Content uploaded by L. G. Pee
Author content
All content in this area was uploaded by L. G. Pee on Aug 10, 2018
Content may be subject to copyright.
Page 1 of 25
Trust in User-Generated Information on Social Media during Crises:
An Elaboration Likelihood Perspective
Abstract
Social media are increasingly being used as a source of information during crises such as
natural disasters and civil unrests. Nevertheless, there have been concerns about the
quality and truthfulness of user-generated information. This study seeks to understand
how users form trust in information on social media. Based on the elaboration likelihood
model and the motivation, opportunity, and ability framework, this study proposes and
empirically tests a model that identifies the information processing routes through which
users develop trust, and the factors influencing the use of the routes. Findings from a
survey of Twitter users seeking information about the Fukushima Daiichi nuclear crisis
indicate that personal relevance and level of anxiety moderate individuals’ use of
information processing routes. This study extends the theorization of trust in
user-generated information. The findings also suggest practical approaches for managing
social media during crises.
Keywords: User-generated information; trust; elaboration likelihood model; motivation,
opportunity, ability framework; crisis information
Cite as: L. G. Pee and J. Lee (2016) Trust in User-Generated Information on Social Media during Crises: An Elaboration
Likelihood Perspective, Asia Pacific Journal of Information Systems, 26 (1), pp. 1-21
Page 2 of 25
Trust in User-Generated Information on Social Media during Crises:
An Elaboration Likelihood Perspective
1. Introduction
Social media are increasingly being used as a source of up-to-the-minute information
about what is happening on the ground during large-scale crises (Westerman et al. 2014)
such as natural disasters and civil unrests (e.g., street riot, political reform). For instance,
during the Oklahoma Grassfires and the Red River Floods that occurred in the United
States in 2009, millions of messages containing information about the location of
affected areas, evacuation sites, damages, and injuries were shared on Twitter (Vieweg et
al. 2010). User-generated information is seen as helpful for improving situation
awareness, which is the perception of elements in a crisis, the comprehension of their
meaning, and the projection of their status in the near future (Yin et al. 2012). Situation
awareness helps individuals assess their personal situation or gain a broad understanding
of the crisis. At times, the use of social media during crises even surpasses other media
because Internet access often remains robust when landlines, base stations of mobile
phones, and power lines become congested or damaged (Ichiguchi 2011).
Despite the informational uses and benefits of social media during crises, there
have been concerns about the quality and truthfulness of user-generated information.
Social media often contain unverified information, misinterpretations, and even
fabricated content. It is sometimes considered as a collective rumor mill that propagates
misinformation, gossip, and, in extreme cases, propaganda (Mendoza et al. 2010; Oh et al.
2010). Many users find it difficult to distinguish between true and false information on
social media (Acar and Muraki 2011). Trusting false information not only leads users to
make wrong decisions, it can also have dire social consequences such as fueling mass
panic. For instance, it is widely believed that rumors spread through social media such as
Twitter and Facebook triggered the mass unrest in the 2011 England Riots (Grimmer
2011). It is therefore important to understand how users evaluate and develop trust in
information on social media (Mendoza et al. 2010). The objective of this study is to
propose and empirically assess a model that identifies two information processing routes
through which individuals develop trust, based on the elaboration likelihood model (Petty
and Cacioppo 1986). Further, drawing on the motivation, opportunity, ability (MOA)
Page 3 of 25
framework, we propose that individuals’ use of information processing routes depends on
the personal relevance of information, level of anxiety, and prior knowledge.
The proposed model contributes to theoretical development and the management of
social media in several ways. First, it focuses on the consumption of user-generated
information and narrows a gap in research which has mainly focused on the motivation to
generate content (e.g., Chai 2011; Kim et al. 2009). Second, it sheds light into the
informational processes through which individuals form trust in user-generated
information. Although the elaboration likelihood model has been applied to study trust in
electronic words of mouth (Cheung and Thadani 2012), it is not clear whether the model
is applicable to the crisis context, in which people often frantically seek information from
all available sources to inform their actions and trusting false information can be
especially dangerous. Our proposed model considers the effect of anxiety, which is
particularly relevant in the crisis context. Third, this study focuses on the crisis context
wherein information is critical yet understudied. The model was assessed with data
collected from individuals who sought information about a real-life crisis. Overall, the
proposed model is theoretically grounded and practically relevant. In the following
sections, the conceptual background for the proposed model as well as the study and
findings will be detailed.
2. Conceptual Background
This section first reviews prior studies on trust in user-generated information to identify
gaps in research. This is followed by a discussion of the information processing routes
according to the elaboration likelihood model. Individuals’ motivation, opportunity, and
ability affecting their use of the information processing routes are then described.
2.1. Trust in User-Generated Information
In this study, trust is defined as the extent to which one feels secure and comfortable
about relying on the information on social media. Our review of studies on trust in
user-generated information (see Table 1) shows that prior studies have examined the
effect of trust on factors such as attitude (Bartle et al. 2013) and social media use (Anish
et al. 2014; Chu and Kim 2011). Prior studies have also found that trust is affected by the
source of information (e.g., authority, reputation, integrity; Burgess et al. 2011;
Dickinger 2011), history of interactions with source (Kim and Ahmad 2013) and content
characteristics (e.g., informativeness, quality, volume; Flanagin and Metzger 2013).
Page 4 of 25
Another stream of research has proposed methods for quantifying the trustworthiness of
user-generated information (Lenders et al. 2008; Moturu and Liu 2011).
The review shows that prior studies have mostly examined trust among users facing
purchase decisions or seekers of travel, lifestyle, and health information. This study seeks
to extend research on the topic by examining individuals’ trust in crisis-related
information, which is an important type of information that is increasingly sought and
used during crises to cope with uncertainty.
It can also be observed from the review that one important means through which
individuals determine their trust in user-generated information is processing and
evaluating the content of the information. This suggests that information processing
theories may augment our understanding of the phenomenon. This study is one of the
first to draw on one such theory, the elaboration likelihood model (ELM), to theorize
trust in user-generated information by identifying two routes of information processing as
well as the conditions affecting their use. ELM will be detailed next.
Page 5 of 25
Table 1. Studies of Trust in User-Generated Information
Source Key Findings and Key Constructs* Method and Sample
Anish et al.
(2014) Trust did not influence the amount of usage of user-generated review platforms Survey of 72 users of Yelp.com
Bartle et al.
(2013) - Trust was associated with strong positive attitude towards cycling as a commuter mode
- Information shared within a group inspired greater trust amongst participants than “official”
information, largely because it was seen as emanating from the experience of “real people”
- There were both calculus trust, arising from the intrinsic quality of the information, and
relational trust, associated with the relationship between information-giver and receiver
Case study of a map-based
website in the United Kingdom
Burgess et al.
(2011) - Greater trust was placed in online travel comments when they were on a specific travel
website than when they were on a generic social networking website
- The highest level of trust was afforded to information provided on state government websites
Survey of 12,000 Australian travel
consumers
Chu and Kim
(2011) Trust influenced a) opinion seeking, b) opinion giving, and c) opinion passing of electronic
word of mouth (eWOM) on social networking sites (SNS) Survey of 400 undergraduate
students who used SNS
Dickinger
(2011) a) Informativeness, b) integrity, c) benevolence, and d) ability increased overall trust of
online channels Survey of 453 tourists in Vietnam
who were also Internet users
Flanagin and
Metzger
(2013)
Information volume increased a) perceived information credibility, b) reliance on the
information, c) confidence in accuracy, and d) congruence with others’ evaluation of the
information
Experiment involving 1,207
Internet users who viewed a
fictitious movie rating website
Kim and
Ahmad (2013) - Distrust was subjective and based on direct experience rather than statements from others
- Trust needed strong evidence like a cumulative history of positive direct experience or a high
public reputation in order to distinguish from lack of confidence interactions
Analyses of 1,560,144 reviews
and 12,668,319 ratings for reviews
provided by 326,983 users on
Epinions
Lenders et al.
(2008) The proposed secure localization and certification service helped content consumers to establish
the trust level of contents Geotagging service
Moturu and
Liu (2011) The proposed approach helped to quantify the trustworthiness of shared content on social
media, based on aspects such as author reputation, content performance (e.g., number of edits,
number of references), and appearance
Content on Wikipedia and Daily
Strength (a health social network)
* Constructs in quantitative studies are indicated in bold font
Page 6 of 25
2.2 Elaboration Likelihood Model (ELM)
ELM posits that information can change individuals’ attitude through the central and
peripheral routes of information processing (Petty and Cacioppo 1986). The central route
of information processing involves scrutinizing the content of information to determine
its inherent merits prior to forming an attitude. That is, information quality is the main
determinant of individuals’ attitude. High-quality information is likely to be perceived as
more trustworthy because it can better support sense making and improve decision
accuracy (O'Reilly 1982). In line with this, studies of health information websites
observed that users find it less risky to trust high-quality information (Koo et al. 2014;
Luo and Najdawi 2004). The other route of information processing is the peripheral route,
which involves the use of cues or heuristics (e.g., source credibility, opinion of the
majority; Diane 1987) to form an attitude and therefore requires less cognitive effort than
the central route.
ELM is often studied in social psychology and marketing research and is
increasingly being applied in information systems (IS) research (Bhattacherjee and
Sanford 2006). The model has been adapted to explain how individuals form attitudes
towards IS, which in turn influence their adoption of IS (Angst and Agarwal 2009) and
intention to continue using IS (Li 2013). ELM has also served as the basis for
understanding individuals’ acceptance and use of information accessed through
technologies such as expert systems (e.g., Dijkstra 1999; Mak et al. 1997) and
e-commerce websites (e.g., Ho and Bodoff 2014; Yang et al. 2006; Zhou 2012). This
indicates that ELM can potentially offer insights into individuals’ trust in information on
social media.
Although ELM identifies the opinion of others as an important heuristic for
processing information and forming attitude (Petty and Cacioppo 1986), the effect of this
heuristic is seldom examined in IS studies applying ELM. The opinion of others
represents social influence and is especially relevant in the context of social media whose
key feature is enabling socializing. Therefore, this study considers the opinion of others
through the construct of majority influence, which reflects the extent to which most
people in a social group hold similar opinion about an issue (Nemeth 1986). On social
media, majority influence may manifest in terms of the extent of agreement (e.g., number
of readers expressing support) or the spread of the information among different users
Page 7 of 25
(e.g., number of times a piece of information is forwarded or reposted). When a piece of
information is supported by many people, it may be perceived as having been endorsed
and validated by the majority and therefore more trustworthy (Chaiken and Maheswaran
1994). This is in line with the concept of social proof, which suggests that individuals
facing uncertainties determine what is correct based on what others think is correct
(Cialdini 1993).
In ELM, the extent to which individuals use the central route (i.e., evaluate
information quality) and the peripheral route (i.e., use the heuristic of majority influence)
depends on their elaboration likelihood, which is influenced by their motivation and
ability to evaluate information (Petty and Cacioppo 1986). Individuals with strong
motivation and ability to process information are likely to expend more cognitive
resources to evaluate the quality of information and rely less on peripheral heuristics.
Motivation and ability are also keys aspects of the MOA framework, which is originally
proposed to explain consumers’ processing of brand information in advertisements. In
addition to motivation and ability, the MOA framework suggests that the opportunity to
process information can influence the amount of attention allocated to a piece of
information. Therefore, we extend ELM with the MOA framework by considering
opportunity to process information in our proposed model. The MOA framework is
detailed next.
2.3 Motivation, Opportunity, Ability (MOA) Framework
MacInnis et al. (1991) propose that consumers’ processing of advertising information is
influenced by their motivation and ability, as well as the opportunity to do so. Motivation
refers to the driving force that generates desire and increases willingness to process
information; opportunity is the extent to which distractions or limited exposure time
affect individuals’ attention to process information; ability refers to the knowledge or
skills relevant to the information to be processed. The MOA framework has been adapted
to study many different behaviors beyond consumer research, including individuals’ use
of social networking sites (e.g., Leung and Bai 2013).
As mentioned earlier, motivation and ability are expected to influence elaboration
likelihood in ELM. Motivation is conceptualized in terms of personal relevance and
ability is based on one’s prior knowledge in ELM (Petty and Cacioppo 1986). Similarly,
prior IS studies applying ELM have predominantly conceptualized motivation and ability
Page 8 of 25
in terms of these constructs (Angst and Agarwal 2009; Bhattacherjee and Sanford 2006).
Accordingly, we consider motivation and ability in terms of personal relevance and prior
knowledge in our proposed model.
With regard to the opportunity to process information, an important source of
distraction that can draw one’s attention away from information processing during crises
is anxiety. Anxiety is common among individuals facing a crisis. A diminution of
available attention can be expected when individuals are anxious and fearful since these
negative emotions often require an immediate, active response. As a form of arousal,
anxiety also leads to an increase in self-focused attention which may distract a person
from thoroughly processing the external social environment (Wilder 1993). Anxiety can
distract individuals from attending to their environment and cause them to rely more on
available cognitive structures such as social stereotypes in making judgment of others
(Sarason 1988; Wilder 1993).
3. Research Model and Hypotheses
Based on ELM, our proposed model considers two routes of information processing
through which individuals determine trust in user-generated information. The central
route involves evaluating information quality and is specified in the model as the effect
of information quality on trust. The peripheral route relies on majority influence and is
specified as the effect of majority influence on trust. According to ELM, the use of the
routes depends on elaboration likelihood, which is determined by personal relevance (a
motivation factor) and prior knowledge (an ability factor). In terms of modeling, personal
relevance and prior knowledge are expected to moderate the impact of information
quality and majority influence on trust (e.g., information quality has a stronger impact on
trust when personal relevance is high). As discussed earlier, we extend ELM with the
MOA framework by considering anxiety as a distraction that reduces the opportunity to
process information. Therefore, anxiety is also expected to moderate the impact of
information quality and majority influence on trust (see Figure 1). The hypotheses in the
moderated model are detailed next.
Page 9 of 25
Figure 1: Model of Trust in User-Generated Crisis Information on Social Media
3.1 Moderating Effects of Personal Relevance
Personal relevance is the extent to which an issue is expected to have a significant
consequence on one’s life (Apsler and Sears 1968). When personal relevance is high, the
consequence of being incorrect is experienced strongly and personally. For instance, for
those who live within an area where a natural disaster has been forecasted, trusting false
information that the disaster would not occur can endanger their lives directly. Personal
relevance increases individuals’ sufficiency threshold in information processing (Chaiken
et al. 1989). This prompts individuals to engage in more information processing to satisfy
the increased need. Personal relevance also motivates individuals to increase their
judgmental confidence to avoid the dire consequence of trusting false information. They
are therefore likely to allocate more cognitive resources to assess the quality and validity
of information and rely less on peripheral heuristics (Chaiken et al. 1989; Petty and
Cacioppo 1986). This suggests that when the personal relevance of crisis information is
high, individuals rely more on the central route of information processing and less on the
peripheral route. In other words, the effect of information quality on trust is likely to
strengthen while the effect of majority influence (which is a peripheral heuristic) is likely
to weaken.
H1: As personal relevance increases, the effect of information quality on trust in
user-generated crisis information increases.
H2: As personal relevance increases, the effect of majority influence on trust in
user-generated crisis information decreases.
Page 10 of 25
3.2 Moderating Effects of Anxiety
Anxiety involves the selective processing of information perceived as signifying a threat
or danger to one’s personal safety or security (Beck and Clark 1997). At the cognitive
level, anxiety involves: a) certain sensory-perceptual symptoms such as feelings of
unreality, hypervigilance, and self-consciousness; b) thinking difficulties such as poor
concentration, inability to control thinking, blocking, and difficulty in reasoning; and c)
conceptual symptoms like cognitive distortions, fear-related beliefs, frightening images
and frequent automatic thoughts (Beck and Clark 1997). In general, anxiety engages
cognitive resources in mental activities such as worrying, leaving less capacity for
tackling other cognitive tasks (Eysenck et al. 2007). In support, studies have shown that
individuals under high anxiety exhibit lower performance in tasks that demand cognitive
resources (e.g., Ashcraft 2002). The reduced cognitive capacity is likely to have
implications for the elaboration and processing of information, which can be viewed in
terms of the amount of thought or scrutiny devoted to a piece of information (Petty and
Cacioppo 1986). With lowered cognitive capacity, anxious individuals are more likely to
rely on peripheral cues which demand less effort to process and less likely to evaluate
information quality. In support, a study observed that high-trait-anxiety individuals are
often persuaded by the peripheral cue of source attractiveness regardless of argument
quality, while low-anxiety individuals are persuaded by argument quality regardless of
source attractiveness (DeBono and McDermott 1994). Nevertheless, there has been a lack
of study on the effect of state anxiety (how one feels in a particular situation) on the use
of information processing routes during crises. The following hypotheses are assessed to
narrow this gap:
H3: As anxiety increases, the effect of information quality on trust in user-generated
crisis information decreases.
H4: As anxiety increases, the effect of majority influence on trust in user-generated crisis
information increases.
3.3 Moderating Effects of Prior Knowledge
Prior knowledge refers to one’s familiarity, expertise, and experience with an issue
(Kerstetter and Cho 2004). Prior knowledge can serve to disambiguate information
(Chaiken et al. 1989). When individuals have strong prior knowledge about an issue, they
are better able to scrutinize the content of information and there is therefore less need to
Page 11 of 25
rely on peripheral heuristics (Bhattacherjee and Sanford 2006). In contrast, individuals
with little prior knowledge lack the ability to process information critically and they are
therefore forced to rely more on peripheral heuristics (Petty and Cacioppo 1986).
Accordingly, we hypothesize that:
H5: As prior knowledge increases, the effect of information quality on trust in crisis
user-generated information increases.
H6: As prior knowledge increases, the effect of majority influence on trust in
user-generated crisis information decreases.
4. Research Method
The target population of this study is individuals who sought crisis-related information
on social media. Data were collected in a survey of individuals who sought information
about the Fukushima Daiichi nuclear crisis on Twitter. On 11 March 2011, a tsunami
triggered by the magnitude 9.0 Tohoku earthquake led to a nuclear meltdown involving
three of the six nuclear reactors at a Fukushima nuclear plant, causing the largest nuclear
incident since the Chernobyl disaster in April 1986 and the only (after Chernobyl) to
measure level 7 on the International Nuclear Event Scale. After the incident became
publicly known, many individuals within and outside Japan turned to social media for
up-to-date information about the extent and effects of radiation on air quality and food
sources (Acar and Muraki 2011). Millions of messages containing information related to
the nuclear crisis were posted on social networking sites, including Twitter (Doan et al.
2012).
In Twitter, information on a topic can be accessed by searching “tweets”, which are
text-based messages of up to 140 characters. A message can include links to other
webpages, which are typically used to provide further supporting information beyond the
140-character limit. Tweets are by default open to the public and can be retrieved by
anyone with Internet access (Shi et al. 2013). As of March 2011, the average number of
tweets per day was about 140 million. On 11 March 2011, the day of the Tohoku
earthquake, the average number of daily tweets increased to 177 million (Smith 2011).
Messages such as the following abounded Twitter (Zax 2011):
“Nuclear Ash Cloud Of Radiation Raining On Tokyo From Burning Of Radioactive
Fukushima Sewage Sludge... http://fb.me/W3FIGu2C”
"The specialists in the nuclear sites are getting less and less -- who will be left to
Page 12 of 25
work on them? Leave Tokio and go south for now -- at least and take the OLD
People with you!"
"Luckily I have been able to get a seat on a flight to Okinawa today. I am catching
the 2000 flight from Haneda. Those still around, be careful not to get rained on."
"Don't believe government reassurances radiation levels are safe -- get out of
Japan now."
"The situation at the nuclear plants in Fukushima is getting worse and worse, and I
am getting very afraid of it. Now, I am going out for grocery shopping with my sick
child in search for more water and other supplies."
Compared to other social networking sites such as Facebook, Twitter is quite open
and loose. The relationship between the message poster and reader often cuts across long
(real-world) social distances (Shi et al. 2013). It has been shown that any retweets (i.e.,
messages that are reposted) on Twitter reach an average of 1,000 users regardless of the
number of subscribing followers in the original message and can be read by people who
are four degrees of separation away from the source within minutes (Kwak et al. 2010).
Twitter therefore more closely resembles an information broadcasting site than a
traditional social network and is particularly relevant for testing our proposed model. The
development of survey instrument and data collection are described next.
4.1. Development of Survey Instrument
The constructs in the proposed model were operationalized based on instruments
validated in prior studies as much as possible (see Table 2). The items measuring
information quality, personal relevance, anxiety, and trust were adapted from validated
scales while items measuring majority influence and prior knowledge were developed
based on their conceptualizations. The items measuring information quality were scored
on semantic-differential scales while the others were scored on five-point Likert scales.
4.2. Data Collection and Sample Demographics
The invitation to participate in the survey was posted in online forums that discussed
topics related to the Fukushima nuclear crisis. Users of Twitter who sought information
about the Fukushima crisis were invited to complete a web-based survey. As an incentive
for participation, respondents had the option of entering a lucky draw of vouchers for an
international shopping website. The survey was open to individuals residing within as
well as outside of Japan to ensure sufficient variance in personal relevance, which is one
Page 13 of 25
of the constructs of interest in our study. We received a total of 198 responses. Most of
the respondents were residing in Japan (26.8 percent; see Table 3) and the United States
(19.7 percent). Male respondents constituted 53 percent and 69.2 percent of the
respondents aged from 21 to 35. Most respondents had one to two years of experience
using Twitter (54.6 percent) and more than five years of experience using the Internet
(66.7 percent).
Table 2. Survey Instrument
Construct Item and Source
Information
quality I think the information related to nuclear radiation on Twitter is generally …
1. subjective/objective
2. unverifiable/verifiable
3. has insufficient/sufficient breadth or coverage
4. has insufficient/sufficient depth or detail
5. outdated/up-to-date
6. difficult/easy to understand
(Scored on semantic-differential scales; All items adapted from Lee et al.
2002)
Majority
influence 1. On Twitter, most people hold largely similar views about the effects o
f
radiation
2. On Twitter, most people share consensus about the effects of radiation
3. On Twitter, there is general agreement about the effects of radiation
(All items developed based on Martin et al. 2002)
Personal
relevance
(formative
measure)
1. There is a high possibility that I will experience the negative effects o
f
nuclear radiation in future
2. My physical health makes it more likely that I will experience the negative
effects of nuclear radiation
3. My geographic location makes it more likely that I will experience the
negative effects of nuclear radiation
4. My occupation makes it more likely that I will experience the negative
effects of nuclear radiation
(All items adapted from Champion 1984; Clarke 1999)
Anxiety 1. I feel anxious (worrying, anticipation of the worst) about the Fukushima
nuclear crisis
2. I feel tense (trembling, feeling of restlessness, unable to relax) due to the
Fukushima nuclear crisis
3. I have difficulty falling asleep due to the Fukushima nuclear crisis
4. I feel depressed due to the Fukushima nuclear crisis
(All items adapted from Hamilton 1959)
Prior knowledge 1. I have professional expertise in domains related to nuclear radiation
2. I had personally experienced the effects of nuclear radiation
3. I had spent a lot of time reading about nuclear radiation on sources othe
r
than Twitter
(All items developed based on Kerstetter and Cho 2004)
Trust in
user-generated
crisis information
1. In general, I trust the information related to nuclear radiation on Twitter
2. I feel secure using the information related to nuclear radiation on Twitter in
decision making
3. I feel comfortable using the information related to nuclear radiation on
Twitter in decision making
(All items developed based on Komiak and Benbasat 2006)
Page 14 of 25
Table 3. Sample Demographics
Variable Value Percentage Count
Age 18 to 20 2.53% 5
21 to 25 19.19% 38
26 to 30 26.77% 53
31 to 35 23.23% 46
36 to 40 13.64% 27
41 to 45 7.07% 14
46 to 50 4.04% 8
> 50 3.54% 7
Gender Male 53.03% 105
Female 46.97% 93
Country of Residence Japan 26.77% 53
US 19.70% 39
Canada 9.60% 19
Australia 8.59% 17
China 12.12% 24
Singapore 13.64% 27
Malaysia 9.60% 19
Experience using Twitter Less than 1 year 21.21% 42
1 to 2 years 54.55% 108
3 to 4 years 18.18% 36
5 to 6 years 6.06% 12
Experience using Internet Less than 1 year 0.00% 0
1 to 2 years 1.52% 3
3 to 4 years 31.82% 63
5 to 10 years 53.54% 106
>10 years 13.13% 26
5. Data Analysis and Results
The data were analyzed using Partial Least Squares (PLS), a structural equation modeling
technique that concurrently tests the measurement model and structural model (Chin et al.
2003). PLS was used because it is able to account for formative and reflective constructs
jointly occurring in a single structural model. A reflective construct has indicators that
are affected by a single underlying latent construct and removing an indicator should not
alter the conceptual domain of the construct (Jarvis et al. 2003). On the other hand, a
formative construct is a composite of multiple indicators and excluding an indicator may
alter the conceptual domain of the construct (Jarvis et al. 2003). In this study, personal
relevance is a formative construct because its items tap into different themes and the
items are not interchangeable. For example, physical health (second measurement item)
and geographic location of a person (third item) may not always be correlated (see Table
Page 15 of 25
2). The other constructs are considered reflective. All data were standardized prior to
analyses.
5.1. Tests of Measurement Model
The survey instrument was tested for reliability, convergent validity, and discriminant
validity. Reliability of each construct was assessed with Cronbach’s alpha coefficient (see
Table 4). All constructs achieved scores above the recommended value of 0.70 (Hair et al.
2009). Convergent validity was assessed by examining composite reliability and average
variance extracted (AVE) by each construct (see Table 4). All composite reliabilities and
AVEs were above the recommended value of 0.70 (Hair et al. 2009), indicating that the
instrument had satisfactory convergent validity.
Table 4. Tests of Measurement Model
Construct Cronbach’s
Alpha Composite
Reliability
Average
Variance
Extracted Mean
Standard
Deviation
Information Quality (IQ) 0.91 0.93 0.68 3.50 0.60
Majority Influence (MI) 0.88 0.93 0.81 3.73 0.62
Anxiety (AX) 0.84 0.89 0.67 3.63 0.74
Prior Knowledge (PK) 0.74 0.85 0.66 4.25 0.54
Trust in User-Generated
Information (TI) 0.71 0.84 0.64 3.80 0.70
Discriminant validity was assessed by factor analysis (see Table 5) and comparing
construct correlations with square root of AVEs (see Table 6). The results indicated that
all items loaded highly on their stipulated constructs (i.e., with value exceeding 0.70) but
not highly on other constructs. All constructs correlated more highly with their own items
than with items measuring other constructs (Fornell and Larcker 1981). These indicate
that discriminant validity was satisfactory. We also assessed multicollinearity by
calculating variance inflation factor (VIF). The resultant values ranged from 1.02 to 2.98,
which were below the threshold value of 3.33 (Diamantopoulos and Winklhofer 2001).
Table 5. Item Loading in Factor Analysis
Construct Item 1 Item 2 Item 3 Item 4 Item 5 Item 6
Information Quality (IQ) 0.75 0.80 0.84 0.81 0.90 0.86
Majority Influence (MI) 0.92 0.92 0.86
Anxiety (AX) 0.81 0.80 0.85 0.82
Prior Knowledge (PK) 0.78 0.73 0.92
Trust in User-Generated
Information (TI) 0.82 0.84 0.92
Page 16 of 25
Table 6. Correlation and Square Root AVE
Construct 1 2 3 4 5
1. Information Quality (IQ) 0.82*
2. Majority Influence (MI) 0.28 0.90
3. Anxiety (AX) 0.25 0.28 0.82
4. Prior Knowledge (PK) 0.30 0.34 0.34 0.81
5. Trust in User-Generated Information (TI) 0.63 0.66 0.43 0.40 0.80
*Bold diagonal entries are square root of AVE
For the formative construct of personal relevance, these tests are not applicable.
Instead, significance of item weights was examined to determine the contribution of
items constituting the construct. The results were favorable, as all item weights were
significant at p<0.05. VIFs were also calculated and they were all below the
recommended threshold of 3.33, indicating that the items captured different aspects of
personal relevance.
5.2. Tests of Structural Model
The PLS latent variable modeling approach for analyzing interaction effects (Chin et al.
2003) was used to test the moderating hypotheses. The procedure involves computing
interaction terms by multiplying the predicting and moderating constructs. For
interaction terms involving the formative construct of personal relevance, the two-step
score construction procedure suggested by Chin et al. (2003) was used to create
underlying construct scores for the predictor and moderator variables before creating the
interaction terms.
Our hypotheses specify the sign of path coefficients based on ELM and MOA
framework and were thus assessed with one-tailed p-values (Kock 2015). Results of the
structural model are shown in Table 7 and Figure 2. We found that all hypotheses were
supported except for the moderating effects of prior knowledge. Prior knowledge also did
not have a significant direct effect on trust. Among the control variables, age had a
significant negative effect on trust, but not the level of education, gender, number of
years using Twitter, and number of years using the Internet. The proposed model
explained 67.5% of the variance in trust.
Page 17 of 25
Table 7. Results of Hypothesis Testing
Relationship Path
Coefficient T
Value Result
Information quality trust 0.09 0.73 Not significant
Majority influence trust 0.61*** 4.59 Significant
Personal relevance trust 0.25** 2.60 Significant
Personal relevance * information quality trust 0.15* 1.78 H1 is supported
Personal relevance * majority influence trust -0.14* 1.98 H2 is supported
Anxiety trust 0.06 0.97 Not significant
Anxiety * information quality trust -0.13* 1.81 H3 is supported
Anxiety * majority influence trust 0.12* 1.75 H4 is supported
Prior knowledge trust -0.07 1.43 Not significant
Prior knowledge * information quality trust 0.00 0.05 H5 is not supported
Prior knowledge * majority influence trust -0.05 0.73 H6 is not supported
*p<0.05; **p<0.01; ***p<0.001
*p<0.05; **p<0.01
Figure 2: Path Coefficients of Structural Model
The significant moderating effects are plotted in Figure 3. It can be observed that
for information with high personal relevance, the effect of information quality on trust
strengthens (in Figure 3a, the solid-line slope is steeper than the broken line slope) while
the effect of majority influence weakens (in Figure 3b, the solid-line slope is gentler).
This provides support for hypotheses H1 and H2. For high-anxiety individuals, the
opposite is observed – the effect of information quality on trust is weaker (in Figure 3c,
the solid-line slope is gentler) while the effect of majority influence is stronger (in Figure
3d, the solid-line slope is steeper). Interestingly, all the slopes related to information
Page 18 of 25
quality are much gentler (see Figure 3a and 3c) compared to the slopes related to
majority influence (see Figure 3b and 3d), indicating that social media users tend to be
more strongly affected by majority influence than information quality. The implications
of this and other findings are discussed next.
Figure 3a
Figure 3b
Figure 3c
Figure 3d
Figure 3: Plots of Significant Moderating Effects
6. Discussion
This study set out to develop and empirically test a model that identifies (1) the different
information processing routes through which social media users form trust in
user-generated crisis information, and (2) the factors moderating individuals’ use of the
routes. Based on ELM, our proposed model considers two routes: central and peripheral.
Extending ELM with the MOA framework, we hypothesize that the use of the routes is
moderated by personal relevance, anxiety, and prior knowledge. Findings from a survey
indicate that individuals use the central route more when the crisis information has strong
Page 19 of 25
personal relevance or when their anxiety level is low. In contrast, individuals use the
peripheral route more when the crisis information has less personal relevance or when
their anxiety level is high. Contrary to our hypotheses, the moderating effects of prior
knowledge were not significant.
The insignificance of prior knowledge is unexpected considering that there has been
strong evidence for its role in information processing, as discussed in the hypothesis’s
justification. In retrospect, the unusual scale of the nuclear crisis in our study might have
led the respondents to believe that it could spin out of control and prior knowledge might
not be applicable, thereby limiting the effect of prior knowledge. Rather than concluding
that prior knowledge does not come into play when individuals assess crisis information
on social media, we suggest that it is necessary to test the proposed model further in
other types of crisis (e.g., flood, earthquake, civil unrest) and crises of different
magnitude. The implications of this and other findings for research and practice are
discussed next.
6.1. Implications for Research
This study contributes to theoretical development in several ways. First, the proposed
model identifies two important information processing routes through which individuals
form trust. Along with the central route which relies on information quality, the model
accounts for the peripheral route which relies on majority influence and is clearly
pertinent in the context of social media. According to our literature review, this is one of
the earliest IS studies to consider social influence in the formation of trust in
user-generated information. Second, our proposed model clarifies the factors moderating
the use of different information processing routes. We found that their use depends on
personal relevance and level of anxiety. This enhances our understanding of how the
formation of trust is shaped by individual factors. Third, our proposed model focuses on
a critical yet understudied context. Considering the prevalence of turning to social media
for information during crises, studying the formation of trust in the context can inform
the management of crises, such as the spread of false information on social media. Our
model accounts for factors such as personal relevance, anxiety, and majority influence,
which are highly relevant to the crisis context and medium. Fourth, the proposed model
was assessed in an empirical field survey set in a real crisis rather than a fictitious
scenario and realism was thus maintained.
Page 20 of 25
This study is limited in several ways that could be improved in future studies. First,
a complete list of the population (i.e., individuals who sought crisis information on social
media) was not available and random sampling was therefore not viable. The list is
unlikely to become available in the foreseeable future but the generalizability of our
findings can be enhanced by studying other samples, social media, and types of crisis.
Second, our proposed model accounted for only one each of the motivation, opportunity,
and ability factors. Since the findings largely support the moderating effects of
motivation and opportunity, future research can extend the model by considering other
relevant factors such as curiosity (a source of motivation) and time pressure (which could
limit the opportunity to process information).
The findings also suggest further research opportunities. The observation that
users tend to be more affected by majority influence (a peripheral route) is well-matched
to the nature of social media. This may reflect the general personality of social media
users – they can be characterized as having stronger external locus of control and are
therefore more easily swayed by social influence then those who seek information from
other media. Since social influence is prevalent in social media, more research on the
nature of the influence is needed. For example, some interesting questions include: what
are the personal characteristics of social media users who are likely to be influenced?
What are the informational characteristics of influential messages? What are the social
mechanisms through which users are influenced by user-generated information? How do
technological features (e.g., display of access statistics, naming and positioning of the
repost button) affect the extent of social influence? Since social media can potentially
spread false information and rumors during crises (Sutton et al. 2008), understanding the
nature of social influence can help to identify ways to manage undesirable influences.
This also narrows a gap in IS studies applying ELM, which often leave out the opinion of
others even though ELM identifies it as an important heuristic for processing information
(Petty and Cacioppo 1986).
6.2. Implications for Practice
Understanding the formation of trust has several practical implications for the
management of social media during crises. Trust is likely to be more accurate if it is
based on the central route of information processing (i.e., evaluation of information
quality). Therefore, the use of the central route should be promoted to curb the spread of
false information during crises. Our findings suggest that this can be achieved by
Page 21 of 25
increasing personal relevance and reducing the level of anxiety.
With regard to personal relevance, social media websites can organize and
present crisis information to users according to their proximity to the crisis. This can
increase the relevance of information to individual users and entice them to assess
information quality and thereby form a more accurate judgment of its trustworthiness.
For instance, for users who are geographically close to the location of the crisis,
information about evacuation should be emphasized; for users who are distant from the
crisis, information about the crisis’s broad and long-term implications could be
highlighted. Information can be personalized based on users’ personal profile (e.g.,
country of residence, occupation, education, and age), geographical location, as well as
browsing history. A combination of expert recommendation, peer recommendation, and
automated recommendation can be used to enhance accuracy of information
personalization.
To a certain extent, the level of anxiety can be reduced by modifying website
design in several ways. First, it has been observed that the color, background music, and
layout of website content can reduce arousal levels and elicit emotions such as
peacefulness, calmness, and hopefulness (e.g., Wu et al. 2008). During crises, these
elements of social media websites may be temporarily adjusted to reduce anxiety among
information seekers and thereby promote the use of the central route of information
processing. Second, messages signaling social support can be displayed to reduce the
level of anxiety. This is supported by studies that have observed a negative correlation
between social support and the level of distress after natural disasters (e.g., Cook and
Bickman 1990). Messages that convey information about emotional support (expressions
of assurance, affection, closeness), informational support (verified situation updates,
evacuation instructions), and tangible support (donation, shelter, transportation) can be
displayed in the form of banners to enhanced the perceived social support among
high-anxiety individuals.
7. Conclusion
This study recognizes the double-edged-sword quality of social media as an information
source during crises and the importance of forming accurate trust in user-generated
information. This departs from prior IS studies which have mainly focused on increasing
trust to promote IS behaviors (e.g., use of online shopping, adoption of new
Page 22 of 25
technologies). The proposed model addresses a gap in our understanding by shedding
light on the informational processes though which social media users form trust and how
the use of the processes is affected by individuals’ motivation and opportunity. This study
serves as a step stone for further inquiry into the consumption of user-generated
information and contributes to a more complete theorization of the phenomenon, which is
imperative as social media have become integral and even critical to many aspects of our
lives.
References
Acar, A. and Muraki, Y. 2011. "Twitter for Crisis Communication: Lessons Learned
from Japan's Tsunami Disaster," International Journal of Web Based Communities
(7:3), pp. 392-402.
Angst, C.M. and Agarwal, R. 2009. "Adoption of Electronic Health Records in the
Presence of Privacy Concerns: The Elaboration Likelihood Model and Individual
Persuasion," MIS Quarterly (33:2), pp. 339-370.
Anish, P., Carl, B., Mihaela, V., Barbara, A., and Doug, N. 2014. "Motives for
Reading and Articulating User-Generated Restaurant Reviews on Yelp.Com,"
Journal of Hospitality and Tourism Technology (5:2), pp. 160-176.
Apsler, R. and Sears, D.O. 1968. "Warning, Personal Involvement, and Attitude
Change," Journal of Personality and Social Psychology (9:2), pp. 162-166.
Ashcraft, M.H. 2002. "Math Anxiety: Personal, Educational, and Cognitive
Consequences," Current Directions in Psychological Science (11:5), pp. 181-185.
Bartle, C., Avineri, E., and Chatterjee, K. 2013. "Online Information-Sharing: A
Qualitative Analysis of Community, Trust and Social Influence Amongst
Commuter Cyclists in the UK," Transportation Research Part F: Traffic
Psychology and Behaviour (16), pp. 60-72.
Beck, A.T. and Clark, D.A. 1997. "An Information Processing Model of Anxiety:
Automatic and Strategic Processes," Behaviour Research and Therapy (35:1), pp.
49-58.
Bhattacherjee, A. and Sanford, C. 2006. "Influence Processes for Information
Technology Acceptance: An Elaboration Likelihood Model," MIS Quarterly (30:4),
pp. 805-825.
Burgess, S., Sellitto, C., Cox, C., and Buultjens, J. 2011. "Trust Perceptions of
Online Travel Information by Different Content Creators: Some Social and Legal
Implications," Information Systems Frontiers (13:2), pp. 221-235.
Chai, S. 2011. "Information Sharing on Blogosphere: An Impact of Trust and Online
Privacy Concerns," Asia Pacific Journal of Information Systems (21:3), pp. 1-18.
Chaiken, S., Liberman, A., and Eagly, A.H. 1989. "Heuristic and Systematic
Information Processing within and Beyond the Persuasion Context," in Unintended
Thought, J.S. Uleman and J.A. Bargh (eds.). Guilford Press.
Chaiken, S. and Maheswaran, D. 1994. "Heuristic Processing Can Bias Systematic
Processing: Effects of Source Credibility, Argument Ambiguity, and Task
Importance on Attitude Judgment," Journal of Personality and Social Psychology
(66:3), pp. 460-473.
Champion, V.L. 1984. "Instrument Development for Health Belief Model
Constructs," Advances in Nursing Science (6:3), pp. 73-85.
Page 23 of 25
Cheung, C.M.K. and Thadani, D.R. 2012. "The Impact of Electronic Word-of-Mouth
Communication: A Literature Analysis and Integrative Model," Decision Support
Systems (54:1), pp. 461-470.
Chin, W.W., Marcolin, B.L., and Newsted, P.R. 2003. "A Partial Least Squares Latent
Variable Modeling Approach for Measuring Interaction Effects: Results from a
Monte Carlo Simulation Study and an Electronic-Mail Emotion/Adoption Study,"
Information Systems Research (14:2), pp. 189-217.
Chu, S.-C. and Kim, Y. 2011. "Determinants of Consumer Engagement in Electronic
Word-of-Mouth (eWOM) in Social Networking Sites," International Journal of
Advertising (30:1), pp. 47-75.
Cialdini, R. 1993. Influence: Science and Practice. New York, U.S.: Harper Collins.
Clarke, R. 1999. "Control of Low-Level Radiation Exposure: Time for a Change?"
Journal of Radiological Protection (19:2), pp. 107-115.
Cook, J.D. and Bickman, L. 1990. "Social Support and Psychological
Symptomatology Following a Natural Disaster," Journal of Traumatic Stress (3:4),
pp. 541-556.
DeBono, K.G. and McDermott, J.B. 1994. "Trait Anxiety and Persuasion: Individual
Differences in Information Processing Strategies," Journal of Research in
Personality (28:3), pp. 395-407.
Diamantopoulos, A. and Winklhofer, H.M. 2001. "Index Construction with Formative
Indicators: An Alternative to Scale Development," Journal of Marketing Research
(38:2), pp. 269-277.
Diane, M.M. 1987. "Systematic and Nonsystematic Processing of Majority and
Minority Persuasive Communications," Journal of Personality and Social
Psychology (53:1), pp. 41-52.
Dickinger, A. 2011. "The Trustworthiness of Online Channels for Experience- and
Goal-Directed Search Tasks," Journal of Travel Research (50:4), pp. 378-391.
Dijkstra, J.J. 1999. "User Agreement with Incorrect Expert System Advice,"
Behaviour & Information Technology (18:6), pp. 399-411.
Doan, S., Vo, B.-K., and Collier, N. 2012. "An Analysis of Twitter Messages in the
2011 Tohoku Earthquake," in Electronic Healthcare, P. Kostkova, M. Szomszor
and D. Fowler (eds.). Springer Berlin Heidelberg, pp. 58-66.
Eysenck, M.W., Derakshan, N., Santos, R., and Calvo, M.G. 2007. "Anxiety and
Cognitive Performance: Attentional Control Theory," Emotion (7:2), pp. 336-353.
Flanagin, A.J. and Metzger, M.J. 2013. "Trusting Expert- Versus User-Generated
Ratings Online: The Role of Information Volume, Valence, and Consumer
Characteristics," Computers in Human Behavior (29:4), pp. 1626-1634.
Fornell, C. and Larcker, D. 1981. "Evaluating Structural Equation Models with
Unobservable Variables and Measurement Error," Journal of Marketing Research
(18:1), pp. 39-50.
Grimmer, D. 2011. "UK Riot Latest: Norfolk Police Blast Riot Rumours." Accessed
on Dec 1, 2011,
http://www.northnorfolknews.co.uk/news/uk_riot_latest_norfolk_police_blast_riot
_rumours_1_990163
Hair, J.F., Black, W.C., Babin, B.J., and Anderson, R.E. 2009. Multivariate Data
Analysis, (7th ed.). Upper Saddle River, New Jersey, USA: Prentice Hall.
Hamilton, M. 1959. "The Assessment of Anxiety States by Rating," British Journal
of Medical Psychology (32:1), pp. 50-55.
Page 24 of 25
Ho, S.Y. and Bodoff, D. 2014. "The Effects of Web Personalization on User Attitude
and Behavior: An Integration of the Elaboration Likelihood Model and Consumer
Search Theory," MIS Quarterly (38:2), pp. 497-520.
Ichiguchi, T. 2011. "Robust and Usable Media for Communication in a Disaster,"
Science & Technology Trends Quarterly Review (41), pp. 44-55.
Jarvis, C.B., MacKenzie, S.B., and Podsakoff, P.M. 2003. "A Critical Review of
Construct Indicators and Measurement Model Misspecification in Marketing and
Consumer Research," Journal of Consumer Research (30:3), pp. 199-218.
Kerstetter, D. and Cho, M.-H. 2004. "Prior Knowledge, Credibility and Information
Search," Annals of Tourism Research (31:4), pp. 961-985.
Kim, H.J., Song, S.-M., and Lee, H.G. 2009. "Understanding User Motivations and
Behavioral Process in Creating Video Ugc: Focus on Theory of Implementation
Intentions," Asia Pacific Journal of Information Systems (19:4), pp. 125-148.
Kim, Y.A. and Ahmad, M.A. 2013. "Trust, Distrust and Lack of Confidence of Users
in Online Social Media-Sharing Communities," Knowledge-Based Systems (37), pp.
438-450.
Kock, N. 2015. "One-Tailed or Two-Tailed P Values in PLS-SEM?" International
Journal of e-Collaboration (11:2), pp. 1-7.
Komiak, S.Y.X. and Benbasat, I. 2006. "The Effects of Personalization and
Familiarity on Trust and Adoption of Recommendation Agents," MIS Quarterly
(30:4), pp. 941-960.
Koo, C., Lim, M.K., and Park, K. 2014. "E-Smart Health Information Adoption
Processes: Central Versus Peripheral Route," Asia Pacific Journal of Information
Systems (24:1), pp. 67-94.
Kwak, H., Lee, C., Park, H., and Moon, S. 2010. "What is Twitter, a Social Network
or a News Media?" Proceedings of the 19th International Conference on World
Wide Web, pp. 591-600.
Lee, Y.W., Strong, D.M., Kahn, B.K., and Wang, R.Y. 2002. "AIMQ: A Methodology
for Information Quality Assessment," Information & Management (40:2), pp.
133-146.
Lenders, V., Koukoumidis, E., Zhang, P., and Martonosi, M. 2008. "Location-Based
Trust for Mobile User-Generated Content: Applications, Challenges and
Implementations," Proceedings of the 9th ACM Workshop on Mobile Computing
Systems and Applications, pp. 60-64.
Leung, X.Y. and Bai, B. 2013. "How Motivation, Opportunity, and Ability Impact
Travelers' Social Media Involvement and Revisit Intention," Journal of Travel &
Tourism Marketing (30:1-2), pp. 58-77.
Li, C.-Y. 2013. "Persuasive Messages on Information System Acceptance: A
Theoretical Extension of Elaboration Likelihood Model and Social Influence
Theory," Computers in Human Behavior (29:1), pp. 264-275.
Luo, W. and Najdawi, M. 2004. "Trust-Building Measures: A Review of Consumer
Health Portals," Communications of the ACM (47:1), pp. 108-113.
MacInnis, D.J., Moorman, C., and Jaworski, B.J. 1991. "Enhancing and Measuring
Consumers' Motivation, Opportunity, and Ability to Process Brand Information
from Ads," The Journal of Marketing (55:4), pp. 32-53.
Mak, B., Schmitt, B.H., and Lyytinen, K. 1997. "User Participation in Knowledge
Update of Expert Systems," Information & Management (32:2), pp. 55-63.
Martin, R., Gardikiotis, A., and Hewstone, M. 2002. "Levels of Consensus and
Majority and Minority Influence," European Journal of Social Psychology (32:5),
pp. 645-665.
Page 25 of 25
Mendoza, M., Poblete, B., and Castillo, C. 2010. "Twitter under Crisis: Can We Trust
What We RT?" First Workshop on Social Media Analytics, pp. 71-79.
Moturu, S. and Liu, H. 2011. "Quantifying the Trustworthiness of Social Media
Content," Distributed & Parallel Databases (29:3), pp. 239-260.
Nemeth, C.J. 1986. "Differential Contributions of Majority and Minority Influence,"
Psychological Review (93:1), pp. 23-32.
O'Reilly, C.A., III. 1982. "Variations in Decision Makers' Use of Information
Sources: The Impact of Quality and Accessibility of Information," The Academy of
Management Journal (25:4), pp. 756-771.
Oh, O., Kwon, K.H., and Rao, H.R. 2010. "An Exploration of Social Media in
Extreme Events: Rumor Theory and Twitter During the Haiti Earthquake 2010,"
International Conference on Information Systems pp. 1-15.
Petty, R.E. and Cacioppo, J.T. 1986. Communication and Persuasion: Central and
Peripheral Routes to Attitude Change. New York, U.S.: Springer-Verlag.
Sarason, I.G. 1988. "Anxiety, Self-Preoccupation and Attention," Anxiety Research
(1:1), pp. 3-7.
Shi, Z., Rui, H., and Whinston, A.B. 2013. "Content Sharing in a Social Broadcasting
Environment: Evidence from Twitter," MIS Quarterly (38:1), pp. 123-142.
Smith, C. 2011. "Twitter User Statistics Show Stunning Growth," The Huffington
Post. Accessed on 28 April 2015,
http://www.huffingtonpost.com/2011/03/14/twitter-user-statistics_n_835581.html
Sutton, J., Palen, L., and Shklovski, I. 2008. "Backchannels on the Front Lines:
Emergent Uses of Social Media in the 2007 Southern California Wildfires,"
Proceedings of the 5th International Information Systems for Crisis Response and
Management Conference, pp. 624-632.
Vieweg, S., Hughes, A.L., Starbird, K., and Palen, L. 2010. "Microblogging During
Two Natural Hazards Events: What Twitter May Contribute to Situational
Awareness," Proceedings of the ACM SIGCHI Conference on Human Factors in
Computing Systems, pp. 1079-1088.
Westerman, D., Spence, P.R., and Van Der Heide, B. 2014. "Social Media as
Information Source: Recency of Updates and Credibility of Information," Journal
of Computer-Mediated Communication (19:2), pp. 171-183.
Wilder, D.A. 1993. "The Role of Anxiety in Facilitating Stereotypic Judgments of
Outgroup Behavior," in Affect, Cognition, and Stereotyping: Interactive Processes
in Group Perception, D.M. Mackie and D.L. Hamilton (eds.). pp. 87-109.
Wu, C.-S., Cheng, F.-F., and Yen, D.C. 2008. "The Atmospheric Factors of Online
Storefront Environment Design: An Empirical Experiment in Taiwan," Information
& Management (45:7), pp. 493-498.
Yang, S.-C., Hung, W.-C., Sung, K., and Farn, C.-K. 2006. "Investigating Initial
Trust toward E-Tailers from the Elaboration Likelihood Model Perspective,"
Psychology and Marketing (23:5), pp. 429-445.
Yin, J., Lampert, A., Cameron, M., Robinson, B., and Power, R. 2012. "Using Social
Media to Enhance Emergency Situation Awareness," IEEE Intelligent Systems
(27:6), pp. 52-59.
Zax, D. 2011. "Tweeting in Japan: The Good, the Bad, and the Panicked," Fast
Company. Accessed on December 1, 2012,
http://www.fastcompany.com/1738847/the-uses-of-twitter-in-a-crisis
Zhou, T. 2012. "Understanding Users’ Initial Trust in Mobile Banking: An
Elaboration Likelihood Perspective," Computers in Human Behavior (28:4), pp.
1518-1525.