving missions or orders); both deal with collective activities and organizational setting effects.
The role of metaknowlege in learning as well as in performing a task must be stressed: It intervenes in evaluating knowledge complexity, the “distance” between one’s knowledge and knowledge required for given problem solving; studies on pilots have shown that the choice of a strategy in mission execution depends on pilots’ representations of their operational abilities. This is indirectly related to individual emotional and motivational properties involved in decision-making, as well as in social interactions. This metaknowledge--self-knowledge--plays a functional role in regulating individual activities and managing internal resources and constraints, in particular under high external temporal constraints (Valot & Amalberti, 1992)
The role of psycho-social representations is developed by Norros in chapter 9, this volume. Her paper presents existing interactions between metaknowledge (awareness and language for making knowledge and strategies explicit) and collective competence. Her analysis converges with Hutchins’ study on team navigation, which underlines the fact that supporting the continuity of team over time through career cycles of team members (reproduction function vs. production function) involves knowledge overlapping. Overlapping of knowledge is facilitated by “openness” of operative tools and by the ability of more expert team members to make knowledge and strategies explicit. In turn, it “allows the system to respond robustly to failure of individual team members and to avoid breakdowns” (p. 218) (Hutchins, 1990).
This leads to the fact that beyond the production function of tasks which guarantees a set level of production, two other functions are present, and more or less fulfilled: a reproduction function of the system which performs the production tasks and a training function to improve operators’ competences. Training situations are mainly oriented towards the first and main function: production. More attention should be paid to possible interactions with the two others.
MAKING COMPLEX TASKS IN DYNAMIC ENVIRONMENT MANAGEMENT TRACTABLE IN TRAINING SITUATIONS
Managing task complexity
Decomposing complex tasks into unitary ones, and dividing long term work into more restricted time units provides solutions to some of the problems of managing task complexity, particularly when the training purpose is focused on skills automatization. However knowledge tends to be acquired at the lowest level, with local structuration; experience only deals with well-defined situations, while problem solving is concerned with “pocket problems.” Articulating such local fields of operational knowledge or automatized procedures is not a simple process and requires specific training actions. For instance, “Line Oriented Flight Training” was integrated in the pilot training as a response to the problem of articulating previously acquired skills on sub-tasks (Amalberti, 1993; Green, 1990).
Managing complexity due to task dependence (“parallel tasks” for an operator or interdependent tasks for a team) can be done in two ways: choosing tasks parameters in order to simplify all tasks other than the target-task, and increasing complexity in a controlled manner (Bisseret & Enard, 1972), or focusing on the target-task by delegating other tasks to trained subjects or instructors. Observations of high level firemen officers indicated that it is more efficient than training on separate sub-tasks, because it preserves the “logic of the global task,” which is the key-point to be acquired and is easily lost in task decomposition.
In a similar way, integration of technical knowledge (using operative tools, or acquiring specific notions related to the process) into training about organization and planning of operations involved in DEM requires a dialectical solution: Technical knowledge acquires meaning and operational “virtue” through real problem solving, while problem solving requires previous technical knowledge. In-service training observations show that a solution could be found in taking the long-term effects in the didactical processes and the constructive role of reflective thinking (metaknowledge) into account.
Another dimension of task complexity needs to be managed in training situations, which involves the problems encountered in information gathering or action implementation when operators have to mediatize their activity through devices or human beings. Simplifying real tasks can be done in training situations where “all external conditions are fulfilled”: no errors in communication flow; correct implementation when using “complex” operative devices or human mediation; no interaction with “external factors” (such as cabin problems in flight simulation or pressure of media on decision-makers in emergency management training).
Time management raises many open problems. Such constraints as constructing a correct mental representation of process tempo (as for instance flight simulation in Computer Assisted Learning) are unavoidable. However, as artifacts simulators can suspend time: It is a way to “discretize” a continuous environment, as well as modify time units for slowing an overly fast tempo. Metaknowledge may be used in managing action control: Suspending time enables proximal debriefing based on immediate feedback. As we know more about existence than about nature of complexity introduced by interaction between time parameters, managing task complexity through these parameters must be explored with caution.
Control and command proximity can be modified by “envisionment,” introducing a symbolic and fictive “situation visibility”, as in Brehmcr’s or Dorner’s fire simulators. This evisionment could possibly play the role of an operative dynamic image with respect to real situations (see discussion in Samurçay’s paper, Chapter 7, this volume): Validation of such a hypothesis requires studies on professional operators. Introducing feedback (retroaction) and/or anticipation of consequences of actions (proactive feedback) may be a productive means of shortening the control-command loop. As simulated situations do not have productivity constraints, which are concerned with general values above specific task evaluation criteria, and they allow for systematic exploitation of the principle of “knowledge of results” and “action shaping” (Leplat, 1989).
Training can also avoid or limit “over-complexity” of a task produced by the trainees’ actions which could make the situation worse, or create new problems to be solved in interaction with previous ones. This may be seen as a “positive one-way” for interaction between trainees’ actions and process evolution, but requires tight simulation-trainer interaction.
Design of unusual tasks and biases in tasks transposition
Transposition of functionalities of real situations to simulated ones is not always necessary or possible. On the one hand, training purposes may lead to design tasks which do not have any equivalent in real situations; on the other hand, there are avoidable biases due to the training situation itself.
In fact, training may require designing unusual tasks: modifying real task functionalities or defining highly unrealistic tasks with respect to work situations, or confront trainees with more complex tasks than are in real, common situations. Training purposes may consist of enabling trainees to tackle unforeseen or rare events; to confront them with specific points of operational knowledge in an appropriate problem solving situation, or to disentangle specific questions from “technical” problem solving, for instance by introducing “role plays” to initiate cooperative work.
Elsewhere, biases are due to the training situation itself: For instance in flight simulation, trainees expect troubles while in real situations incidents are perhaps expected (decreasingly with progress in automation), but accidents tend to be unexpected. It changes the orientation phase of the control task drastically in a general case, except for emergency management where accidents are initial conditions of action in real professional situations. Another unavoidable bias is related to risk transposition: No physical risk is involved in training situations; there are only cognitive and narcissistic issues in decision-making. Nature and possible effects of stress can not be managed in a simple way.
Conflicts between the didactical and professional “contract” may also arise in simulated situations. What the goals really are, the systems of values, who is the task prescriber, who evaluates trainees’ activities, the trainer or the professional prescriber? In high level firemen officers training, trainee’s evaluation of simulated situations showed that two approaches were possible: an “operational approach,” which was expressed through comments on situation ecological validity and representativity, and a “didactical approach,” with comments centered on the adequacy of the situation for implementing a previously taught method for dynamic environment management. Even in “full size” simulations, this conflict has had to be managed as a significant variable.
CONCLUSION
We defined a task as a node in a network, with partial order linking prescribers and executive operators; a task involves three components: goals to be achieved, evaluation criteria, resources and constraints. This definition allowed us to analyze task functionalities which should be conserved, and define those which can be modified for individual or collective training in dynamic management environment.
Beyond the well-known issues concerning simulation fidelity, a number of features have been emphasized. The first point is the need to recognize that the target task, as well as its goals and the actors involved in performing it, is “embedded” in a socio-technical system. Local functionalities of the “current task” have to be related to more global funtionalities of the embedding tasks. Organizational impact has to be evaluated in designing training situations where trainees may have to cope with conflicts about evaluation criteria or even goals, negotiate resources and constraints, and handle the ways in which their own decisions can propagate effects on lower and higher processing levels. Adequate mental representations on the part of actors about their own work position may in fact be crucial in unexpected situations, overriding their technical skills and their level of responsibility.
Another point is the issue of managing temporal task dimensions: Preserving the tempo of the dynamic environment is an unavoidable constraint and acts to alleviate conflicts on the skill-based control level in task performance. Conserving the temporal context also involves integration of (simulated) tasks into the flow of long term activities. Actually, several temporal horizons have to be considered in complex dynamic situations; they tie actors at various levels, from management to executive operators. In any case, the preparation phase (briefing in collective activities) and the phase of action evaluation (debriefing) should be included systematically in training settings using simulation.
Modifying task functionalities in order to make difficult situations tractable requires a cautious analysis of possible negative side effects, such as oversimplifying interactions between tasks, communication constraints, or complexity due to human limitations and possibilities for error. One solution could be to design simulations with “parametrizable” cognitive models of interacting actors (a solution which requires much research for cognitive modeling), and/or in designing training situations involving actors (trainees) at various levels in the operative setting, and various levels in previous training and competence.
The advantages of simulated settings, such as action recording, clock stopping, experimenting, etc., could also be used to develop self-knowledge at an individual and a collective level, that is metaknowledge about own knowledge, personal ways of thinking and reacting, possible biases. Methodological training about task and activity self-analysis could reinforce the purpose of developing reflexivity, aiming not only at increasing reliability in task performing but also engaging in a dynamics of training through the work situations themselves. From this point of view, initial or in-service training situations should provide opportunities to “anchor” in all actors the use of collective discussions for transforming individual experiences into shared knowledge, that can surpass knowledge and experience limitations due to the rareness of severe incidents, a consequence of the improvements in systems safety.
REFERENCES
Amalberti, R. (1993). Safety in flight operations. In B. Wilpert & T. Qvale (Eds.), Reliability and safety in hazardous work situations. Approaches to analysis and design (pp.171–193). New York: Lawrence Erlbaum Associates.
Bærentsen, K. B. (1991, September). Knowledge and shared experience. Paper presented at the Third European Conference Cognitive Approaches to Process Control. Cardiff, UK.
Bainbridge, L. (1989). Cognitive processes and training methods. In L. Bainbridge & R. Quintanilla (Eds.), Developing skills with information technology (pp. 177–192). Chichester, UK: Wiley.
Baron, M., Rogalski, J., & Samurçay, R. (1989). Un système de traitement de situations dynamiques complexes: Analyse de la décision avec une méthode de raisonnement tactique (MRT); conséquences pour l’élaboration d’aides informatiques et la formation [A system for managing complex dynamic situations: decision analysis through a method for tactical reasoning, MTR; consequences for designing support systems and training]. First European Congress on System Science (pp. 697–709). Paris: AFCET.
Bisseret, A., & Enard, C. (1969). Le problème de la structuration de l’apprentissage d’un travail complexe: Une méthode de formation par interaction constante des unités programmées (MICUP) [Structuring training for complex work. A training method based on continual interaction between programmed units: MCIPU]. Bulletin de Psychologie, 284, 632–648.
Brehmer, B., & Allard, R. (1991). Dynamic decision making: The effects of task complexity and feedback delays. In J. Rasmussen, B. Brehmer & J. Leplat (Eds.), Distributed decision making: Cognitive models for cooperative work (pp. 317–334). London: Wiley.
De Keyser, V. (1990). Fiabilité humaine et la gestion du temps dans les systèmes complexes [Human reliability and time management in complex systems]. In J. Leplat & G. de Terssac (Eds.), Les facteurs humains de la fiabilité dans les systèmes complexes (pp. 85–108). Marseille: Octarès.
Green, R. (1990). Human errors on the flight deck. In D.E. Broadbent, J. Reason, & A. Baddeley (Eds.), Human factors in hazardous situations (pp. 503–512). Philosophical Transactions of the Royal Society of London, B 327.
Hoc, J.-M. (1989). Cognitive approaches to process control. In G. Tiberghien (Ed.), Advances in cognitive science, vol 2: Theory and applications (pp. 178–202). Chichester, UK: Horwood.
Hunt, J.G., & Phillips R.L. (1991). Leadership in battle and garrison: A framework for understanding the differences and preparing for both. In R. Gal & A.D. Mangelsdorff (Eds.), Handbook of military psychology (pp. 411–429). Chichester, UK: Wiley & Sons.
Hutchins, E. (1990). The technology of team navigation. In J. Galegher and C.Kraut (Eds.), Intellectual team work (pp. 191–220). Hillsdale, NJ: Lawrence Erlbaum Associates.
Jansens, L., Grotenhuis, H., Michiels, H., & Verhaegen, P. (1989). Social organizational determinants in nuclear power plants: Operators training in the management of unforeseen events. Journal of Occupational Accidents, 1, 121–129.
Leplat, J. (1989). Simulation and simulators in training: some comments. In L. Bainbridge & R. Quintanilla (Eds.), Developing skills with information technology (pp. 277–291). Chichester, UK: Wiley.
Norros, L. (1989, September). Responsibility and risk control as elements of process operators professional expertise. Paper presented at the Second European Meeting on Cognitive Sciences Approaches to Process Control. Siena, Italy.
Olsen, S.E., & Rasmussen, J. (1989). The reflective expert and the prenovice. In L. Bainbridge and R. Quintanilla (Eds.), Developing skills with information technology (pp. 9–33). Chichester, UK: Wiley.
Rasmussen, J. (1986). Information processing and human-machine interaction. Amsterdam: North-Holland.
Rogalski, J. (1989, September). Cooperative work in emergency management: Analysis of verbal communications in distributed decision making. Paper presented at the Second European Meeting on Cognitive Science Approaches to Process Control. Siena, Italy.
Rogalski, J., & Samurçay R. (1991, September). Effects of expertise on knowledge representation in dynamic environment management. Paper presented at the Third European Conference Cognitive Approaches to Process Control. Cardiff, UK.
Rogalski, J., & Samurçay, R. (1993). Représentations: Outils cognitifs pour le contrôle d’environnements dynamiques [Representations: Cognitive tools for dynamic environments management]. In A. Weill-Fassina, P. Rabardel & D. Dubois (Eds.), Représentations pour l’action (pp. 183–207). Toulouse: Octarès.
Rogalski, J. (1991). Distributed decision making in emergency management: using a method as a framework for analyzing cooperative work and as a decision aid. In J. Rasmussen, B. Brehmer & J. Leplat (Eds.), Distributed decision making: Cognitive models for cooperative work (pp. 291–301). Chichester, UK: Wiley.
Rogalski, J. (1993, September). Designing operative cognitive tools for open dynamic environment management. Paper presented at the Fourth European Conference Cognitive Science Approaches to Process Control. Hilerod, Denmark.
Rogalski, J. (in press). Formation aux activités collectives [Training for collective activities]. Le Travail Humain.
Samurçay, R., & Rogalski, J. (1993). Cooperative work and decision making in emergency management. Le Travail Humain., 56, 1, 53–77.
Sheperd, A., Marshall, E.C., Turner, A., & Duncan, K. (1977). Diagnosis of plant failures from a control panel: A comparison of three training methods. Ergonomics, 20, 347–361.
Swezey, S.R, & Salas, E. (Eds.). (1992). Teams: Their training and performance. Norwood, NJ: Ablex.
Valot, C., Amalberti, R. (1992). Metaknowledge for time and reliability: Luxury or necessity? Reliability Engineering and System Safety, 36, 199–206.
Van Daele, A., & De Keyser, V. (1991). Distributed decision making and time in the control of continuous processes. In J. Rasmussen, B. Brehmer & J. Leplat (Eds.), Distributed decision making: Cognitive models for cooperative work (pp. 261–273). Chichester, UK: Wiley.
1Vocabulary choices require some comments. The term “conservation” is chosen in reference to the Piagetian notion of invariants, conserved through transformations. “Competences” is used for its wide range of meaning including knowledge as well as skills. “Didactical” is preferred to “didactic” because the term “didactics” is frequently used in the international community of researchers on mathematics education (for the scientific discipline itself). “Mediatization” is a neologism, as in French, used to translate the german word “Indirekheit.” The term “operational device” is used not only in its concrete meaning but also in a more abstract meaning.