963

1 INTRODUCTION

The vast majority of maritime accident analyses

identify human error as a main or contributing cause.

Recent estimates range from 60 % (Butt et al., 2013) to

89.5 % (EMSA 2022, p. 5). However, according to the

classical accident pyramid, accidents are only the tip

of an iceberg, with incidents at the visible base and a

large amount of near misses and unsafe acts under the

surface (William 1959; Grech, Horberry & Koester,

2008, p. 17).

Human factors research could attribute unsafe acts

in many cases to a lack of situation awareness (SA), a

concept derived from aviation and elaborated by

Endsley (1995a). Endsley defines SA as a mental

process consisting of three successive levels: the

perception of relevant elements in the environment

within a volume of time and space (level 1), the

comprehension of their meaning (level 2), and the

projection of their near-future state (level 3). In the

maritime domain, 71% of human error in accident

reports (Grech, Horberry & Smith, 2002), or

correspondingly 50% of all accidents (Stratmann &

Boll 2016) were found to be related to a SA problems.

Loss of level 1 SA (perception) was reported most

frequently.

The human factor has long been neglected in the

maritime domain and has received increasing

attention only in recent years (Grech et al. 2008). A

focus of the latest empirical research is human

decision-making, mostly in the context of developing

maritime collision avoidance support systems (e.g.

Aylward et al. 2022, Butler et al. 2022, Fan et al. 2023,

Kizilay et al. 2023, Kartoglu et al. 2022), applying or

citing a variety of methods, including ship simulators,

interviews, surveys, observations, case studies, and

How Does Maritime Situation Awareness Depend on

avigation Automation and Mental Workload? A Sea

imulator Experiment

G. Müller

-Plath, J. Lehleitner, J. Maier, J. Silva-Löbling, H. Zhang, X. Zhang, & S. Zhou

Technische

Universität Berlin, Berlin, Germany

ABSTRACT: A good situation awareness (SA) of the navigator is essential for the safety of the ship, especially in

coastal areas. In this study, the Unity 3D engine was used to simulate the navigation of a coastal trading vessel

along predefined routes in the Baltic Sea. The SA of the helmsman, who was either Chinese or European, was

assessed several times with the SAGAT test (Endsley, 1995b, 20-

21) and compared between low and high

workload conditions and between manual and autopilot navigation. High workload and automated navigation

both reduced SA significantly and in an additive manner. No difference was found between Chinese and

European participants. In contrast to previous accident analyses of SA, we found that SA level 3 (projection of

future states) was most strongly affected by both factors, while SA levels 1 (perception of relevant information)

and 2 (comprehension of the current situation) suffered to a lesser extent. Further research is needed to establish

specific relationships between types of automation on ships, types of workload, and SA problems in order to

design countermeasures.

http://www.t

ransnav.eu

the

International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 17

Number 4

December 2023

DOI: 10.12716/1001.17.04.

964

even SA questionnaires (e.g. Clemente et al. 2014). In

this type of research, SA is one of several

psychological factors that determine decision making.

However, what in turn are the determinants of SA?

Endsley & Jones (2011, ch. 3) suggested eight,

mostly interacting, factors that may impair SA at

different levels: Attentional tunnelling, memory

failures, stress due to anxiety, fatigue etc., mental

overload, misplaced salience/distraction, creeping

complexity, errant mental models, out-of-the-loop

syndrome due to automation (Wickens 2002,

Parasuraman et al. 2008). Although most of them have

been identified in analyses of maritime accident

reports (see e.g. Stratmann & Boll 2016), to our

knowledge, it has not been empirically demonstrated

in the maritime domain that one or more of these

factors causally affects SA at a subthreshold level, i.e.

without a subsequent incident or accident (see the

current literature review by Fan at al., 2023). The

present pilot study is a first step towards filling this

gap

In a bridge simulator experiment, we tested the SA

of the helmsman of a coaster vessel during navigation

while manipulating two of the aforementioned

factors: high versus low mental workload, and

automated versus manual navigation. The graph in

Figure 1 visualises the proposed hypotheses in the 2 x

2 experimental design: (H1) Main effect of workload:

Averaged over the two levels of automation, high

workload was expected to reduce SA compared to

low workload, due to cognitive overload (Wickens

2002; Parasuraman et al. 2008). (H2) Main effect of

automation: Averaged over the two levels of

workload, automated routing and steering was

expected to reduce SA compared to manual, due to

the out-of-the-loop syndrome (the operator loses

vigilance because his mental presence is not needed in

operating the ship, Endsley & Kiris 1995). (H3) No

interaction: As automation reduces workload

(Parasuraman et al. 2008; Endsley & Kaber 2011),

there could be an interaction between automation and

workload such that the impact of high workload on

SA is less severe in the automated than in the manual

condition. Due to a lack of literature on this, we do

not claim such an interaction a priori.

Figure 1. Hypothesised SA of the helmsman in navigating a

simulated coaster under conditions of low and high mental

workload, and manual versus automated route planning

and steering.

2 METHOD

2.1 Participants

16 students of the Technische Universität Berlin (23-33

years of age, 8 female/8 male) participated as

volunteers. In order to ensure intercultural validity, 8

were European (fluent German-speaking), and 8 were

Chinese (native language Mandarin). All participants

had normal or corrected-to-normal vision and

hearing, no neurological impairments, no medication

that impairs driving ability, and did not possess a

recreational boat license or nautical patent. Since they

were not familiar with maritime navigation, they

conducted a training session before taking part in the

two test sessions, which are described in section 2.3

below. They were paid 40 Euros for participation.

2.2 Simulator

The sea simulator was programmed in our lab with

the Unity 3D engine (Release Unity2019.1.6f1). The

participant was shown the Point of View (POV) from

the bridge of a coaster trading vessel with overall

length 75 m, width 12 m, draught 4 m, projected onto

a screen of 2 x 3 m (see Fig. 2). At the top of the screen,

some navigation relevant data were displayed (water

depth, speed, current course in degrees, and elapsed

time). The participant was seated in a mockup ship

bridge at a distance of 3.5 m to the screen and

operated the ship’s movement with a joystick, and the

POV with a mouse. The area to be navigated was a

coastal region of the Baltic Sea, the Isefjord in

Denmark (see Fig. 3), in good weather condition. It

contained several harbours and marinas, isles and

islets, a buoyed fairway, and some narrow

anchorages. Water depth was on average 5 - 7 m, but

shallow near the coast, in harbour entrances, and at

some single spots marked with cardinal buoys. The

water depths in the simulator corresponded exactly to

those indicated in the chart with linear interpolation

in between. Other navigation aids included buoys,

harbour buildings with sailboat masts, and coastlines

with trees and cylindrical towers. Also, a variety of

other ships were visible consisting of four distinct

types: coaster, ferry, sailing boat, motor boat. A tablet

was used to simulate the ECDIS with the marked

route, distance to the next waypoint, and Automatic

Identification System (AIS) icons for other ships in

proximity (red: within a critical distance of 0.5

nautical mile (nm), green: outside a radius of 0.5 nm;

see Fig. 7).

965

Figure 2. Unity model of the coaster to be navigated through

the Isefjord in Denmark, as seen from external (above) and

from the bridge (below, point-of-view of the participants).

2.3 Procedure

Participants were tested singly in the simulator for

about two hours. Smooth communication was

guaranteed by matching the spoken language of

experimenter and participant (Mandarin or German).

As our participants were not familiar with

maritime navigation, they started with a training

session consisting of the following tasks: Steer a

course of 000 degrees (north), round a red buoy, head

for a green buoy, speed up and slow down in forward

and reverse, stop the ship, explain displays. They

were also presented with the audio signals relevant

for their workload condition (see Table 2), to which

they had to respond with the appropriate action.

Finally, a full SAGAT test (Table 1) was conducted.

In two subsequent test sessions, participants

navigated two routes that were set up in advance on

the "ECDIS" (the tablet). Each route consisted of seven

waypoints and had a total length of 7.5 nm of which

two segments of 1.5 and 2 nm each had to be actually

sailed (see Figure 3). There was a 15 minute break

between the two routes..

Situation Awareness (SA) was tested three times

along each route at unpredictable points. As a

measurement tool, we applied the Situation

Awareness Global Assessment Technique (SAGAT)

developed by Endsley (1995b, 2021). It uses the "freeze

technique", where a situation is stopped and the

participant answers questions at all three levels of SA.

Here, the simulator was set to freeze and masked,

seven questions were asked (see Table 1) and the

answers were recorded along with the true situation.

Two questions referred to SA level 1 (perception), two

to level 2 (comprehension), and two to level 3

(projection). The questions are shown in Table 1.

Figure 3. Simulated test area: Isefjord in the Baltic Sea,

Denmark, with the two routes (A and B), waypoints (WP),

the segments to be sailed, and the six SAGAT test points

(three on each route). The position of SAGAT test 6 and the

positions of preceding sound signals are marked as

examples. participants).

An additional seventh question, which was always

asked first, was a control question that had no direct

relation to ship navigation, but served to control for

the specificity of our navigation-related SAGAT: we

expected the six SAGAT questions to be affected by

automation and workload, as hypothesised in Fig. 1,

but not the control question. Some questions could

only be solved by keeping a good lookout, while

others required constant attention to the "ECDIS" or

"bridge instruments" at the top of the screen.

The automation of navigation was manipulated by

the use of an autopilot: On one route, the participant

steered the ship manually (the waypoints marked on

the "ECDIS" had to be reached, and an SOG of 8 and

10 knots respectively was recommended for the two

segments), while on the other route the ship was

navigated by the autopilot at those SOGs. Every

participant started with the same route, but the

mapping of the two routes to the two levels of

automation was randomised across participants.

Mental workload was manipulated using audio

signals indicating specific secondary tasks, which

appeared at irregular intervals before and after the

SAGAT tests. Participants had to complete the task

immediately after the audio signal. Low workload

was induced by presenting only one of two possible

audio signals, linked to simple tasks. To induce high

workload, two more signals with more complex tasks

were added and applied 2-4 times between SAGAT

tests (see Table 2). While automation of navigation

was manipulated within participants (each participant

performed both levels), mental workload was

implemented as a between-subjects factor, with n = 8

participants in the low and n = 8 in the high workload

group (4 Chinese, 4 German speaking each).

966

Table 1. SAGAT questions 1. – 6., source of information, and

SA level, as asked three times along each of the two routes.

The first question 0 served as a control question unspecific

to maritime SA.

________________________________________________

Question Source SA Level

________________________________________________

0. What man-made objects or structures lookout (control)

are visible on the coast ahead?

1. How many vessels are currently in "ECDIS" 1

close range? (AIS symbols in red)?

2. What type of vessel are they (sailing lookout 1

boat, small motorboat, coaster, ferry)?

3. What is your current position? "ECDIS" 2

(Draw in the other tablet)?

4. How much water do you currently depth 2

have under your keel (in m/cm)? display

5. What is the distance to the "ECDIS" 3

next waypoint?

6. How will you change course there? "ECDIS" 3

(Indicate in degrees and whether to

port/left or starboard/right.)

________________________________________________

Table 2. Audio signals and tasks to induce low/high

workload. Low workload: 1 signal from S1, S2 between

every two SAGAT points. High workload: 2-4 signals from

S1, S2, S3, S4 between every two SAGAT points.

________________________________________________

Audio signal Task to be executed

________________________________________________

S1 short beep, middle frequency do nothing

S2 sequence of beeps ("fuel alarm") cancel with button press

S3 sequence of 3 short beeps rotate your POV by 360°

using the mouse

S4 cell phone ring tone memorize a given 4-digit

number and recall it after

a minute (new beep)

________________________________________________

2.4 Data Analysis

At each SAGAT query (three per route), the real data

of the situation with regard to the seven questions

were recorded at the moment the participants left the

situation (the simulator was "frozen"). Dependent on

the impact of errors for safe navigation, the accuracy

of each answer was scored with 0, 1, or 2 points and

summed up across the three tests of each route. The

scores were averaged within each level and across all

levels. Thus, for each of the two routes, i.e. for the

automated and the manual navigation condition, each

participant received six scores for the six questions,

three average scores for SA level 1, 2, 3, and a total SA

score, all ranging from 0 to 6 points. The statistical

analysis of the data followed the procedure described

in the next section.

3 RESULTS

Figure 4 shows the mean total SA scores of the 8

participants of the high workload group and the 8

participants of the low workload group, with manual

and automated navigation each. Figure 5 shows the

same separately for the three levels of SA and the six

individual SAGAT questions. In order to generalise

from the small sample of participants to a population,

we conducted statistical significance tests of the two

main effects (hypotheses H1, H2) and the interaction

(H3) of the factors workload and automation in the 2 x

2 variance analytic design for each of the ten graphs

shown in Figures 4 and 5. Since hypotheses H1 and

H2 were formulated as directed effects and the factor

automation was a within-subjects factor, these

hypotheses were tested with one-tailed t-tests for two

independent samples instead of variance analytic F-

tests in order to maximise statistical power. (The

independent samples t-test on the appropriate sum or

difference scores of the within-factor, respectively, is

equivalent to testing the main effects or the interaction

in a mixed 2 x 2 ANOVA; the homogeneity of

variances is not critical here because the two groups

have equal size; Posten 1984.)

Figure 4. Mean total SAGAT scores (range 0-6) under

manual and automated navigation for the two groups of

participants with low workload (black dots, n = 8) and high

workload (white dots, n = 8). The stars indicate statistically

significant (p < 0.05) main effects.

Figure 5. Left column of graphs: Mean SAGAT Level 1,

Level 2, and Level 3 scores (range 0-6) under manual and

automated navigation for the two groups of participants

with low (black dots, n = 8) and high workload (white dots,

n = 8). Middle and right columns of graphs: Mean SAGAT

scores (range 0 – 6) for the individual questions, see Table 1.

The stars indicate statistically significant (p < 0.05) main

effects.

The results graph in Fig. 4 looks almost exactly like

the hypotheses graph (Fig. 1). The two hypotheses H1

967

and H2 were also confirmed statistically: Averaged

over the two conditions of automation, high workload

reduced SA compared to low workload (t = 2.269, df =

14, p = 0.019). Likewise, averaged over the two

workload conditions, autopilot navigation reduced

SA compared to manual navigation (t = 3.914, df = 14,

p = 0.0008). However, despite its significance, the

effect of automation does not seem very large. Further

insight is gained from a detailed look at the SA levels

and the individual questions in Fig. 5: The proposed

pattern of results was most evident only in SA level 3,

projection of perception into future states. Here,

automation and workload had a strong and

significant impact (t = 4.817, df = 14, p = 0.0001 for the

directional main effect of automation; t = 1.784, df =

14, p = 0.048 for the directional main effect of

workload), and both questions were affected. SA level

2, comprehension of the current situation, was

significantly impaired only by the high workload (t =

1.829, df = 14, p = 0.044 for the directional main effect

of workload) but not by the automated navigation (t =

0.732, df = 14, p = 0.238 for the directional main effect

of workload). This seems to be due to the fact that the

latter effect was not consistent in the two questions of

this SA level. The most basic SA level 1, perception of

relevant information, was not significantly affected by

either of the factors (t = 1.417, df = 14, p = 0.089 for the

directional main effect of automation; t = 0.73, df = 14,

p = 0.249 for the directional main effect of workload).

However, it might be worth noting that the

performance on one of the test items, the question of

how many other vessels were in close range, was

significantly impaired by the autopilot navigation (t =

2.054, df = 14, p = 0.029 for the directional main effect

of automation).

Figure 6. Mean scores (range 0 – 6) in the control question.

Symbols as in Fig. 4.

A control question was included in the SAGAT

(Table 1, first row) to check whether workload and

navigation automation affect on-board perception in

general or navigation-related situational awareness in

particular. Interestingly, as Figure 6 shows, the result

of this test item was completely different from all the

"real" SAGAT items: Buildings on shore were

discriminated even better under autopilot navigation

than under manual navigation. There was a small

positive effect of high workload. (As we had no

hypotheses for this item, no significance tests were

carried out.)

Further analyses showed that neither the native

culture of the participants, Chinese or European, nor

the mapping of automated and manual navigation on

the two routes A and B, had a statistically significant

effect on the SA results.

4 CONCLUSION

Human error, mostly due to a lack of situation

awareness (SA), is responsible for the vast majority of

all accidents at sea. Maritime human factors research

sees human error not as the end of an investigation,

but as the starting point. It is not the cause of a

problem but the effect of a deeper trouble, a sensor

indicating that something is wrong in the human-

machine system. Reason (2000) coined the famous

Swiss cheese model of human error which states that

a variety of latent and active failures by various

actors, from individual human operators to designers

and economic and legal organisations, must align like

holes in a Swiss cheese in order to produce an

accident. The small proportion of actual accidents

relative to latent unsafe human acts (accident

pyramid) is illustrated herein by the low probability

of multiple holes in a cheese being aligned.

To date, most research on the determinants of

maritime SA has focused on accident reports. This has

mostly identified problems at SA level 1 (perception).

Our research has taken the first step in a different

direction: We investigated SA problems at a

subthreshold level, where no accident occurred. In

order to causally examine the detrimental effects of

high workload and of an out-of-the-loop state due to

automated navigation, we conducted a controlled

experiment in a sea simulator, simulating coastal

navigation. We confirmed harmful effects of both

factors, with no interaction between them. That is, the

effects of workload and automation were additive,

with the worst SA under high workload and

automated navigation. The idea that automation

might compensate for workload by reducing it was

not supported. Furthermore, in contrast to the

analysis of accident reports, we found the most severe

reduction in SA not at SA level 1 (perception) but at

level 3 (projection). It is not surprising that this result

does not emerge from accident analyses: When people

are simply asked what they did or did not perceive,

only SA level 1 is addressed. Almost no one is able to

report that he or she was unable to project their

perception into the future. The SAGAT test applied

here allows a much deeper insight into the cognitive

processes underlying SA loss. Despite some

drawbacks mentioned below, Burmeister et al. (2021,

p. 22-23) pointed out that the SAGAT is the most valid

test of SA available.

968

Figure 7. Screenshot of the simulated "ECDIS / AIS" display

at one SAGAT test point ("freeze"). In the simulated AIS,

vessels in close range, defined as a radius of 0.5 nm, were

displayed as bright red icons, other vessels as green icons.

The black circle, which was not part of the original display,

shows the red icons for reasons of clarity in the black-and-

white print.

Although SA level 1 (perception) was not much

affected by our manipulation of workload and

automation, one result is worth highlighting: In our

SAGAT questions 1 and 2, participants had to indicate

how many other vessels were in close range and of

what type. This is an everyday problem in coastal

navigation, as vessels of many types (ferries, motor

yachts, sailing boats and other fishing or trading

vessels) frequently cross fairways and harbour

entrances. In our simulation, the AIS icons of vessels

in close range were displayed in bright red (Fig. 7).

We found a significant decrease in Question 1 scores

when sailing on autopilot compared to manual. We

argue that this effect was indeed a consequence of the

out-of-the-loop syndrome, as our control question

(indicate how many man-made structures are visible

on shore) was answered even better when sailing on

autopilot.

There are two limitations to our study, apart from

the fact that it is only a simulation. First, the SAGAT

technique for measuring SA suffers from the problem

that it has to be repeated several times to give reliable

results, but then the participant knows the questions

to be asked and can prepare mentally. However, this

implies that participants have an unrealistically good

SA because of this preparation, and in reality SA

problems may be even worse. This problem can only

be addressed with a much larger number of

participants, each of whom takes only one SAGAT.

Secondly, our participants were students, not

professional officers at sea. However, in our previous

research on maritime navigation in the simulator and

at sea, we found no systematic differences between

the two populations (Müller-Plath et al. 2018; Müller-

Plath 2019), which is probably due to fundamental

laws of human-machine interaction being involved.

Future research should firstly validate the results

with experts in maritime navigation and a larger

sample. Second, it should extend the line of study

outlined here and investigate more specific questions

such as what kind of automation affects what kind of

perceptions, understandings, and projections in which

way? And thirdly, how can training or tools be

designed to counteract the loss of SA, as simply more

automation does not seem to be the "silver bullet"?

ACKNOWLEDGEMENTS

The authors would like to thank Richard Gross for

programming the sea simulator and solving many technical

problems in a short time.

REFERENCES

Aylward, K., Weber, R., Lundh, M., MacKinnon, S.N., &

Dahlman, J. 2022). Navigators’ views of a collision

avoidance decision support system for maritime

navigation. The Journal of Navigation, 1-14.

Butler, G.L., Read, G.J., & Salmon, P.M. 2022.

Understanding the systemic influences on maritime pilot

decision-making. Applied Ergonomics, 104, 103827.

Burmeister, H.-C., Grundmann, R., Hochgeschurz, S.,

Hohnrath, P., Ujkani, A. 2021. Increasing Maritime

Situational Awareness by Augmented Reality Solutions.

Hamburg, Germany: Fraunhofer Center for Maritime

Logistics and Services CML

Butt, N., Johnson, D., Pike, K., Pryce-Roberts, N., & Vigar,

N. 2013. 15 Years of Shipping Accidents: A review for

WWF. Southampton Solent University.

http://awsassets.panda.org/downloads/15_years_of_ship

ping_accidents_a_review_for_wwf_.pdf. Last access 08

April 2023.

Clemente, S., Loia, V., & Veniero, M. 2014. Applying

cognitive situation awareness to collision avoidance for

harbour last-mile area safety. Journal of Ambient

Intelligence and Humanized Computing, 5, 741-745.

European Maritime Safety Agency (EMSA) 2022. Annual

Overview of Marine Casualties and Incidents 2021.

https://www.emsa.europa.eu/we-do/safety/accident-

investigation/download/6955/4266/23.html. Last access

08 April 2023.

Endsley, M.R.. 1995a. Toward a theory of situation

awareness in dynamic systems. Human Factors 37(1):

32-64.

Endsley, M.R. 1995b. Measurement of situation awareness

in dynamic systems. Human Factors 37(1): 65-84.

Endsley, M.R. 2021. A systematic review and meta-analysis

of direct objective measures of situation awareness: A

comparison of SAGAT and SPAM. Human Factors,

63(1), 124-150.

Endsley, M.R. & Jones, D.G. 2011: SA Demons: The Enemies

of Situation Awareness. In: Designing for Situation

Awareness: An Approach to User-Centered Design, ch.

3, pp. 31–41. Boca Raton: CRC Press.

Endsley, M.R. & Kaber, D.B. 1999. Level of automation

effects on performance, situational awareness and

workload in a dynamic control task’, Ergonomics, vol.

42, no. 3, pp. 462-492.

Endsley, M.R. & Kiris, E.O. 1995. The Out-of-the-Loop

Performance Problem and Level of Control in

Automation. Human Factors, vol. 37, no. 2, pp. 381–394.

Fan, S., Blanco-Davis, E., Fairclough, S., Zhang, J., Yan, X.,

Wang, J., & Yang, Z. 2023. Incorporation of seafarer

psychological factors into maritime safety assessment.

Ocean & Coastal Management, 237, 106515.

Grech, M. & Horberry, T. 2002. Human error in maritime

operations: Situation awareness and accident reports.

Paper presented at the 5th International Workshop on

Human Error, Safety and Systems Development,

Newcastle, Australia.

Grech, M. R., Horberry, T. J., & Koester, T. 2008. Human

factors in the maritime domain. Boca Raton, FL: CRC

Press

Kartoglu, C., Senol, Y.E., & Kum, S. 2022. Situational

awareness of navigators in high-speed craft bridge

navigation operations. Australian Journal of Maritime &

Ocean Affairs, 1-15.

Kizilay, F.E., Arslan, O., Akyuz, E., & Kececi, T. 2023.

Prediction of human error probabilitiy for officers

969

during watchkeeping process under SLIM approach.

Australian Journal of Maritime & Ocean Affairs, 1-18.

Müller-Plath, G., Jung, D. & Müller, M. 2018. Research-

based design and usability guidelines for electronic

charting systems (ECS) in yachting and boating.

International Journal of e-Navigation and Maritime

Economy, 10, 32-48.

Müller-Plath, G. 2019. How does digital navigation on

sailboats affect spatial abilities at sea?. TransNav:

International Journal on Marine Navigation and Safety

of Sea Transportation, 13(2), 299-310.

Parasuraman, R., Sheridan, T.B., & Wickens, C.D. 2008.

Situation awareness, mental workload, and trust in

automation: Viable, empirically supported cognitive

engineering constructs. Journal of cognitive engineering

and decision making, 2(2), 140-160.

Posten, H. O. (1984). Robustness of the two-sample t-test. In:

D. Rasch & M.L. Tiku (eds.). Robustness of statistical

methods and nonparametric statistics: 92-99. Dordrecht,

Holland: Kluwer Academic Publishers Group.

Reason, J. (2000). Human error: models and management.

BMJ, 320, 768-770.

Stratmann, T. C. & Boll, S. 2016. Demon Hunt-The Role of

Endsley’s Demons of Situation Awareness in Maritime

Accidents. In 6th International Conference on Human-

Centred Software Engineering (HCSE)/8th International

Conference on Human Error, Safety, and System

Development (HESSD) (pp. 203-212). Springer

International Publishing.

Wickens, C. D. (2002). Situation awareness and workload in

aviation. Current directions in psychological science,

11(4), 128-133.

William, H. H. 1959. Industrial accident prevention: A

scientific approach. Boston: McGraw-Hill Book

Company.