SIGCHI Conference Paper Format

8 downloads 0 Views 833KB Size Report
found a high correlation between aesthetics and usability. .... 57.9%, 24.5% and 8.2% had high school, bachelor and ..... [20] Lu, Y. and Rastrick, K. 2014.
The Interplay of Aesthetics, Usability and Credibility in Mobile Websites and the Moderation by Culture Kiemute Oyibo University of Saskatchewan Saskatoon, Canada [email protected] ABSTRACT

The relationships between aesthetics, usability and credibility have been widely investigated in humancomputer interaction (HCI). However, in the mobile domain, limited empirical evidence exists showing the interplay among these three constructs and the role culture plays. To address this, we carried out a survey on four systematically designed mobile websites among 233 subjects from Canada and Nigeria, which belong to low- and high-context cultures respectively. Using path modeling, we investigated the relationships among the perceptions of these three HCI constructs and the possible differences that exist between the two cultures. Our results show: 1) it is the perception of aesthetics that predominantly drives the judgment of mobile web credibility at the global and subgroup levels, with the effect size being greater for the Canadian group than for the Nigerian group; 2) the direct effect of aesthetics on usability is moderated by culture, with the magnitude being stronger for the Nigerian group than for the Canadian group; and 3) our models explain 29-83% of the variance of credibility. Author Keywords

Model; mobile web; aesthetics; usability; credibility; culture ACM Classification Keywords

H.5. Information interfaces and presentation; H.5.2 User Interfaces (D.2.2, H.1.2, I.3.6): Evaluation/methodology. INTRODUCTION

In the HCI research community, in respect of user interface (UI) design, there has been an ongoing debate regarding: 1) which of aesthetics and usability is more important in predicting users’ judgment of credibility [21]; and 2) whether what is beautiful is usable [38]. Despite abundant research in the web domain showing that the credibility of a website is predominantly determined by the perception of aesthetics, some researchers and practitioners still disagree, often citing unattractive sites which are popular, e.g., Craigslist, as an Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected]. IHC '16, October 04-07, 2016, São Paulo, Brazil © 2016 ACM. ISBN 978-1-4503-5235-2/16/10…$15.00 DOI: http://dx.doi.org/10.1145/3033701.3033711

Julita Vassileva University of Saskatchewan Saskatoon, Canada [email protected] example [22]. Some even argue that aesthetics is not important if there is good usability. Yet, others argued that aesthetics in HCI should only be considered and measured in terms of how information processing is facilitated, and not how the user is engaged in a pleasing experience [36]. This tends to relegate aesthetics to the background [17], making it important for more systematic research in a different platform than the web in order to ground the findings. In prior studies, products or systems are scarcely chosen systematically [13], as existing ones are mostly used, preventing the possibility for the systematic variation of UI features [21]. Moreover, fewer studies have focused on the mobile domain [20] and the role culture plays. To bridge this gap, we carried out a study on the perception of four systematically modified mobile sites among 233 subjects from a low-context (LC) culture (Canada) and a high-context (HC) culture (Nigeria). A HC culture is that type of culture which communicates in a style in which much is left unsaid by the sender and expected to be inferred by the receiver from the context. On the other hand, a LC culture is that type of culture in which information is communicated by the sender to the receiver explicitly, with little or nothing left to be inferred from the context [11]. Research has shown that most western countries (e.g., Germany, Canada, etc.) are LC cultures; while most African countries (e.g., Nigeria, Ghana, etc.), Asian countries (e.g., China), and South American countries (e.g., Brazil) are HC cultures. By using two different types of cultures, we set out to investigate whether both notions—1) aesthetics is more important than usability in predicting credibility and 2) aesthetics has a direct effect on usability—hold among a homogeneous as well as a culturally heterogeneous population. Also, we aimed to investigate the notion that aesthetics may become unimportant in predicting credibility if there is good usability [22]. In summary, our findings show: 1) it is the perception of aesthetics that predominantly drives mobile web credibility at the global and subgroup levels; 2) the effect size of aesthetics on credibility is greater for the Canadian group (CG) than for the Nigerian group (NG); 3) the direct effect of aesthetics on usability is moderated by culture, with that of the NG having a higher magnitude than the CG; and 4) our models account for 29-83% of the variance of credibility. The rest of this paper is organized as follows. Sections 2 and 3 focus on related work and method respectively. Sections 4 and 5 dwell on result and discussion respectively. Finally, Section 7 focuses on conclusion and future work.

RELATED WORKS

In this section, we provide a conceptual background on the three HCI constructs of interest and review the extant literature dealing with their empirical relationships. Conceptual Background

Aesthetics and usability, according to Tractinsky et al. [38], represent two important and independent dimensions of HCI. While the definition of usability may be easy, that of aesthetics is not. Aesthetics is often regarded as an abstract concept in HCI and a controversial topic in arts. Hence, information technologists and artists possess different points of view of aesthetics [1]. Moreover, while usability in HCI can be relatively measured using objective metrics (e.g., efficiency), aesthetics cannot, as it refers to non-quantifiable, subjective, and emotion-based perception and/or experience of a system [3], [38]. Against this backdrop, in the rest of this subsection, we provide a brief definition and explanation of each construct, including credibility, within the context of HCI and our study, to set the tone for the rest of the paper. Aesthetics

Aesthetics, as cited in [17], can be defined as “an artistically beautiful or pleasing appearance” (p. 272). In HCI, it is often conceptualized as having two dimensions: classical and expressive. Classical aesthetics captures the traditional notion of aesthetics, which is often associated with terms such as “clear,” “clean,” “well-organized,” etc. These terms, according to [37], reflect “orderly and clear design,” which is closely related to a large number of the design rules proposed by usability experts. Expressive aesthetics, on the other hand,

reflects the creativity and expressive power of the designer, and is often associated with terms such as “originality,” “fascinating design,” “special effects,” etc. [17]. Usability

Usability can be defined as “the effectiveness of interaction between human operators and their machines” (p. 60) [3]. It is conceptualized as comprising a number of related terms, such as “user-friendly,” “ease-of-use,” “intuitive,” etc. It can be measured either objectively or empirically. Objectively, it is measured based on “how well intended users can master and perform tasks on the system” (p. 60) [3], e.g., task completion time [13]. Empirically, it is measured by “observing and recording actual user performance of frequent or critical tasks on the system” (p. 60) [3]. Credibility

Credibility can be defined as believability, which is based on perceived trustworthiness and perceived expertise of the producers or designers of a product [6]. In the context of our study, credibility is subjective (based on users’ perceptions). Relationships between HCI Constructs

The beauty and usability of a website are key to its credibility and success and, by extension, those of its sponsors. As a result, a lot of research [2], [13], [17], [18], [19], [38] has been focused on investigating the links between these three important constructs of HCI (aesthetics, usability and credibility) from the standpoint of first-time users. Owing to

the many debates, which the relationships among these three constructs have elicited in the literature, a number of studies have made efforts on investigating which of aesthetics and usability is more important in determining the credibility or intention of use of a website, especially in domains such as banking [20], government [40], [15] and health [32] . On one hand, some researchers have argued and provided empirical evidence showing that aesthetics is more important than usability in users’ judgment of credibility. Robins and Holmes [29], for example, found that the effect of aesthetics on website credibility occurs “within the first few seconds in which a user views a web page” (p. 386), and when the same content was presented to users at different levels of aesthetic treatment, the one with better aesthetic treatment elicited better credibility rating. Norman [25], who was once a leading advocate of usability considerations in HCI design in the 1990s, became later convinced in 2004 [24] that aesthetics, which he described as visceral, plays an important role in the purchase of everyday products [25]. On the other hand, other researchers have found that usability also plays a vital role in the judgment of web credibility [33], [40]. Some [14], [23], [39] have even argued or implied that, in general, usability is more important than aesthetics in predicting the credibility or intention of use of a system, with Nielsen advocating that more emphasis be placed on usability in designing products to facilitate and sustain their use [24]. Still, others, e.g., Kurosu and Kashimura [16], have found a high correlation between aesthetics and usability. Based on their ATM-based study, where they classified usability into “apparent” and “inherent,” they advised designers to strive to focus not only on inherent usability but apparent usability (correlating with perceived aesthetics) as well, as they found less correlation between both types of usability. They argued that there was no use talking about the inherent usability if apparent usability failed to motivate and convince users to use a system in the first place. Further, in an experimental context, Tractinsky et al. [38] found that “what is beautiful is usable” (p. 127). However, Hassenzahl [12] could not replicate this finding, and thus rejected this notion, claiming Tractinsky et al.’s research suffered from methodological problems. Tractinsky et al. had argued and shown that users perceived aesthetics early in the HCI process, which, due to the “halo effect,” later influenced their judgement of usability. However, Li and Yeh [18] in a study of m-commerce websites using 200 subjects and structural equation modeling, found no significant impact of “design aesthetics” on “perceived usefulness” or “ease of use.” Thus, the mixed results necessitates more research in a different domain than the web and among a culturally different population in order to enrich the literature. So far, in the mobile domain, limited research has investigated the relationships among the three constructs in a systematic fashion [20], and how they differ between a LC and a HC culture. Moreover, most prior studies have been concentrated on the Western and Asian cultures. It is this gap we attempt to fill using a mixed sample from Canada and Nigeria.

METHOD

In this section, we present our research method, including the research design, the instruments used for measuring the three constructs under investigation and the demographics of the surveyed participants. Research Design

The aim of our study is to answer the research question: How do the three important web user interface constructs (aesthetics, usability and credibility) influence one another in the mobile domain? Does this depend on culture and UI design? To answer the above research question, we designed four systematically modified (hypothetical) mobile websites (see Figure 1) using a transformation framework we called artifact-action framework [26]. The x- and y- axes represent UI treatment (or actions) carried out on an artifact (in a given quadrant) to realize a new artifact (in the next quadrant) in a clockwise fashion. For example, to produce a new UI design B from A, we carry out the action “make gray and add icon.” Next, to produce C from B, we carry out the action “make unicolor.” This UI transformation continues till we arrive at A from where we started. Our survey was submitted to and approved by our university’s ethics board, after which it was posted on the university’s online bulletin and social networks, and emailed to potential subjects as well for anonymous participation. Participants were offered the chance of winning a gift card of $50. Further, to model our collected data, we built a hypothetical path model (Figure 2), informed by prior research. For example, according to [9], “users draw upon aesthetic factors to judge usability and credibility” (p. 1). In other words, aesthetics influences both usability and credibility. Additionally, research has shown that the perception of usability influences credibility as well [33], [40]. Figure 2 is a hierarchical component model built using a two-step approach [30]. Usability and credibility were directly measured, while aesthetics, a high-order construct, was indirectly measured using its two-dimensional low-order constructs: classical aesthetics and expressive aesthetics. Based on the path model, we formulated our hypotheses as follows: H1: Aesthetics (A) will be more important than Usability (U) in predicting Credibility (C) at the global and subgroup levels. In other words, AC > UC. H2: The magnitude of influence of Aesthetics on Credibility and Usability will be higher for the Nigerian group than for the Canadian group. In other words, A C (NG) > AC (CG) and AU (NG) > AU (CG). H3: The magnitude of influence of Usability on Credibility will be higher for the Canadian group than the Nigerian group. In other words, UC (CG) > UC (CG). H4: The mediation of the direct effect of Aesthetics on Credibility by Usability will be higher for the grid-layout mobile webpage than the list-layout mobile webpages for both cultures. In other words, UC (D) > UC (A, B, C).

H3: The level of influence of Usability on Credibility is

Figure1. Systematically designed user interface

Figure 2. Hypothesized path model

The first hypothesis (H1) is due to the fact that most prior studies [7], [18] in the web domain have shown that aesthetics is more important to users than usability in the judgment of credibility. Our second and third hypotheses (H2 and H3) are based on UI research by Sun [34], who found that HC cultures, such as Brazil and China, prefer visual and colorful websites, suggesting that they care more about aesthetics, while LC cultures, e.g., Germany, prefer hierarchically organized websites, suggesting they care more about usability. Thus, in the mobile domain, we surmised that aesthetics perception, on one hand, would impact credibility and usability perceptions for the NG more than it does for the CG since the former belongs to a HC culture, which tends to prefer visual and colorful to minimalist

websites (aesthetics). On the other hand, we surmised that usability perception would impact credibility perception for the CG more than the NG since the former belongs to a LC culture, which tends to care more about organization of content (usability) to appearance [34]. In other words, we hypothesized the higher the preference for a given type of website over another (an aesthetic website over a usable website, or vice versa) by a given culture, the higher the corresponding construct would influence credibility for that culture than the other. Lastly, the fourth hypothesis (H4) is informed by the psychological “decoy effect” phenomenon, which research has shown, affects users’ judgment in ecommerce. The “decoy effects indicate that in a decision situation a certain target option (product) T can look much nicer compared to a certain competitor option C when a decoy element D is added to the set of alternative items” (p. 324) [35]. In our case, T is Usability, C is Aesthetics, and D is the grid layout design, which we intended to use to shift the attention of participants from Aesthetics towards Usability as the predictor of Credibility. Due to the fact that mobile webpage D has a different layout (grid) from the rest webpages (list), coupled with the fact that we presented it last to participants in the survey, we surmised that its unique layout (grid) would catch the attention of participants and, as a result, Usability might have more impact on the Credibility of D than it would on the Credibility of A, B and C. Measure

To measure the three constructs of interest with respect to the four UI designs in Figure 1, we adopted previously validated scales. Aesthetics was measured using the 6-item validated version of Lavie and Tractinsky’s aesthetics scale [17] by Schaik and Ling [31]. Usability was measured using Lavie and Tractinsky’s 5-item scale [17]. It has been validated as well in a previous study [32]. Each item in both scales was based on a 7-point Likert scale, which ranged from Strongly Disagree (1) to Strongly Agree (7). Finally, credibility was measured using a single-item scale, adopted in a prior study [32]. It is based on a 7-point Likert scale, which ranged from Very Bad (1) to Very Good (7). The four mobile webpages in our online survey were presented to participants in this order (C, A, B, D). However, the blocks of items in the aesthetic and usability scales were not administered to participants in an orderly manner (i.e., one block followed by the other). Rather, they were mixed up and presented in a randomized fashion in a single block in order to avoid revealing to participants the exact constructs we are measuring if we were to use two separate blocks of items for both constructs. Participants

A total of 233 participants (146 Nigerians and 87 Canadians) took part in the online survey. Table 1 shows the participants’ demographics: gender, age, internet experience and highest educational qualification. Overall, 54.5% of the participants were males, while, 45.5% were females. Moreover, 53.2% have up to and over 10 years of internet experience and about 57.9%, 24.5% and 8.2% had high school, bachelor and postgraduate qualifications respectively.

Criterion Gender Age Years on Internet Qualification

Group Male Female 18-24 >24 Unidentified =10 High School Bachelor Postgraduate Others

CG 19 68 48 39 0 10 77 35 37 10 5

NG 108 38 110 32 4 99 47 100 20 9 17

No. 127 106 158 71 4 109 124 135 57 19 22

Percent 54.5% 45.5% 67.8% 30.4% 1.7% 46.8% 53.2% 57.9% 24.5% 8.2% 9.4%

Table 1. Participants’ demographics RESULT

In this section, we present our result using R’s Partial Least Square Path Modeling (PLSPM) package [8], the validation of the global measurement and structural models, and the multi-group analysis (MGA). Evaluation of Global Measurement Model

We began our PLSPM by assessing the measurement models to ensure indicator reliability, internal consistency reliability, convergent validity and the discriminant validity criteria are met before proceeding to analyze the respective structural models [30], [10]. Due to limited space, we are unable to present the table showing the values of the assessed criteria. Indicator Reliability

Except for “pleasant” in classical aesthetics for webpage B with a value of 0.66, all the indicators in our measurement models had an outer loading greater than 0.7 [10]. Thus, the reliability criterion, to a great degree, was met, as the communality value for all indicators were greater than 0.5, except for the “pleasant” indicator. Internal Consistency Reliability

The internal consistency reliability of the constructs was evaluated using the composite reliability criterion, DG.rho (ρ). The ρ values for all constructs with respect to all four mobile web designs were greater than 0.7 [10]. Convergent Validity

The Average Variance Extracted (AVE) was used to assess convergent validity. The AVE for each model construct was above the recommended value, i.e., greater than 0.5 [10]. Discriminant Validity

The crossloading of each construct was also assessed. We found no indicator that loaded higher on any other construct than the one it was meant to measure [10]. Verification of First Hypothesis (H1)

We verified the first hypothesis (AC > UC) in two stages. First, we analyzed the global model. Second, we analyzed the subgroup models. The global analysis was meant to verify whether the significant influences (direct effects) at the subgroup level could be replicated in order to uncover/confirm those influences that transcended cultures.

Global model

Figure 3 shows the global model in two forms. Figure 3a shows the partial least square path model (PLS-PM) [30] for all four mobile webpages (A, B, C and D), represented by red, black, blue and green colors respectively, while Figure 3b replicates the influences in a line graph for easy visualization of the trend as we transition from one UI to another. We compressed all four global models for all four webpages into one overarching model for compactness, quick and easy comparison among all four models. The model shows the influences exerted by the exogenous constructs on the endogenous constructs; the coefficients of determination (R2) of the endogenous constructs, i.e., the amount of variance of an endogenous construct explained by the exogenous constructs; and the goodness of fit (GOF). The GOF represents the predictive power of each model or how well each model fits its data. The number of asterisks indicates how significant each of the direct effects is. It ranges from one to four asterisks, representing the p-value of 0.05, 0.01, 0.001 and 0.0001 respectively. The p-value of 0.05 indicates our chosen minimum confidence level (α = 95%) regarding the direct effects or path coefficients (β). Figure 3a shows that the global models fit their respective data well due to the high GOF > 0.7 for all four models, with webpages A and D having the highest values of 0.78 and 0.77 respectively and webpages B and C having the least values of 0.73 and 0.74 respectively. The R2 for Usability and Credibility ranges from moderate to high for all four models. Based on the PLS-PM standard, R2 < 30 is low, 0.30 < R2 < 0.50 is moderate and R2 > 0.60 is high [30]. For Credibility, the R2 values range from 0.48 (moderate) for webpage D to 0.70 (high) for webpage A. Further, we see that the β’s for AC are significantly greater than the corresponding path coefficients for UC for all four webpages, as clearly shown in the line graph in Figure 3b. For example, for webpages A and B, the β’s for AC are 0.76 and 0.78 respectively (significant at p AU (CG), is validated, as we see all four differences, regarding AU, between the two groups are significant at p UC (NG). Just like AC in the second hypothesis, in Figure 4b, we see that UC for the CG is not consistently higher than the UC for the NG for all four webpages; rather it alternates with that of the NG. Also, except for the significant difference between the UC for the CG and the NG for webpage C, the other differences, for webpages A, B and D, are not significant (Table 2). Therefore, our H3 is not validated. Verification of Fourth Hypothesis (H4)

To verify our fourth hypothesis, we carried out mediation analysis by computing the variance accounted for (VAF), i.e., the ratio of the significant indirect effect to direct effect [10]. As shown in the global model (Figure 3a), the indirect path UC is not significant for all four mobile webpages, so Usability does not mediate the direct effect of Aesthetics on Credibility (AC) at the global level. However, we find that for a number of the webpages in the subgroup model (Figure 4a), the indirect path UC is significant alongside AU, indicating the possibility of usability mediation. This led us to compute the VAF for mobile webpages A, C and D for the CG and for mobile webpages B and D for the NG, where both AU and UC are significant. Figure 5 shows the VAF values. According to [10], VAF < 0.2 indicates no mediation, 0.2 ≤ VAF ≤ 0.8 indicates partial mediation and VAF > 0.8 indicates full mediation. Thus, for the CG, Usability partially mediates the direct effect of Aesthetics on Credibility for webpages A, C and D given that their VAF values are 0.27, 0.26 and 0.21 respectively. Similarly, for the NG, Usability partially mediates the direct effect of Aesthetics on Credibility for webpages B and D given that their VAF values are 0.23 and 0.73 respectively, with the latter having the strongest partial (almost full) mediation. Thus, our hypothesis that Usability would mediate the influence of Aesthetics on Credibility for D more than for the other UIs is not supported at the global level. However, at the subgroup level, it is only supported for the NG but not for the CG. One possible explanation is that the higher the perceived Usability of a UI the higher may be its mediation

Figure 5. Usability mediation effect of aesthetics on credibility

of the direct effect of Aesthetics on Credibility. In our prior study [26], we found that the NG, unlike the CG, perceived the usability of webpage D (with grid layout) significantly higher than that of the other three designs (with list layout). DISCUSSION

The result of our study shows that it is the perception of aesthetics, for the most part, that drives the judgment of credibility of a mobile website, irrespective of culture. This is consistent with previous findings [7], [18], [19], [29]. Our result also shows that usability perception is highly influenced by aesthetics perception in the mobile domain, which seems to lend credence to the notion that what is beautiful is usable [38] within the context of UI perception. Our result also indicates that Usability can mediate the influence of Aesthetics on Credibility if the perception of usability is high. Moreover, according to [28], it has become very important for researchers to report the effect sizes (f2 or ES) of driver constructs on the target construct whenever possible. Effect size, according to Cohen, is the extent to which the phenomenon of interest is present in the population [4]. We computed the effect sizes of the driver constructs to measure how much effect Aesthetics and Usability have on Credibility at the global and subgroup levels. Figure 6 shows a line graph of the effect sizes for the driver constructs on Credibility. It clearly shows that at both global and subgroup levels, it is the perception of aesthetics that predominantly drives credibility. At both levels, Aesthetics (A) has a medium to very large effect size on Credibility (C) for all four UIs, except for webpage D for the NG, where Usability (A) mediates the influence of Aesthetics on Credibility far more than for the CG and any other webpage (see Figure 5). Based on Cohen’s guidelines [4], effect sizes of 0.02, 0.15 and 0.35 are regarded as small, medium and large respectively. In the global model, we see that the ES’s for AC (i.e., ES_AC in Figure 6) are 0.89, 0.94, 0.69 and 0.23 for webpages A, B, C and D respectively. Similarly, in the CG subgroup model, the ES’s are 0.46, 0.79, 0.38 and 0.61 respectively, while in the NG subgroup model, the ES’s are 0.32, 0.31, 0.27 and 0.004 respectively. The ES values at the global level, except for that regarding D, happen to be higher than the corresponding subgroup values, confirming once again that it is aesthetics perception that drives mobile web credibility at the global level. However, regarding UC at both global and subgroup levels, the ES’s are small, ranging from approximately zero (for the global webpages) to 0.16 (for the CG’s webpage A).

Figure 6. Effect sizes of driver construct on credibility

The medium-to-large ES’s for AC and the zero-to-small ES’s for UC at the global and subgroup levels confirm that our first hypothesis (AC > UC) is supported. Regarding our second hypothesis [AC (NG) > AC (CG) and AU (NG) > AU (CG)], we find that it is only the second part of the hypothesis (perception of aesthetics influences usability for the NG more than for the CG) that is supported. This indicates that in judging usability for all four webpages, the way the NG rated might have been influenced by the perception of aesthetics more than it had for the CG. This indicates how important aesthetics is to a LC culture (as found by Sun [34] and Oyibo et al. [26]) and how it can strongly influence the perception of other constructs, such as Usability. However, regarding effect size, we find that the ES of Aesthetics on Credibility (ES_AC) for all four webpages are stronger for the CG than for the NG. The CG’s effect sizes [ES_AC (Can)] are all large; while the NG’s effect sizes [(ES_AC (Nig)] range from approximately zero for mobile webpage D to medium for mobile webpages A, B and C. This, coupled with the influence AU in Figure 4b, indicates where the CG and NG majorly differ with respect to the perception of aesthetics determining credibility or usability. While the direct effect (AU) is significantly higher for the NG than for the CG [AU (NG) > AU (CG)], the effect size (ES_AC) is higher for the CG than for the NG [ES_AC (CG) > ES_AC (NG)]. This suggests that the perception of aesthetics influences usability for a HC culture more than for a LC culture, while the effect size of Aesthetics on Credibility is higher for the latter than the former. We controlled for internet experience. We conducted MGA for only participants having >=10 years of internet experience. The results turned out to be the same. Regarding our third hypothesis, we hypothesized that UC (CG) > UC (NG), but Figure 4b shows that this hypothesis is not supported, as we do not have an all significantly higher UC for the CG than for the NG. Also, we did not see a significant difference between UC (CG) and UC (NG), except for mobile webpage C where the influence is positive (0.30) for the CG but negative (-0.33) for the NG (see Table 2). However, with respect to effect size, we notice a little difference between both subgroups, with the CG having a more substantial magnitude. For example, for the CG, the ES_UC regarding webpages A and C are 0.16 and 0.13, which are medium and almost medium respectively, while those for the NG are 0.019 and 0.004 respectively, amounting to no effect size. While this is not sufficient for us to claim that ES_UC (CG) > ES_UC (NG), the fact that ES_UC for A, C and D are higher for the CG than for the NG (see Figure 6), coupled with ES_AC for all four webpages being higher for the CG than for the NG, may indicate that the effect sizes of the two driver constructs are generally stronger for a LC than a HC culture. However, this calls for further empirical investigations. Regarding our fourth hypothesis [UC (D) > UC (A, B, C)], we could only validate it for the NG. We discuss in the next paragraph some possible reasons why this hypothesis was only validated for the NG, but not the CG.

Furthermore, regarding the first hypothesis (AC > UC) for webpage D, the inability for us to validate it for the NG does not mean that perception of aesthetics does not influence credibility in the least for the NG. Rather, it only indicates, technically, such an influence is highly mediated by the perception of usability, as we saw in the validation of H4 [UC (D) > UC (A, B, C)] for the NG and evident in the high value of VAF of 0.73 (see Figure 5). This finding is plausible given that webpage D has a layout uniquely different from the others, which we intended, within the context of decoy effect, to use to shift the attention of participants from aesthetics towards usability concerns. This layout in our previous papers on users’ perceptions shows that participants (CG and NG [26]), and males and females [27] alike) preferred the grid layout over the list layout. Thus, they rated D higher on most criteria, even when the only visual and structural difference between D and C (its high-level-design counterpart) is layout. As a result, it is not surprising that, for webpage D, Usability is mediating the influence of Aesthetics on Credibility for both the NG and the CG, with the magnitude being higher for the former, perhaps because it perceived D more usable than the latter did and the other three webpages [26]. According to [10], when the indirect paths (AU and UC) are controlled for (i.e., are included in the path model), a previously significant path changes its value significantly and/or even becomes non-significant as we see regarding webpage D. From our effect size analysis, when we excluded Usability from the path model for mobile webpage D for the NG, the direct path between Aesthetics and Credibility (AC) was significant (β=0.5) at p