3
The intersection of these characteristics adds complex-
ity. For example, gender and ethnicity together influence
trust in technology adoption [33], while age and education
interact to shape trust in online transactions [34].
Emergency Alert System
Alert systems commonly use voice and text messages to
inform people about emergencies situations. Sending text
messages are popular and mostly preferred over voice mes-
sages since text messages can be viewed later. It also offers
advantages like lower bandwidth needs and clearer infor-
mation [35].
Public alert systems differ significantly from mine emer-
gency alerts due to miners’ unique demographics, urgent
decision-making context, and limited evacuation routes.
While text messages are effective in public safety guidance,
limited research exists on miners’ responses to such alerts,
likely due to recruitment challenges [36].
Studies on human response to emergency notifications,
guided by communication theory and agent-based model-
ing, outline a multi-stage process of message processing and
response. A foundational three-step model involves hear-
ing, perceiving, and responding to alerts [37], while more
recent frameworks describe six steps: receiving, understand-
ing, believing, personalizing the threat, deciding to act,
and confirming the message with additional information
[38, 39].
Research highlights how alert content, style, and
receiver context critically influence responses to emergency
alerts. Receiver characteristics such as gender, age, and
familiarity with the environment significantly impact min-
ers’ responsiveness to evacuation messages [37]. The litera-
ture suggests five essential alert elements: hazard, location,
guidance, time, and source, which supports the hypothesis
that higher specificity in alerts may improve compliance
among miners. Although text messages can serve as effective
communication channels in emergencies, their limitations,
including length, deliverability, and risk of being over-
looked, present unique challenges. These insights underpin
hypotheses regarding miners’ demographic factors, message
specificity, and communication channels, aiming to opti-
mize emergency messaging for enhanced miner safety.
From the literature available, there are no works that
investigate the influence system type and amount of infor-
mation has on trust and perceived safety. To fill this gap, this
study investigates how safety, trust and decision-making
during gas emergencies are influenced by knowledge of the
gas monitoring type and amount of information provided
about the gas monitor. Findings aim to improve safety by
increasing trust that miners have in gas monitoring systems.
METHODOLOGY
This study was conducted to test 2 hypotheses: H1: AI
gas monitoring system has a significant effect on peoples
perceived safety and H2: High amount of information
provided about the gas monitoring system positively influ-
ences safety, trust and decision making. This was achieved
through an online survey experiment with participants
recruited through Prolific to test if perceived safety, trust
and decisions during gas emergencies can be influenced
by knowledge of the gas system type (human vs. AI) and
amount of information provided about the system. A 2×2
factorial experiment with these two factors was conducted.
Each participant recruited was randomly assigned to one of
the 4 conditions.
The questionnaire was designed to mimic a scenario of
gas exposure in an underground mine, where participants
play the role of workers in the mine. The participants were
given a background about gas monitoring in an under-
ground mine and were informed about the specific type
of gas monitoring system in accordance with their experi-
mental condition. Participants were asked to indicate their
initial perception about safety in an underground mine,
after which they received an alert message about the gas
situation in the mine. Two different alert messages were
sent out with the first message indicating a gas leakage with
concentrations below threshold and a second message indi-
cating a gas leakage with concentration above threshold.
The second message spells out the next action the partici-
pant should take to ensure their safety. Each message was
followed by questions that measured their perceived safety
in the mine and trust in the alert system. There were addi-
tional questions that seek to find out if participants would
like to replace the gas monitoring system sending them the
alert message with another system type, delegate the duty of
the gas monitoring to an AI system, their preference of AI
system over Human system and vice versa, and their gen-
eral trust in gas monitoring systems. The questionnaire was
audited and approved by Missouri University Institutional
Review Board.
The responses of participants from the study were ana-
lyzed in R studio.
Study Participants
An online survey experiment was conducted in July 2024
with participants recruited through Prolific. Using the
general characteristics of miners in terms of gender, educa-
tion, age, quotas were set in Prolific to ensure participants
have similar distribution as miners in US. Each partici-
pant indicated their consent to partake in the study after
The intersection of these characteristics adds complex-
ity. For example, gender and ethnicity together influence
trust in technology adoption [33], while age and education
interact to shape trust in online transactions [34].
Emergency Alert System
Alert systems commonly use voice and text messages to
inform people about emergencies situations. Sending text
messages are popular and mostly preferred over voice mes-
sages since text messages can be viewed later. It also offers
advantages like lower bandwidth needs and clearer infor-
mation [35].
Public alert systems differ significantly from mine emer-
gency alerts due to miners’ unique demographics, urgent
decision-making context, and limited evacuation routes.
While text messages are effective in public safety guidance,
limited research exists on miners’ responses to such alerts,
likely due to recruitment challenges [36].
Studies on human response to emergency notifications,
guided by communication theory and agent-based model-
ing, outline a multi-stage process of message processing and
response. A foundational three-step model involves hear-
ing, perceiving, and responding to alerts [37], while more
recent frameworks describe six steps: receiving, understand-
ing, believing, personalizing the threat, deciding to act,
and confirming the message with additional information
[38, 39].
Research highlights how alert content, style, and
receiver context critically influence responses to emergency
alerts. Receiver characteristics such as gender, age, and
familiarity with the environment significantly impact min-
ers’ responsiveness to evacuation messages [37]. The litera-
ture suggests five essential alert elements: hazard, location,
guidance, time, and source, which supports the hypothesis
that higher specificity in alerts may improve compliance
among miners. Although text messages can serve as effective
communication channels in emergencies, their limitations,
including length, deliverability, and risk of being over-
looked, present unique challenges. These insights underpin
hypotheses regarding miners’ demographic factors, message
specificity, and communication channels, aiming to opti-
mize emergency messaging for enhanced miner safety.
From the literature available, there are no works that
investigate the influence system type and amount of infor-
mation has on trust and perceived safety. To fill this gap, this
study investigates how safety, trust and decision-making
during gas emergencies are influenced by knowledge of the
gas monitoring type and amount of information provided
about the gas monitor. Findings aim to improve safety by
increasing trust that miners have in gas monitoring systems.
METHODOLOGY
This study was conducted to test 2 hypotheses: H1: AI
gas monitoring system has a significant effect on peoples
perceived safety and H2: High amount of information
provided about the gas monitoring system positively influ-
ences safety, trust and decision making. This was achieved
through an online survey experiment with participants
recruited through Prolific to test if perceived safety, trust
and decisions during gas emergencies can be influenced
by knowledge of the gas system type (human vs. AI) and
amount of information provided about the system. A 2×2
factorial experiment with these two factors was conducted.
Each participant recruited was randomly assigned to one of
the 4 conditions.
The questionnaire was designed to mimic a scenario of
gas exposure in an underground mine, where participants
play the role of workers in the mine. The participants were
given a background about gas monitoring in an under-
ground mine and were informed about the specific type
of gas monitoring system in accordance with their experi-
mental condition. Participants were asked to indicate their
initial perception about safety in an underground mine,
after which they received an alert message about the gas
situation in the mine. Two different alert messages were
sent out with the first message indicating a gas leakage with
concentrations below threshold and a second message indi-
cating a gas leakage with concentration above threshold.
The second message spells out the next action the partici-
pant should take to ensure their safety. Each message was
followed by questions that measured their perceived safety
in the mine and trust in the alert system. There were addi-
tional questions that seek to find out if participants would
like to replace the gas monitoring system sending them the
alert message with another system type, delegate the duty of
the gas monitoring to an AI system, their preference of AI
system over Human system and vice versa, and their gen-
eral trust in gas monitoring systems. The questionnaire was
audited and approved by Missouri University Institutional
Review Board.
The responses of participants from the study were ana-
lyzed in R studio.
Study Participants
An online survey experiment was conducted in July 2024
with participants recruited through Prolific. Using the
general characteristics of miners in terms of gender, educa-
tion, age, quotas were set in Prolific to ensure participants
have similar distribution as miners in US. Each partici-
pant indicated their consent to partake in the study after