International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 1

Number 3

September 2007

243

AIS Contribution in Navigation Operation-

Using AIS User Satisfaction Model

A. Harati-Mokhtari

Liverpool John Moores University, UK & Chabahar Maritime University, Iran

P. Brooks, A. Wall & Jin Wang

Liverpool John Moores University, UK

ABSTRACT: AIS was introduced in 2002 and its phased implementation programme completed in 2004.

Problems still exist in its reliable use for navigational operation. Our paper is part of a wider evaluation of

AIS. This paper considers the users view of AIS and we have attempted to measure the extent of navigators’

satisfaction with AIS in their navigation activities by using an AIS User Satisfaction Model. This paper

evaluates the validity of the AIS User Satisfaction Model using questionnaire data as a suitable structure for

measuring the degree of navigators’ satisfaction and usage of AIS, and probably applies for other similar

technologies. This, in turn, could help to determine the measures that need to be adopted in order to improve

quality and use of AIS as an effective navigation and anti-collision tool.

1 INTRODUCTION

Introduction of Automatic Identification System

(AIS) in marine industry was aimed at promotion of

efficiency and safety of marine navigation. Its

mandatory phased implementation programme

completed in December 2004, and consequently all

SOLAS Convention vessels should have installed

the equipment on the bridge by this date. Results of

AIS data studies focused on the accuracy of the

information transmitted by AIS, carried out at

Liverpool John Moores University (Harati-Mokhtari

et al, 2007), revealed that data provided by AIS are

not reliable in many cases, especially the data

entered into the equipment manually. Human

failures were observed at different levels in AIS

application for navigation, which are:

− Failures by frontline operator

− Installation Failures

− Design failures

− Training and management failures

− Regulatory failures

Therefore, the AIS could not wholly be trusted,

and AIS usage and its data quality may even be

deteriorated furthermore.

The reliable operation of AIS in early stages of

introduction and correct implementation strategy are

important concerns that could influence navigators’

impressions, attitude, and behaviour toward their

acceptance and future use of the system. The aim of

this paper develops a suitable model for evaluation

of AIS usage for navigation operation, particularly

for anti-collision, by navigators on the ship’s bridge.

This study examines influence of some important

factors in navigators’ satisfaction with AIS

technology that could affect its actual improved

usage for the intended purposes in navigation.

2 TECHNOLOGY USAGE BEHAVIOUR AND

IMPLEMENTATION ENVIRONMENT

Technology implementation may be based on

voluntarily or mandatory adoption. Voluntarily

adoption is a situation where adoption and use of

244

technology is not obligatory and determined by the

user’s optional preference. The opposite is

mandatory adoption is a situation where adoption

and use of a system is directed from higher level

than the user. In mandatory adoptions users are

obliged to use the system to perform their job

(Brown, et al, 2002). According to Adamson and

Shine (2003), even in mandatory adoption and use of

technology, some users may not comply with such

mandate if they believe that the system is not

satisfactorily supporting their work tasks, or they

may use it as their only available choice but with a

negative job satisfaction result.

3 AIS USER SATISFACTION MODEL

(AISUSM)

Identifying appropriate functions and characteristics

of new technology, such as AIS, will help in

delivering the accurate and useful system to its end

users. Such identifications could be carried out by

evaluation of the acceptability of the system by the

user through understanding his responses and

satisfaction level in system use. By identifying

appropriate functions and characteristics demanded

from AIS, required modifications could be made to

the system and its implementation strategy.

According to Venkatesh (2000), a significant

progress has been made recently in explaining and

predicting users acceptance of technology at work,

especially information technology. Most of commonly

used theories and models of technology acceptance

by end user have examined the acceptance of

technology in voluntarily environment where adoption

and use of technology are based on volitional choices,

and only few of them have considered technology

adoption and use in mandatory environment.

However, End User Satisfaction Model (EUS)

(Adamson and Shine, 2003) is considered to be

suitable model for measuring system satisfaction in

mandatory environment. It was argued (Venkatesh,

2003) that the role of social influence of subjective

norm (according to Ajzen and Fishbein (1980),

subjective norm is the end user’s belief about how

other people would view him/her if he performed the

behaviour) is only important in initial stages of

introduction of technology when user experience

with technology is at low levels but it eroded over

time and finally become insignificant with continued

usage. Since the AIS has mandatory been used on all

SOLAS Convention vessels from end of December

2004, the influence of subjective norm in this study

is insignificant.

Therefore AIS User Satisfaction Model is adopted

from EUS Model for assessment of navigators’

satisfaction with AIS without consideration of

subjective norm (figure 1).

Fig. 1. AIS User Satisfaction Model (adapted from Adamson

and Shine, 2003)

4 METHODOLOGY

A questionnaire that was already designed to assess

the navigators’ perception about different aspects of

the AIS will be used to analyse validity of the AIS

User Satisfaction Model for explaining and

predicting navigators’ acceptance of AIS at work.

Apart fro demographic factors, there were 39 other

items included in the questionnaire that are grouped

to fit into the AISUS Model. The relevant groups

are:

− System Quality (SQ)- The degree to which end

user believes on ease of retrieving data, the

system’s response time, accuracy and reliability

(Adamson and Shine, 2003).

− Self-Efficacy (S-E)- the level of individuals’

beliefs on their ability to perform specific tasks

successfully, with consideration of the degrees of

efforts required in challenging situations

(Adamson and Shine, 2003).

− Perceived Usefulness (PU)- the degree to which

an individual believes that using a particular

system will enhance his/her job performance and

productivity (Davis, 1986).

− Perceived Ease of Use (PEOU)- the degree to

which an individual believes that using a

particular system will be free of effort (Davis,

1986).

− AIS User Satisfaction (AISUS)

The analysis will be carried out with the use of

the computer software SPSS version 14. Cronbach’s

alpha coefficient (α), as one of the most commonly

used indicators of the scale reliability (Pallant. 2004,

and Field, 2005), will be used for analysing internal

consistency of the measurement. Pallant (2004) and

Field (2005) stated that Cronbach’s alpha ranges in

value from 0 to 1, and values of above 0.7 are

acceptable values of alpha, but higher the score, the

more reliable the generated scale is. The construct

validity (relationship among variables) will be

explored through statistical technique of multiple

regression, which according to Tabachnick and

245

Fidell (2000), Pallant (2004), and Field (2005) is

used as a popular technique that can deal with

variety of questions, especially in predicting a

dependent variable (DV) from several continuous

independent variables (IV), in many disciplines. The

goal of regression in this research is to arrive at a set

of regression coefficients (β values) for the IVs.

Tabachnick and Fidell (2000) further pointed out

“regression analysis would only reveal the

relationships among variables but do not indicate

causality of the relationships”. Therefore, since our

data (one sample) are normally distributed multiple

regression is considered to be the most suitable

technique for our analysis.

4.1 Data manipulation

Scores for negatively worded items (high scores

indicate low satisfaction) were reversed, and total

scales scores were calculated for the model

measurement constructs. The total scored named as

TSQ, TSE, TPU, TPEOU, and TAISUS.

5 ANALYSIS

5.1 Data manipulation

Five of the items were found to have low reliability

figures (Cronbach’s alpha less than 0.7) and they

were dropped from the final analysis. The remaining

items were all reliable (with alpha value > 0.7). The

total scale was reliable with alpha values of 0.8.4,

further scales for TSQ, TSE, TPU, TPEOU, and

TAISUS all were reliable with alpha values of 0.74,

0.71, 0.77, 0.74, and 0.70, respectively.

The distribution of the total scores for five

variables examined by relevant histograms, and

further crosschecked by calculating z-scores. The

data were normally distributed.

5.2 Final analysis

Final data analysis is carried out for each of the five

individual sub scales, and the results are given in

following sub-sections.

5.2.1 Correlations

According to the value of Pearson correlation

coefficient in correlation matrix, both of the

TSQ and TSE scales correlate positively with TPU

(R = 0.504, p < 0.001 and R = 0.380, p < 0.001,

respectively). But TSQ has a larger positive

correlation with TPU, than TSE. Thus it is likely that

TSQ will best predict TPU. Apparently, there is not

any correlation between TSQ and TSE (R= -0.049).

One-tailed significance values show that both the

positive correlations of TSQ with TPU and TSE with

TPU are very significant as

p < 0.001.

Pearson correlation coefficients also show that

TSQ correlates substantially with TPEOU (R = 0.360,

p < 0.001), but TSE has a smaller positive correlation

with TPEOU (R = 0.132, p < 0.001) than TSQ.

Although TSE had a lower positive correlation with

TPEOU, it is still significant in predicting TPEOU.

Bivariate Correlation between TSQ and TSE is -

0.049. One-tailed significance values indicates that

both the correlations between TSQ and TPEOU, and

between TSE and TPEOU are positive and very

significant, p <0 .001.

Pearson correlation coefficient for both the scales

of TPU and TPEOU are above 0.3, (R = 0.543, p <

0.001, and R = 0.311, p < 0.001, respectively) which

show important correlations with TAISUS. But TPU

has a larger positive correlation with TAISUS, than

TPEOU. Bivariate correlation between TPU and

TPEOU is 0.407 and bellow maximum limit of 0.9.

One-tailed significance values indicate that positive

correlations are very significant (p < 0.001) in both

the cases.

5.2.2 Evaluation

Model summary for Total Perceived Usefulness

of the AIS shows that 41.8% (R squared = 0.418) of

the variance in TPU is explained by the model,

which includes the TSQ (R squared = 0.254) and

TSE (R square = .164). Adjusted R squared is 0.399

and the shrinkage is equal to 1.9% =

100)399.0418.0( ×−

. Therefore, the percentage of

the variance explained by the model for TPU is very

close to that of the corrected estimate of the true

population.

Result of the analysis of variance (ANOVA)

shows that the improvements due to the regression

models are much grater than inaccuracy within the

models (the F-ratios are 22.099 and 22.939). This is

unlikely to have happened by chance as both of the

F-ratios are very significant with probabilities of <

0.001. Therefore the model is a significant fit of the

data overall and it significantly improves our ability

to predict the outcome variable because the F-ratio

is significant (probability less than 0.05).

Model summary also indicates that 15.2% (R

square = 0.152) of the variance in TPEOU is

explained by the model, which includes the TSQ (R

square = 0.129) and TSE (R square = 0.023).

Adjusted R square is 0.125 and the shrinkage is

equal to 2.7% =

100)125.0152.0( ×−

, which shows

that the percentage of the variance explained by the

246

model is very close to that of the corrected estimate

of the true population.

According to ANOVA both the F-ratio (F = 9.667,

p < 0.003 and F = 5.729, p < 0.005) are significant

and unlikely to have happened by chance. This

indicates that the improvement due regression model

is greater than inaccuracy within the model.

Therefore, the ability to predict the outcome variable

will be significantly improved by the model, and the

model is a significant fit of the data overall due

to the significant F-ratio (significance value is less

than 0.05).

The result also shows that 30.4% (R square =

0.304) of the variance in TAISUS is explained by the

model. This includes the TPU (R square = 0.294),

and TPEOU (R square = 0.010). Adjusted R square

is 0.292, which shows shrinkage of

1.2% =

100)292.304(. ×−

. This means that the

percentage of the variance explained by the model in

not so much away from the corrected estimate of the

true population. TPU causes R

2

to change from zero

to 0.294, which this change in the amount of

variance explained gives rise to a significant F-ratio

of 46.751 with a probability of less than 0.001.

Addition of TPEOU causes R

2

to increase by 0.010,

and the change in the amount of variance that it can

explain gives rise to an F-ratio of 1.5430, which is

not significant with a probability less than 0.217.

According to ANOVA, both the F-ratio for model 1

(F = 46.751), and F-ratio for model 2 (F = 24.261)

are very significant (p < 0.001 for both the cases),

and therefore, it is unlikely to have happened by

chance. These results show that both models 1 (with

TPU as the independent variable) and model 2 (with

addition of TPEOU as second independent variable)

are significant fit of the data overall, and they

significantly improves our ability to predict the

outcome variable, because the F-ratios are

significant (probability less than 0.05).

5.2.3 Model parameters

Summary of the regression model indicates that

the TSQ, with standardised beta of 52.3%, makes a

stronger unique contribution in explaining TPU,

when the variance explained by the TSE is

controlled for. The standardised beta value for TSE

is showing a less contribution with 40.5%. Further,

TSQ and TSE both with a significance value of

0.001 are making a unique, and statistically very

significant, contribution to the prediction of the TPU

scores. This also means no overlap between TSQ

and TSE.

Confidence interval for TSQ is between 0.452

and 0.972, and for TSE is between 0.148 and 0.411,

which both are relatively narrow, and do not cross

Zero. This indicates that the parameters for these

variables are significant and they have positive

relationships.

Further, the zero-order correlations are 0.504 for

TSQ and 0.380 for TSE. The part correlation

coefficients are 0.523 for TSQ and 0.405 for TSE,

indicating that TSQ uniquely explains 27% (0.523

2

)

and TSE uniquely explains 16% (0.405

2

) of the

variance in TPU scores.

In the case of TPEOU, TSQ with β value of 0.367

has 36.7% a unique contribution in explaining

TPEOU, when the variance explained by the TSE is

controlled for. The TSE with β value of 0.150 has

less contribution with 15.0%. The results also

indicate that TSQ with a significance value of 0.002

making a unique, and statistically very significant,

contribution to the prediction of the TPEOU scores.

However contribution of TSE with significance

value of 0.198 is not significant that may be due to

some degrees of overlap between TSQ and TSE.

Confidence interval for TSQ is between 0.219

and 0.955, which is relatively narrow and does not

cross zero. Confidence interval for TSE is between -

0.065 and 0.308, which is narrow but it does cross

zero. This indicates that only the parameters for TSQ

are significant, and it has a positive relationship, but

the parameters for TSE are not significant and it has

a negative relationship.

The zero-order correlation for TSQ is 0.360 and

for TSE is 0.132. These values correspond to the

same values of the Pearson correlation coefficients.

TSQ (with part correlation coefficients of 0.367),

and TSE (with part correlation coefficients of 0.150)

each uniquely explain 13.5% (0.367

2

), and 2.3%

(0.150

2

) of the variance in TPEOU scores,

respectively, when the effects of the other predictors

on the outcome are controlled for.

TPU with β value of 49.9% makes a stronger

unique contribution in explaining TAISUS, when the

variance explained by the TPEOU is controlled for.

The standardised beta value for TPEOU is only

showing a contribution of 10.8%. TPU with a

significance value of 0.001 is making a unique and

very significant contribution to predict TAISUS

scores. But TPEOU with significance value of 0.217

does not make such a unique and statistically

significant contribution to TAISUS scores

prediction, which may be due to some overlap

between TPU and TPEOU.

Confidence interval for TPU is between 0.449

and 0.921, which is relatively narrow and does not

cross zero. The range of confidence interval for

TPEOU is between -0.75 and 0.325, which despite

being narrow, it crosses zero. These confidence

intervals indicate that the parameters for TPU are

247

significant, but the parameters for TPEOU are not

significant and they do not have positive

relationships.

The zero-order correlations (TPU = 0.543, and

TPEOU = 0.311) again correspond to the Pearson

correlation coefficients. The part correlation

coefficients for TPU (0.456) and for TPEOU (0.098)

indicate that TPU uniquely explains about 21%

(0.456

2

) and TPEOU could only uniquely explains

less than 1% (0.098

2

) of the variance in TAISUS

scores, when the effect of the other predictor on the

outcome are controlled.

5.2.4 Multicollinearity assessment

In the case of TPU, the lowest tolerance value is

0.998, which is not less than 0.10. The highest

Variance Inflation Factor (VIF) value is 1.002,

which is well below the critical value of 10. The

tolerance and VIF values confirm that collinearity is

not a problem for this model, and therefore, the

variability of TPU is properly explained by the TSQ

and TSE.

The eigenvalues of the scales are between 2.95

and 0.006, which are fairly close, and condition

index of the final dimension is 22.32, which is not

very large compared to other dimensions. The

variance proportions show that for TSQ highest

percentage of its variance proportion (92% of the

variance of the regression coefficient) is associated

with eigenvalue number 3, and for TSE highest

percentage of its variance proportion (89% of the

variance of the regression coefficient) is associated

with eigenvalue number 2. These data further

indicate that multicollinearity is not a problem in this

model.

In the case of TPEOU, the lowest tolerance value

is 0.998, which is more than 0.10. The highest VIF

value is 1.002, which is well below 10. These values

of tolerance and VIF confirm that the problem of

multicollinearity is not an issue for this model, and

therefore, the variability of TPEOU is properly

explained by the TSQ and TSE.

In addition, the collinearity diagnostics data

shows that the eigenvalues of the scales are between

2.95 and 0.006, which are fairly close. Condition

index of the final dimension is 22.32, which is not

very large compared to other dimensions. The

variance proportions show that for TSQ 92% of the

variance of the regression coefficient is associated

with eigenvalue number 3, and for TSE 89% of the

variance of the regression coefficient is associated

with eigenvalue number 2, which is a sign of no

multicollinearity.

The lowest Tolerance value is 0.834 (more than

0.10), and the highest VIF value is 1.199 (well

below 10). These show that multicollinearity is not a

problem for this model in prediction TAISUS.

In addition, the eigenvalues of the scales are

between 2.974 and 0.010, which are reasonably

close. Condition index of the final dimension is

17.026, which in comparison to other dimensions is

not very large. The variance proportions show that

the highest percentage (80%) of TPU variance

proportion is associated with eigenvalue number 3,

and the highest percentage (100%) of TPEOU

variance proportion is associated with eigenvalue

number 2. These data indicate no multicollinearity.

5.2.5 Casewise diagnostics

Casewise diagnostics result for TPU shows that

out of 116 cases only 3 cases (about 3%) are with

standardised residuals outside the limits. Therefore,

appears that there is not a big difference between

outcome of the sample and the outcome of the

model, and the model is reasonably accurate.

Casewise diagnostics result for TPEOU indicates

that out of 116 cases only 2 cases (less than 2%) are

with standardised residuals outside the limits of

±

2.

Therefore, our sample appears to conform to the

expectation of a reasonably accurate model.

Finally the result of casewise diagnostics for

TAISUS indicates that out of 116 cases only 3 cases

(less than 3%) are with standardised residuals

outside the limits of (

±

2). This means that about

97% of the cases are with standardised residuals

within the limits, and therefore, our sample is

reasonably accurate.

6 DISCUSSION AND CONCLUSION

6.1 Internal consistency

Preliminary data analysis showed that scores of the

grouped questionnaire items, after dropping five

items with low reliability from the analysis, were

normally distributed. The remaining 34 items

included in the questionnaire, for the final analysis

according to AIS user satisfaction model, had a

reliable total scale with an overall Cronbach’s alpha

of 0.804. Reliability figures for makeup items in

model variables were 0.739 for system quality, 0.711

for system self-efficacy, 0.769 for perceived

usefulness, 0.737 for perceived ease of use, and

0.704 for AIS user satisfaction, which are within

acceptable limit.

248

6.1.1 Implications

Pearson correlation coefficients (R) are used to

test relationship between the attitudinal forming

variables of System Quality and Self-efficacy, and

the sample’s Perceived Usefulness and Ease of Use

of AIS. The results are as follows:

SQ: PU (R = 0.504, P < 0.001, 1-tailed)

SE: PU (R = 0.380, P < 0.001, 1-tailed)

SQ: PEOU (R = 0.360, P < 0.001, 1-tailed)

SE: PEOU (R = 0.132, P < 0.140, 1-tailed)

Results show that both the System Quality and

Self-efficacy have a statistically very significant and

positive relationship with Perceived Usefulness.

About Perceived Ease of Use, only System Quality

has a significantly positive relationship with

Perceived Ease of Use. But the positive relationship

of Self-efficacy with Perceived Ease of Use is not

statistically significant (P > 0.05). The strongest

relationship is between SQ and PU with R = 0.504,

and the weakest relationship is between SE and

PEOU with R = 0.132. The relationships show that

the System Quality is strongly related with AIS

Perceived Usefulness and its Perceived Ease of Use.

The results of Pearson correlation coefficients (R)

for perceptual variables of Perceived Usefulness,

Perceived Ease of Use, and AIS User Satisfaction

are as follows:

PEOU: PU (R = 0.407, P < 0.001, 1-tailed)

PU: AISUS (R = 0.543, P < 0.001, 1-tailed)

PEOU: AISUS (R = 0.311, P < 0.001, 1-tailed)

The above correlation coefficients show positive

and statistically very significant relationships between

the PU, PEOU and AISUS. It also can be seen that

there is a relatively strong bivariate relationship

between PU and PEOU. The relationship between

PU and AISUS is stronger than the relationship

between PEOU and AISUS. This means that if the

AIS users perceive that the implemented AIS

technology is useful and easy to use then they are

likely to be satisfied with the system, and therefore,

they more frequently use the AIS for navigational

activities.

Path analysis of the model is drawn in figure 2 to

show the importance of influence of different

variables in predicting dependent variable in AIS

User Satisfaction Model. The diagram includes

standardised beta coefficients (β), which shows the

strength of influence of each predictor variable on

the criterion variable according to the measurement

constructs of the model.

Fig. 2. Path Analysis of the AIS User Satisfaction Model

Path analysis of the AIS User Satisfaction Model,

figure 2, demonstrates that:

Unique influence of each one of the independent

variables on predicting Perceived Usefulness, when

variance explained by other variable is controlled

for, is 52.3% for AIS System Quality and 40.5% for

navigators’ Self-efficacy. These unique importances

of variables in predicting AIS Perceived Usefulness

are both very significant with a probability of 0.001

and without any overlap between them.

Unique influence of each one of the independent

variables on predicting Perceived Ease of Use, when

variance explained by other variable is controlled

for, was 36.7% for AIS System Quality and 15.0%

for navigators’ Self-efficacy. The unique importance

of the System Quality in predicting AIS Perceived

Ease of Use is very significant with a probability of

0.002, but this unique importance is not significant

for navigators’ Self-efficacy (P = 0.195, which is

more than 0.05). There is possibility of overlap

between System Quality and Self-efficacy.

Unique influence of each one of the independent

variables on predicting Perceived AIS User Satisfac-

tion, when variance explained by other variable is

controlled for, is 49.9% for AIS Perceived

Usefulness and 10.8% for AIS Perceived Ease of

Use. The unique importance of the Perceived

Usefulness in predicting AIS User Satisfaction is

very significant with a probability of 0.001, but the

unique importance of Perceived Ease of Use is not

significant (P = 0.217, which is more than 0.05).

Some degrees of overlap might exist between

Perceived Usefulness and Perceived Ease of Use.

Confidence intervals show that the parameters for

AIS System Quality and navigators’ Self-efficacy in

predicting Perceived Usefulness are significant with

positive relationships. According to part correlation

values, AIS System Quality uniquely explains 27%,

and navigators’ Self-efficacy 16% of the variance in

Perceived Usefulness of the AIS for navigation. A

shrinkage of 1.9% shows that the difference in

percentage of the variance in AIS Perceived

249

Usefulness explained by the model and the corrected

estimate of the true population is very low. The

result shows that model was a good fit and it

significantly improves prediction of Perceived

Usefulness.

Parameters for AIS System Quality in predicting

Perceived Ease of Use is very significant with

positive relationships, but parameters for self-

efficacy in predicting Perceived Ease of Use are not

significant and with negative relationships. AIS

System Quality uniquely explains 13.5%, and

navigators’ Self-efficacy 2.3% of the variance in

Perceived Ease of Use of the AIS for navigation. The

difference in percentage of the variance in AIS

Perceived Ease of Use explained by the model and

the corrected estimate of the true population is 2.7%.

The result also shows that model is a significant fit

of the data overall for Perceived Ease of Use.

Parameters for AIS Perceived Usefulness in

predicting AIS User Satisfaction are significant with

positive relationships. But parameters for Perceived

ease of use in predicting AIS User Satisfaction are

not significant and with negative relationships. AIS

Perceived Usefulness uniquely explains 21%, and

AIS Perceived Ease of Use uniquely explains less

than 1% of the variance in Perceived AIS User

Satisfaction for marine navigation. The difference in

percentage of the variance in Perceived AIS User

Satisfaction explained by the model and the

corrected estimate of the true population is 1.2%.

The result also shows that the model is a significant

fit of the data overall for Perceived Ease of Use.

The model shows significant goodness-of-fit in

predicting the Perceived AIS User Satisfaction.

It is also observed that the problem of multicollin-

earity due to perfect or strong correlation between

independent variables does not exist in the model.

Therefore, the regression coefficients are uniquely

estimated in the model. Casewise diagnostics shows

that the regression models are reasonably accurate as

the maximum percentage of the cases with

standardised residuals outside the limits is 3%.

Therefore, there is not a big difference between

outcome of the sample and outcome of the model.

The path analysis (figure 2) shows that the there

is not a significant unique influence of the

navigators’ Self-efficacy on predicting Perceived

Ease of Use. It is also revealed that the unique

influence of Perceived Ease of Use is not significant

on the AIS User Satisfaction. But a unique influence

from navigators’ Self-efficacy on the Perceived

Usefulness was observed in the model, which is not

included in the original model.

REFERENCES

Adamson, I., and Shine, J. (2003) Extending the New

Technology Acceptance Model to Measure the End User

Information Systems Satisfaction in a Mandatory

Environment: A Bank’s Treasury. Technology Analysis &

Strategic Management, Vol. 15, No. 4, p. 441-455.

Brown, S.A., Massey, A.P., Montoya-Weiss, M..M., and

Burkman, J.R. (2002) Do I Really Have to? User

Acceptance of Mandated Technology. European Journal of

Information Systems, Vol. 11, No. 4, p. 283-295.

Davis, F. D. (1986) A Technology Acceptance Model for

Empirically Testing New End-User Information Systems:

Theory and Results. Ph.D. Dissertation, MIT Sloan School

of Management, Cambridge, MA.

Harati-Mokhtari A, Wall A, Brooks P & Wang J (2007) AIS:

Data reliability and human error implications., to be

published for publication in the Journal of Navigation,

September 2007, Vol. 60, No. 3.

Venkatesh, V. and Davis, F. D. (2000) A Theoretical Extension

of the Technology Acceptance Model: Four Longitudinal

Field Studies. Management Science, Vol.46, No. 2, p.186-204.

Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D.

(2003) User acceptance of information technology: Toward

a unified view. MIS Quarterly, Vol. 27, No. 3, p. 425-478.