4
going through a consent statement, after which they went
through the various sections of the questionnaire.
Of the 200 participants, 82% are male, and 15% are
female, 61% White, 8% Asian, 14% Black or African, 15%
Hispanic and 2% of another ethnicity. As to age, 44% of
the participants are less than 30 years old, 38% are 30 to 44
years, 11% are 45 to 54 years, and 6% of the participants
are 55 years or older. For their level of education, 58% of
participants have a high school degree, 21% have a col-
lege degree, and 11% of the participants have a bachelor’s
degree. In terms of marital status, 69% were single while
23% were married. For their income, about 21% receive
less than $20,000, 20% receive from $20,000 to $40,000,
while 18% of them receive from $40,000 to $60,000.
Measure of Safety and Trust
The study was set to measure participants perceived safety
and trust. To measure their perceived safety, they were asked
to indicate on a 10-point Likert scale how safe they think
underground mines are, how safe they feel after receiving
each of the two alert messages.
An internal reliability analysis was carried out using the
Cronbach alpha method to assess the degree to which our
trust and safety questions were all measuring the same thing.
The reliability test indicated that, if the question seeking
participants initial safety perception about underground
mine and an attention seeking question are removed, the
reliability factor will increase to about 86%. The remain-
ing 12 questions were averaged to obtain a single score for
perceived safety and trust.
Data Analysis Approach
To test our hypotheses, we analyzed the effect of the vari-
ous independent variables on the safety and trust score. The
factors that we hypothesized will affect this score included
the system type (AI or Human), the amount of information
(low or high level) about the gas monitoring system given
to the participant, and other demographic factors that have
been studied by other researchers [40, 37, 5] and seen to
affect decision making and perceived safety, were captured
as independent variables.
RESULTS
The results of the ANOVA test for significance are pre-
sented on Table 1 while Figure 1 shows the effects of system
type and information level on safety and trust scores. From
the results, system type had no effect on trust, F(1, 195) =
0.003, p =0.957, eta2 =2.08×10–5 but higher informa-
tion led to more trust, F(1, 195) =4.733, p =0.031 and
eta2 =2.37×10–2 and there was no interaction between the
two, F(1, 195) =0.141, p =0.708 and eta2 =7.06 × 10–4.
System type did not have a significant impact on partici-
pants safety and trust perception (p =0.957), which implies
that, in terms of trust and safety, knowing the specific type
(AI vs human) of gas monitoring system did not have much
influence.
Effect of Information Level on Safety and Trust
Figure 1 presents the results of the ANOVA test for the
influence of the amount of information provided to partici-
pants on their perceived safety and trust. It shows that pro-
viding participants with more information about the system
that sends out the alert message increases their safety and
trust perception. The amount of system information did
have a significant impact on participants’ safety and trust
perception (p =0.031), and it is worth noting that partici-
pants who received more information had a higher score
than participants who received less information, implying
increasing the amount of information increases participants
safety and trust.
Effect of Demographics on Safety and Trust
All demographic variables were entered into a linear regres-
sion model to reveal their effect on perceptions of safety:
gender, ethnicity, age, education level, marital status,
income level, and number of children. Backward selection
Figure 1. Effect of System Type and Information Level on
Trust and Safety Score
going through a consent statement, after which they went
through the various sections of the questionnaire.
Of the 200 participants, 82% are male, and 15% are
female, 61% White, 8% Asian, 14% Black or African, 15%
Hispanic and 2% of another ethnicity. As to age, 44% of
the participants are less than 30 years old, 38% are 30 to 44
years, 11% are 45 to 54 years, and 6% of the participants
are 55 years or older. For their level of education, 58% of
participants have a high school degree, 21% have a col-
lege degree, and 11% of the participants have a bachelor’s
degree. In terms of marital status, 69% were single while
23% were married. For their income, about 21% receive
less than $20,000, 20% receive from $20,000 to $40,000,
while 18% of them receive from $40,000 to $60,000.
Measure of Safety and Trust
The study was set to measure participants perceived safety
and trust. To measure their perceived safety, they were asked
to indicate on a 10-point Likert scale how safe they think
underground mines are, how safe they feel after receiving
each of the two alert messages.
An internal reliability analysis was carried out using the
Cronbach alpha method to assess the degree to which our
trust and safety questions were all measuring the same thing.
The reliability test indicated that, if the question seeking
participants initial safety perception about underground
mine and an attention seeking question are removed, the
reliability factor will increase to about 86%. The remain-
ing 12 questions were averaged to obtain a single score for
perceived safety and trust.
Data Analysis Approach
To test our hypotheses, we analyzed the effect of the vari-
ous independent variables on the safety and trust score. The
factors that we hypothesized will affect this score included
the system type (AI or Human), the amount of information
(low or high level) about the gas monitoring system given
to the participant, and other demographic factors that have
been studied by other researchers [40, 37, 5] and seen to
affect decision making and perceived safety, were captured
as independent variables.
RESULTS
The results of the ANOVA test for significance are pre-
sented on Table 1 while Figure 1 shows the effects of system
type and information level on safety and trust scores. From
the results, system type had no effect on trust, F(1, 195) =
0.003, p =0.957, eta2 =2.08×10–5 but higher informa-
tion led to more trust, F(1, 195) =4.733, p =0.031 and
eta2 =2.37×10–2 and there was no interaction between the
two, F(1, 195) =0.141, p =0.708 and eta2 =7.06 × 10–4.
System type did not have a significant impact on partici-
pants safety and trust perception (p =0.957), which implies
that, in terms of trust and safety, knowing the specific type
(AI vs human) of gas monitoring system did not have much
influence.
Effect of Information Level on Safety and Trust
Figure 1 presents the results of the ANOVA test for the
influence of the amount of information provided to partici-
pants on their perceived safety and trust. It shows that pro-
viding participants with more information about the system
that sends out the alert message increases their safety and
trust perception. The amount of system information did
have a significant impact on participants’ safety and trust
perception (p =0.031), and it is worth noting that partici-
pants who received more information had a higher score
than participants who received less information, implying
increasing the amount of information increases participants
safety and trust.
Effect of Demographics on Safety and Trust
All demographic variables were entered into a linear regres-
sion model to reveal their effect on perceptions of safety:
gender, ethnicity, age, education level, marital status,
income level, and number of children. Backward selection
Figure 1. Effect of System Type and Information Level on
Trust and Safety Score