Further Investigation of Factors That Affect Task Prioritization on the Flight Deck

The objective of the second study was to further investigate the pilots’ use of a subset of prioritization factors identified in Chapter 4 (Figure 4.8). Pilots flew final approach procedures on a part-task simulator. The retrospective interviewing technique, in coordination with a Challenge Probe Point (CPP) questionnaire was used to probe subjects for factors that influenced their task prioritization strategy. Additionally, task performance data was collected and analyzed with the objective of identifying a relationship between task performance and prioritization factors used by pilots.

The participants for this study were 8 airline pilots, all male, with an average of 6838 total flying hours. They had an average of 2531 hours of single pilot time and 481 hours of "glass cockpit" experience. Their age range was 25 to 52, with an average of 34.6 years. They were paid a stipend and travel expenses for their participation in the study.

The part-task simulator was the NASA Stone-Soup Simulator version 4.1 obtained from NASA-Ames Research Center. The hardware consisted of 3 SGI Indigo2 workstations, running the IRIX 6.2 operating system. The workstations were networked together, two for pilot displays and the other for the experimenter control of the simulator. The simulator flight control was performed with a B&G Systems Flybox and 2 mice connected to the pilot’s workstations. Video equipment included a Panasonic 8mm camcorder, Sony PVM-1910 video monitor, Videonics MX-1 video mixer to obtain the picture-in-picture configuration. Video from the pilot’s workstation was achieved using an SGI Galileo Video board and software for NSTC video output.

Data collection for the experiment consisted of four flight scenarios, designated as Standard, A, B, and C (Figures 5.1, 5.2, 5.3, 5.4). These were all variations of the same arrival scenario, with the differences existing in the Challenge Probe Point (CPP) of a particular scenario. A CPP was an operational situation during the scenario where up to 6 tasks could become active at the same instant. The pilot had to decide the order in which the tasks were to be performed, then actually perform the tasks. For example, in scenario A (Figure 5.2), the CPP was located at the SUNOL intersection, and was the focus of data collection for this scenario. At SUNOL, an ATC call was initiated, giving a speed clearance. Additionally, an engine fire was initiated, requiring the pilot to perform an engine fire checklist. The pilot was faced with the following 6 concurrent tasks that all required attention:

1. initiate a descent to 4000’
2. initiate a turn to a heading of 240
3. respond to an ATC instruction to reduce speed to reduce speed to 180 knots
4. reduce the aircraft’s speed to 180 knots
5. perform the engine fire checklist
6. configure the displays for the ILS

Scenarios B and C also contained similar CPP’s intended for data collection. The Standard scenario (Figure 5.1) does not include a CPP, and was used as a measurement of baseline performance for the pilot. The order that subjects flew the four scenarios was a four by four Latin square design replicated 2 times (Montgomery, 1997). This effort was an attempt to ensure that the learning effects of order on the performance data were counterbalanced.

In each scenario, the retrospective interviewing technique was the knowledge elicitation method used to collect task prioritization data (Salter, 1988; Cooke, 1994). This technique was selected for its low intrusiveness, ease of application and as a result of the study in Chapter 4, where it was found that no differences existed between the intrusive and retrospective techniques in eliciting information from the pilots regarding task prioritization factors.

A very common approach to designing subjective questionnaires is the Likert scale (Friedenberg, 1995). A Likert scale typically consists of a number of statements to which subjects respond on a five-point scale of agreement to disagreement, approval to disapproval, or favorable to unfavorable, depending on how the statement is worded (Graham and Lilly, 1984).

A CPP Questionnaire was developed for each of the 3 scenarios that contained a CPP and can be found in Appendix 5. Each questionnaire contained three sections. First, the pilot was asked the order in which the 6 tasks were actually performed. Next, the pilot was presented with a set of statements regarding the factors that were used to determine the priority that was assigned to three specific tasks. The pilot responded to this Likert-type statement (Graham and Lilly, 1984) with: Strongly Agree, Agree, N/A (not applicable in the current context), Disagree or Strongly Disagree. For each of these statements, the pilot evaluated, in retrospect if his use of this factor was appropriate then was given an opportunity to make any comments regarding the statement or its appropriateness. The third section of the questionnaire gave the pilot an opportunity to re-order the 6 tasks, then justify why he performed the tasks in such an order. The pilot could also decline the opportunity to reorder the tasks if he was comfortable with the order in which he actually performed them.

The model of task prioritization presented in Chapter 4 includes 12 factors that affect task prioritization (Figure 4.8). These 12 factors are the result of pilots verbalizing factors that were used in the prioritization of tasks. In the present study, a subset of these 12 factors was selected for further investigation on the basis of their frequency of reporting in the earlier study. The 6 factors included in the present study are status, procedure, value, urgency, salience and time/effort (for a thorough explanation of these factors, see Chapter 4).

Additionally, a subset of the 6 tasks associated with each CPP was selected for investigation using the questionnaires. At each CPP, exactly 3 tasks were probed for the factors that affected the order in which that the tasks were performed. The tasks were selected based on their associated classification under the Aviate-Navigate-Communicate-Manage Systems (ANCS) classification taxonomy common in both the pilot and research communities. Tasks were selected at each CPP with ANCS classifications. For example, in CPP-A, Navigate, Communicate and Manage Systems tasks were selected for inclusion in the probe questionnaire.

This resulted in the formulation of a 6-factor by 9-task matrix of datapoints. This was the basis for the 54 statements that probed the factors that affect task prioritization, which made up the second section of the CPP questionnaire. Table 5.1 is the structure for the design of these statements. For example, in CPP-A, the statement regarding the status factor of the "perform engine fire checklist" task (statement A3-1 from table 5.1) was the following:

At each CPP, 6 statements for each of 3 tasks was posed to the subjects, who responded with their agreement/disagreement to the statement. The subjects’ response to these statements was the dependent variable in the present study, and provided the evidence for the use of the proposed factors that affect task prioritization.

The CPP questionnaire was administered after each of the data collection scenarios (A, B, C). Immediately after the scenario was over, the subject reviewed a videotape of the CPP and complete the questionnaire. Again, each subject performed three scenarios that contained a CPP and one baseline Standard scenario that was a routine arrival without any unusual circumstances.

This experiment measured the response of the subjects to each of the statements included in the CPP questionnaire. This value was the primary dependent variable upon which an analysis of variance was later applied to determine the extent to which the prioritization factors were used to prioritize tasks during the scenarios. Following is a detailed description of the mixed factor design of the experiment.

Agreement/disagreement with the use of a prioritization factor in determining the order that tasks were performed at a CPP.

A fixed effect variable with 6 levels (status, procedure, value, urgency, salience, time/effort).

A fixed effect variable with 4 levels (aviate, navigate, communicate, manage systems).

A fixed effect variable with 3 levels (scenario-A, scenario-B, scenario-C).

A random effect variable with 8 levels (subjects 1-8).

In addition to the above analysis, two other analyses were performed. First, the order that subjects performed each task in the CPPs was analyzed to examine any trends that might be insightful as to the use of the prioritization factors. Finally, the actual performance data from the simulator was analyzed to distinguish the good from the poor performers.

The tasks performed in the part-task simulator were consistent with the previous experiment (see Chapter 4.)

Navigational equipment was limited to a single very-high-frequency omnirange (VOR) navigation instrument for this study (see Glossary in Appendix 1 for definitions of aircraft instrumentation). While this is a very minimal configuration, it was ample for the scenarios and is an accurate partial representation of true navigational tasks. VOR displays and controls consisted of the navigational display, which included a VOR deviation indicator, distance measuring equipment (DME) distance from the ground-based VOR, course setting and VOR frequency. The VOR frequency control was located on the radio displays, and the VOR radial input selector was located on the mode control panel (see Figure 5.5).

Communications between the pilot and simulated ATC was perform in a direct verbal exchange, as the pilot and experimenter were within approximately 5 feet of each other. Verbal exchanges were restricted to standardized radio procedures and there was no free conversation between pilot and experimenter during data collection scenarios. As a small added communications task, the pilot was required to dial in the proper communications radio frequency before an exchange with ATC.

Other system management tasks were performed in the simulator through various controls and displays. For example, equipment malfunctions illuminated the caution indicator and display a message in the Engine Indication and Crew Altering System (EICAS) display area. The equipment malfunctions could then be acknowledged and reset by performing a series of mouse click inputs to the multifunction display/control panel of the simulator.

A task analysis for this study was simplified and refined from the analyses performed by Alter, K.W., and Regal, D.M. (1992) and McGuire, J.C., et al. (1990). This resulted in the identification of the tasks listed in the pilot questionnaire. This was consistent with the ANCS task taxonomy accepted in both the pilot and aviation research communities.

Pilots arrived for the 5-hour experiment and immediately completed an informed consent document (Appendix 6) and pre-trial questionnaire (Appendix 7) to record flight experience, age and to ensure no extenuating circumstances interfered with the trial, such as excessive caffeine or lack of sleep the night before.

Pilots were then given a brief overview of the experiment. They were notified that their flying performance was being measured in this experiment and that comments made to the experimenter would be separated from their names and would be kept confidential. The sequence of the experiment consisted of approximately 3 hours of training on the simulator, followed by a break, then an additional 30-minute training session. Finally, the four data-collection scenarios were flown on the simulator.

The training consisted of a very structured, consistent presentation of the tasks needed to fly the simulator. Appendix 8 is the syllabus used for training and was adapted from the NASA Aircraft Operations Manual for the Advanced Concepts Flight Station, which is part of the Stone Soup Simulator documentation.

After training, a general description of the data collection procedure was presented, explaining how the interviewing technique was administered. The data collection scenarios would then begin. Once again, the order of the scenarios was determined by a Latin square design, as outlined above.

After the pilot had completed the last data collection scenario, he was immediately given a post-test questionnaire, inquiring about how comfortable the pilot was with flying the scenario and if the training was adequate for testing purposes. Additionally, the pilot was encouraged to discuss his thoughts and feelings about anything related to the experiment. This completed the experiment.

The data collected in this study contained several dimensions and lent itself to several different analyses. The analyses presented below are all directed at the subjects’ use of the proposed factors that affect task prioritization.

Each of the statements in the Challenge Probe Point (CPP) questionnaires (Appendix 5) were scored accordingly:

Subject Response Resulting Score

Strongly Agree +2

Agree +1

N/A 0

Disagree -1

Strongly Disagree -2

These scores were then tallied for each of the 6 prioritization factors and 9 tasks (3 tasks from each of the 3 scenarios). Table 5.2 presents the sum total and average for each of the 54 task-factor combinations, the sum total and average for the 18 CPP-factor combinations and finally the sum total and average for the 6 overall prioritization factors.

The data in Table 5.2 were the resulting scores of the Likert-type questionnaire developed for the CPPs. It is common to perform an analysis of variance statistical analysis of Likert-type data as long as the assumption is met that each statement is unidimensional (Graham and Lilly, 1984). In other words, the statement should measure only one subjective opinion and not two or more. In the case of the CPP statements, this assumption held true, as the dimension being measured was the pilots’ agreement or disagreement with his use of each factor in determining the order in which to perform the CPP tasks. The analyses presented below were performed using the Statgraphics Plus for Windows 3.0 software program from Statistical Graphics Corporation.

The scenario, with three levels (A, B, C) was evaluated using analysis of variance and was found to have an insignificant effect. This establishes that subjects’ use of the prioritization factors were independent of the scenario. Therefore, the scenario factor was removed from the model, resulting in a 3 factor mixed effect model.

The CPP questionnaire values were analyzed using ANOVA. Recall from the experimental design that this is a fixed effect factor having six levels (status, procedure, value, urgency, salience, time/effort). These factors were found to be a significant effect (F(5,35) = 5.14, p < .01). As can be seen in Figure 5.6, over the entire experiment, procedure emerged as the factor most agreed to by pilots for use in task prioritization, followed in descending order by salience, value, time/effort, urgency and finally status.

As depicted in Figure 5.6, a positive value represented supporting evidence that pilots used the factor in determining the priority of a particular task. The corresponding magnitude of the value represented the agreement with the factor, using the (–2, -1, 0, +1, +2) scale developed for the CPP questionnaires. The maximum value a specific factor could be in this analysis is a +2 (strongly agree) score. Similarly, a negative value corresponded to the subject reporting disagreement with using a particular prioritization factor (e.g., the status factor in Figure 5.6).

Each of the nine tasks probed in the CPP questionnaires were categorized into the ANCS classification (see Table 5.1). This effect was found to be significant (F(3,21) = 6.09, p < .01). As can be seen in Figure 5.7, the prioritization factors were most agreed upon by the subjects when performing the aviate tasks. The other tasks, in descending order, are manage systems, navigate, and finally communicate. Factors with a negative score (i.e., the communicate tasks in Figure 5.7) indicate subjects disagree with using the prioritization factors on these tasks.

The prioritization factor-task category interaction was found to be highly significant (F(15,105) = 4.32, p < .001). As can be seen in Figure 5.8, this finding establishes that the subjects’ agreement with the use of the prioritization factors depends on which task is being evaluated for execution. In general, the aviate and manage systems tasks show strong agreement, while the communicate tasks show strong disagreement in the status, value and urgency factors. The navigate tasks show mixed results.

The subjects’ effect on the resulting CPP questionnaire scores were also found to be highly significant (F(7,240) = 4.53, p < .001). As can be seen in Figure 5.9, four subjects reported an overall average agreement with the prioritization factors, while four reported an overall average disagreement. Subject 7 reported the highest agreement, while subject 5 reported the lowest overall score. This result indicates that there is little overall consistency between subjects in their agreement with the use of the factors that affect task prioritization.

Multiple range tests were performed to determine significance differences in the response of the individual subjects. Application of Fisher’s least significant difference procedure at the 99% level resulted in the identification of 4 homogeneous groups within the 8 subjects.

The subject - prioritization factor interaction was found to be significant (F(35,240) = 1.93, p < .01). As can be seen in Figure 5.10, the agreement with particular prioritization factor is highly dependent upon the subject. In other words, subjects show little consistency in their agreement with the use of prioritization factors.

The task execution order at each CPP was analyzed by adding the order in which each task was performed to the data set collected above for the initial analysis of variance. The effect of order was found to be significant (F(5,315) = 2.96, p < .02). Interestingly, as can be seen in Figure 5.11, subjects appear to agree with the use of the prioritization factors most for the task that is performed third. However, a Fisher’s least significant difference test at the 99% level fail to establish a statistical difference between the tasks performed in the first through fifth positions.

Recall from Chapter 3, an assumption made for the present research was that the order that the tasks are performed in is approximately equivalent to the priority assigned to the tasks by the subjects. Therefore, a second dimension of the data collected in the present experiment was related to the order in which the tasks were performed at each of the CPPs. Again, a CPP is an operational situation in the scenarios where up to 6 tasks become active at the same instant. The pilot must decide the order in which the tasks are performed.

The initial part of the CPP questionnaire (Appendix 5) required the subjects to record the order that they performed the multiple tasks. Table 5.3 presents the actual task execution order for each of the 8 subjects performing in each of the 3 scenarios. This was accomplished with a review of videotape of the CPP. The subjects were allowed to replay the video as many times as was necessary to accurately record the order that the tasks were performed.

In CPP-A, only two subjects, 4 and 7 performed the tasks in exactly the same order. Subjects 6 and 8 were very similar, differing in the order of just 2 tasks from the order selected by subjects 4 and 7. In CPP-B, no two subjects performed the tasks in the same order. However, in CPP-C, there was more consistency. Subjects 1 and 2 performed the tasks identically, with all subjects performing the stop descent task first and there was much more consistency in the remaining tasks than observed in the other two CPPs.

Additionally, the last part of the CPP questionnaire gave the pilots an opportunity to evaluate the order that they performed the multiple tasks. There was no feedback from the experimenter regarding performance during the CPP. However, the subjects had just completed an evaluation of the 18 statements related to the prioritization factors, so they had carefully thought about their justification of why they performed the tasks in the order that they did. If the subjects, in retrospect, revised the order that they performed the tasks, then the revised order is presented in Table 5.4. Over the course of the experiment, which included 24 CPPs, 13 times, or 54% of the time, pilots would have performed the tasks in a different order. An analysis of variance on the effect of the subject’s decision to change the execution order resulted in no significant effect on the subject’s responses to the factor statements in the CPP questionnaires.

The performance data collected from the flight simulator was analyzed for errors. In the present study, a performance error was committed only under very well defined and extreme conditions (see Appendix 9). As an example, when ATC issues an altitude clearance, FAA regulations stipulate that a tolerance of +/- 100 feet from the assigned altitude is an allowable deviation. To constitute an altitude deviation error in this study, the subjects had to deviate from the assigned altitude by more than +/- 200 feet. Pilots were fully aware of these error conditions primarily through their familiarity with the Federal Aviation Regulations (FARs). Other aircraft-specific error conditions were explicitly explained to them during the training portion of the experiment.

Over the course of the entire experiment, 7 subjects committed 9 performance errors in the 24 scenarios flown (Table 5.5). Figure 5.12 depicts the average score when an error is committed, while Figure 5.13 is the score when no error is found. An analysis of variance of this data found no significant differences between the error and non-error conditions on the subject’s responses to the factor statements in the CPP questionnaires.

One of the primary findings in this study was the individual differences pilots exhibited in the task prioritization process. This is consistent with the findings of Schutte and Trujillo (1996), in which they concluded CTM was largely dependent on individual differences between flight crews and personal style. While this study considered only single pilot operations, both the statistical significance of the effect of subjects on the questionnaire scores and the corresponding graphical depiction in Figure 5.10, it is apparent that there is no general prioritization vector for pilots as a whole, as suggested by others (e.g., Gopher, 1992; Logan, 1985; Adams and Pew, 1990).

However, this is not to say that prioritization vector is a poor conceptual approach to a representation of the task prioritization process. Since there is truly no single, optimal task execution order in the majority of situations on the flight deck, there may not be a single optimal task prioritization vector. Rather than a prioritization vector that is applicable to pilots in general, it may be more useful to be able to establish several prioritization vectors that all result in equally-optimal task performance. Following, then, a pilot’s individual prioritization vector could be compared to established prioritization vectors.

Another concept in support of the individual differences found in this study was the realization that each pilot has a unique knowledge base, which was used to apply these prioritization factors. The pilots in this study had amazingly different training and experiences, ranging from military pilots, to Alaskan bush pilots, to flying instructors. The expectation that individuals with such diverse experiences all have similar knowledge representations and apply prioritization factors in a consistent manner may be unrealistic.

In the present experiment, subject 1 had the distinction of being the worst performing pilot by committing a performance error at each of the 3 CPPs. The prioritization vector for subject 1 is given in Figure 5.14. The statistical analysis (multiple range tests) of this vector was unable to distinguish a difference between it and the vectors from subjects 4, 7 and 8. Interestingly, subjects 4, 7 and 8 each committed an error in one of the three CPPs with which they were presented.

Subject 2 was the best performing pilot by being the only subject to not commit a single error in any of the CPPs. The prioritization vector for subject 2 is given in Figure 5.15. Again, a multiple range test was unable to establish a statistical difference in the prioritization vectors between subject 2 and subjects 3, 5 and 6. Again, these subjects (3, 5 and 6) committed an error in one of the three CPPs.

With the limited number of data points in this study, it is not possible to establish the envisioned prioritization vector categories, but this may be a direction to pursue. It is methodologically reassuring that subject 1 (worst performer) and subject 2 (best performer) did not have statistically equivalent prioritization vectors.

The conceptual basis of the ANCS task hierarchy is that tasks have inherent priorities that are independent of the current context. In other words, aviate tasks always have a higher priority than navigate tasks, which have a higher priority than communicate tasks, and so on. However, it is not difficult to conceptualize situations in which this prioritization scheme breaks down and the tasks take on a priority inconsistent with the ANCS hierarchy. For example, it may be necessary to perform a manage systems task, such as balancing fuel loads in wing tanks, before the aircraft is maneuverable enough to aviate. In such a situation, the ANCS gives an invalid prioritization strategy.

However, the findings in the current study do reflect some of the characteristics of the ANCS prioritization scheme. For example, the aviate tasks do exhibit the most agreement from pilots in the use of the prioritization factors. Subjects responded with the justification of "fly the airplane first" in both the comments section of the CPP questionnaire and informal discussion after data collection was over.

There are two aspects of the results, however, that appear to contradict the ANCS hierarchy. First, the communicate tasks show an overall negative agreement score (Figure 5.9). Further, it is only the status, value and urgency factors that actually have negative scores (Figure 5.8). With respect to value and urgency, pilots often lowered the priority of the communicate task when they were given the opportunity to change the order in which the performed the tasks. The pilots know that both the value and urgency of communicate tasks are relatively low, yet for some reason they tended to perform these tasks earlier than they thought was appropriate. With respect to the status factor, see the following section.

Second, pilots performing the manage systems tasks appeared to utilize the prioritization factors more than might be expected. This may be explained by inadequacy of the general ANCS approach. There are many tasks on the flight deck that do not fit nicely into the ANCS hierarchy. For example, a non-normal situation, such as an engine fire, is difficult to place at a lower priority than the navigate and communicate tasks. In fact, the results indicate that pilots do not adhere to such a strict prioritization scheme. In the present experiment, many of the manage systems tasks were of a non-normal nature and the pilots prioritized accordingly.

One major inconsistency in the results was the overall negative scores of the status prioritization factor. This is especially disconcerting considering status was the most reported factor in the initial study of prioritization factors (Chapter 4). It was anticipated that status would have been an often-used factor by the subjects.

That the status factor was resulting in disagreement by the subjects was apparent early on in the data collection. Rather than stop the experiment after a few subjects, it was decided to continue data collection, but to add some informal questioning of each subject after the experimental trial was over.

It was discovered that the particular wording of the CPP statements regarding the status prioritization variable was inappropriate for some of the tasks. This is best explained with an example. The statement regarding the status of the manage system task, engine fire checklist, in CPP-A (Appendix 5, statement A3-1) appeared as:

For the subjects, this was a statement consistent with the nature of the engine fire checklist task (i.e., it made sense to the pilots) and the results for the score of this task were, in fact, positive (Figure 5.8, status factor in the manage systems task category).

However, when a similar statement was posed to the subjects regarding an aviate or navigate task, the fact that these tasks are not thought of as ever being "complete" in the context of a flight may have affected the resulting responses by the subjects. For example, a navigate task in CPP-B was to track the Instrument Landing System (ILS) instruments. The statement regarding the status of this task was as follows (Appendix 5, statement B1-1):

To the subjects, this was not an appropriate way to phrase this statement regarding the status of the task. Completion of this task occurs only at landing and it was not the intention of the statement to assess the subject’s evaluation of the task as related to landing the aircraft. Rather, it was the intention to evaluate the deviation from the localizer and glide slop indicators the aircraft was at that particular instant. This statement failed in the accurate assessment of this task and, in retrospect, other tasks in the experiment.

The solution to this situation was, in hindsight, rather simple. If the statement had been worded slightly differently, to reflect the context of each category of task, it is anticipated that the use of the status factor would have shown a much higher score. For example, the rewording of the B1-1 statement could have been as follows:

The informal feedback from the subjects, after data collection was complete, did not explicitly identify other obvious errors in the wording of the CPP statements. Therefore, it is assumed that the other statements resulted in satisfactory responses from the pilots regarding other task-factor combinations.

Recall that the 6 prioritization factors investigated in the present study were a subset of 12 factors identified in Chapter 4. They were selected by the frequency with which they were reported in the initial identification of the factors that affect task prioritization. By again fitting the 6 factors back into the 3 primary prioritization factors of Figure 4.8, the extent to which pilots agree that they used the 3 primary factors of task prioritization can be seen in Figure 5.16. In other words, three factors investigated in the current study, status, salience and urgency are all included in the primary factor status from Figure 4.8. Similarly, time/effort and procedure are included in the primary procedure factor. The value factor stands alone and does not include multiple factors from the earlier study.

With these results, it appears that overall procedure is the most influential factor that affects task prioritization. This is not entirely surprising, as piloting of aircraft is highly procedural and the current training programs reflect this in their highly procedure-based philosophies.

The effect of value was found to have more influence in task prioritization than was initially established in the previous study (Chapter 4). This is reassuring, as the one satisfactory order to perform the tasks would be to always perform the most important tasks to a level of satisfactory status before attending to other, less important tasks. Finally, the status factor did show limited effect in the model, but it is believed to have a bigger effect, as discussed in the previous section.

Current resource limitations enabled the current study to only look at single pilot behaviors in a part-task simulator. Since all of the subjects that participated in this study were current commercial transport pilots, their true operational environment was on a 2 pilot flight deck. It would have been much more ecologically valid to study 2 pilot crews. However, it is anticipated that this would have greatly reduced the workload experienced by the pilots in the CPPs, thus reducing the need for the pilot to prioritize 6 concurrent tasks. Therefore, in the current study of prioritization factors, the high workload situations created were conducive for the objectives of the research.

Although extensive training on the simulator was incorporated into the current experiment, it was still a very different operational context from the actual aircraft the subjects fly on a day to day basis. There are two directions future studies could take regarding this limitation. On the one hand, experimentation could be directed towards a full-motion, high fidelity simulator. This would provide an environment as close to the actual environment as possible, without actually flying in the aircraft. Current high fidelity simulators would provide an excellent research tool, as they are able to capture a wide range of performance statistics that could be used to accurately evaluate pilot performance. However, there are many obstacles to overcome, such as access to such a simulator and the prohibitive costs.

An alternative approach would be to not attempt to increase simulator fidelity, but rather to abstract task characteristics. Computer graphical simulation tools are available to create multiple displays, which could represent multiple concurrent tasks, with remote similarities to flying tasks, but perhaps more generalized to allow non pilots to participate in the experiments. It would be interesting to study factors that affect task prioritization for groups of humans other than highly trained commercial pilots. This approach might diminish the affect of a pilot’s knowledge base, as discussed above, and investigate what might be considered a more general prioritization vector.

The present study started with the conceptualization of the task prioritization process as a vector of factors and attempted to investigate those factors. A primary conclusion is that no general prioritization vector was found in the data. However, it may be that categories of prioritization vectors could be established. Pilot prioritization schemes could then be evaluated according to these categories. The prioritization vector approach may be an interesting research area to pursue.

It is very apparent that even highly trained and skilled humans do not always perform task prioritization optimally when presented with multiple, competing tasks. Pilots, when given the opportunity to (retrospectively) order tasks in a different sequence, chose to do so in more than 50% of the CPPs. This finding is consistent with findings in other domains (Moray, et al., 1991).

The ANCS hierarchy does not hold up rigidly in all circumstances. Subjects often performed communicate tasks much earlier than they should have, and acknowledged this when they reprioritized the tasks in retrospect.

The primary contribution the present work has made is the investigation of factors that affect task prioritization on the flight deck. This is by no means a completed task; it is just the beginning. While the present work has utilized and developed some creative methods for studying prioritization factors, there is a need for better, more robust and reliable methods.

What's New

Listed below, most recent first, are changes made to this page since its creation.

23 Dec 99

• original page