Evaluation criteria for blog design and analysis of causal relationships using factor analysis and DEMATEL

(1)

Evaluation criteria for blog design and analysis of causal relationships using

factor analysis and DEMATEL

Chun-Cheng Hsu

⇑

Department of Communication and Technology, National Chiao Tung University, Hsinchu 300, Taiwan, ROC

a r t i c l e

i n f o

Keywords:

Blog design Factor analysis DEMATEL

Multiple criteria decision-making (MCDM)

a b s t r a c t

The purpose of this study is to find the key factors influencing blog design, and explore the causal rela-tionships between the criteria for each factor. Since design is a multiple criteria decision-making (MCDM) problem, this study adopts a model which is a hybrid of factor analysis and the Decision Making Trial and Evaluation Laboratory method (DEMATEL). The DEMATEL method is used to simplify and visualize the interrelationships between criteria in making a decision. This study found five core factors that influence blog design: visual clarity, interface and usability, content and searchability, programming, and sociabil-ity. In addition, the key criteria for each factor were identified and the impact-relation maps obtained. The results of this study can provide useful guidance to blog designers for developing better blog platforms.

1. Introduction

The phenomenal universality of the Internet has catalyzed the adoption of blog culture into popular culture. Advancements in technology in the form of an abundance of innovative user-friendly applications have allowed it to continuously transform. These applications are of such simplicity, that the growth in popularity of blogging among keen trendsetters is unstoppable. A ‘‘blog’’ or ‘‘weblog’’ is a web page that displays a short journal. An important characteristic of a blog is that its entries are time-stamped and ar-ranged in chronological order (Blood, 2000). The term ‘‘weblog’’ was coined by John Barger in 1997, and later contracted to ‘‘blog’’ by Peter Merholz in 1999. Those who write blogs are called ‘‘blog-gers’’, and ‘‘blogging’’ refers to the act of writing a blog. Another popular term, ‘‘blogosphere,’’ describes the collection of the world’s blogs, often connected to each other in hyperlinked chains. It also collectively refers to blogs, bloggers and online communities in their entirety, and as such is a community or social concept. Four characteristics of blogs are: (1) they are personalized by individu-als, and usually informal in style; (2) are web-based, facilitating fast updates, easy management, and viewed with web browsers; (3) they are built using platforms which allow interconnection to form communities; and (4) they are automated without the use of HTML scripting languages, allowing the blogger to concentrate on the blog content (Du & Wagner, 2006).

In the context of the Internet era, ‘‘human–computer interac-tion’’ has become an important topic in both research and practice. In the past, research focused on the functionality of web pages, but has expanded to include exploration of user’s emotional experi-ence. Furthermore, many studies have demonstrated the relative importance of users’ emotional involvement in websites compari-son with usability.Mullet and Sano (1995)suggested that the uti-lization of graphical elements in visual interface design enhances the attractiveness of websites, and makes them more enjoyable to use.Tractinsky, Cokhavi, Kirschenbaum, and Sharﬁ (2006) indi-cated that users’ emotional enjoyment positively inﬂuences users’ responses to an interface design, and makes them more tolerant of problems with usability, for example being more willing to wait for web pages to download (Preece, Rogers, & Sharp, 2006).

Most evaluation models do not evaluate blog designs from an overall perspective and ignore the interrelationships between cri-teria. Designed to address problems in the ﬁeld of decision-making, the Decision Making Trial and Evaluation Laboratory method (DEMATEL) uses the knowledge of experts to lay out the structural model of a system. It not only provides a way to visualize causal relationships between criteria through an impact-relationship map but also indicates the degree to which criteria inﬂuence each other (Liou, Tzeng, & Chang, 2007). This study uses a combination of factor analysis and DEMATEL to map out the complex relation-ships between criteria and to identify key criteria in blog design.

The goals of this research are as follows: (1) identify the deter-minant criteria inﬂuencing blog design; (2) ﬁnd the causal rela-tionships between the evaluation criteria for each factor and degrees of interrelatedness. The remainder of this study is

doi:10.1016/j.eswa.2011.07.006

⇑ Tel.: +886 935 405 721; fax: +886 3 5509658. E-mail address:[email protected].

Contents lists available atScienceDirect

Expert Systems with Applications

(2)

organized as follows. In Section 2, the literature review is pre-sented. In Section3, the research methods are discussed. In Sec-tion4, an empirical study analysis is provided. Finally, Section5

contains the conclusion and possible directions for future develop-ment of the research.

2. Literature review

This study aims to integrate design evaluation into a decision-making cause and effect model. A thorough literature review is conducted, mainly focusing around two central themes. The ﬁrst theme involves studies on webpage and blog design, while the sec-ond involves the utilization of hybrid models in the ﬁeld of deci-sion-making.

2.1. Web and blog design evaluation

Numerous research efforts have explored this multifaceted dis-cipline, proposing research methods that can be divided into sev-eral types. The simplest method is to conduct a quality evaluation survey such as a service satisfaction survey while a more complex research method would involve utilizing the pro-posed models to scrutinize the relationships between variables; such a method would probably apply the Technology Acceptance Model (TAM). Guo and Salvendy (2009) studied Chinese online shopping websites, using a survey to pinpoint 15 key elements that determined users’ degree of satisfaction. These elements included quality content, service content, and appearance description. Through identifying these key elements, they sought to assist web designers in preparing content for e-business websites that enhances the user’s experience. Nielsen and Tahir (2002) ap-proached the problem from the usability perspective, and sug-gested 113 criteria for webpage design.

Furthermore, in considering the application of the Technology Acceptance Model (TAM) in explaining intranet usage, Horton, Buck, Waterson, and Clegg (2001)found that factors such as use-fulness, ease of use, and intention determined user acceptance.

Lee and Kim (2009)went further by including external factors in examining which elements influence users’ acceptance of a web-site. They identified external factors such as technical support, web experience, task equivocality, and the perceived ease of use as influencing intranet use.

Numerous studies have paid much attention to webpage design, but in comparison, studies on blog design and its visual elements are lagging far behind. Existing researches such asGuan and Liu (2007) have evaluated the usability of blog designs and advised the urgent need for improvement. Hsu and Lin (2008)affirmed the importance of technology and design in blogs, adding that the degree to which bloggers interact is positively associated with the enjoyment they derive from using appropriately designed interfaces. Such intrinsic motivation is an important consideration. In order to narrow the scope of this research, it is necessary to restrict it solely to personal blogs to prevent ambiguity in its find-ings. It explores which key design factors bloggers believe to be important in blog design. On one hand it refers to evaluation crite-ria for web page design, while on the other it applies DEMATEL for the first time to analyze blog design.

2.2. Hybrid model for decision-making

In recent years, many scholars have favored the application of a multiple criteria decision-making (MCDM) approach to problem-solving in the discipline such as social science. This is to overcome inherent limitations that are inevitable in single method models. The multifaceted approach provided by MCDM can often help

identify ingenious resolutions to complex problems and offer am-ple explanations and suggestions. For examam-ple,Tzeng, Chiang, and Li (2007)applied the hybrid MCDM model to analyze the effective-ness of e-learning programs. This hybrid MCDM model they used was based on factor analysis, DEMATEL, fuzzy theory, and the ana-lytic hierarchy process (AHP) method. In concluding their study, the interrelationships between independent criteria and the weights of these criteria were identiﬁed, providing developers with a useful reference for the design of e-learning program websites.

Tseng (2009)used fuzzy measure and the extension of DEMATEL

to integrate hotel service quality perceptions. Chen and Chen

(2010) developed an innovation support system for evaluating

the operative performance of Taiwan’s tertiary education institu-tions. They used a methodology based on DEMATEL, the fuzzy ana-lytical network process (FANP) and the technique for order preference by similarity to an ideal solution (TOPSIS). The results of the analysis served as benchmarks guiding these institutions to-wards maximizing innovation and creativity in structuring better education systems.

DEMATEL was developed in 1971 by the Science and Human Af-fairs Program at the Battelle Memorial Institute’s Geneva research center (Lee, Li, Yen, & Huang, 2010). It has been widely applied to complex problems in scientific disciplines such as technology, environmental science, and anthropology. Conventional quantita-tive analyses rely more on statistical analysis and applications of linear structure models. A good example of this is factor analysis, which extrapolates the degree of user satisfaction from surveys and interviews (Tzeng et al., 2007). Many prior studies focusing on blogs used factor analysis technique, but did not employ meth-odologies that comprehensively examined the interrelationships between influential criteria. DEMATEL, however, is able to com-pensate for these insufficiencies. It allows the causal relationships between evaluation criteria to be determined and a value structure established. DEMATEL is useful in applications in which appropri-ate decisions must be made (Lin & Tzeng, 2009).

Liou et al. (2007)proposed a hybrid model that combined ANP with DEMATEL to extrapolate the weighted signiﬁcance of organi-zational and managerial factors in the evaluation of airline safety management systems. This study found that although traditional airline safety management systems emphasize incident investiga-tion and analysis, the key determinants that inﬂuence the manage-ment of airline safety are in fact strategic and policy factors. Similar research conducted byWu (2008)also proposed a hybrid model that could be used to analyze how a business could effectively uti-lize and transform knowledge into a competitive advantage, and also help it evaluate and choose better information management strategies.

Recently, DEMATEL has been much praised for its invaluable application in the realm of decision-making. In practice decision makers face both decisions that have independent criteria and those that have interacting criteria. Thus, isolating which criteria are key is beneﬁcial in practice. However as DEMATEL has rarely been used as an analytical tool in the design ﬁeld, this research seeks to reveal the potential applicability of DEMATEL to blog de-sign and development by proposing a hybrid model that combines factor analysis with DEMATEL, in order to identify the key factors and their interrelationships of criteria within each factor.

3. Method

This study proposes a hybrid model that combines factor ysis with the DEMATEL method to conduct a comprehensive anal-ysis in order to gain a complete understanding of blog design. The factor analysis method is commonly used to separate criteria into groups and identify the key factors. Additionally, the DEMATEL

(3)

method is used to map out the interrelations between criteria (Tzeng et al., 2007).

In this section, the steps and procedures in the factor analysis and DEMATEL analysis are brieﬂy outlined as follows.

3.1. Factor analysis: ﬁnding factors and criteria

By identifying just a few key influential factors, factor analysis can be used to illustrate the complex interrelationships that exist among variables while preserving the majority of the data. This re-search adopts the exploratory factor analysis (EFA) method for finding the structural frameworks of factors without imposing any constraints, and also the principal component analysis, which is commonly used for factor analysis. Only those factors having eigenvalues greater than 1 are selected in this study. This is to en-sure that the extracted factors can explain at least a specified amount of variance. In practical terms, to be satisfactory, the total amount of variance explained by factors should be at least 95% in the natural sciences, and 60% in the social sciences (Hair, Anderson, Tatham, & Black, 1998; Kaiser, 1958; Tzeng et al., 2007).

3.2. DEMATEL: clarifying the interrelations between criteria of a factor In the initial phase of this study, factor analysis is carried out to simplify and categorize factors. The second phase seeks to apply DEMATEL to gain insight into the causal relationships between the identiﬁed factors and their degrees of inﬂuence. The three basic assumptions of DEMATEL are: (1) clarity in setting research ques-tions: at the research planning stage, researchers must ensure that the research questions they set are clear; (2) clear association in the relatedness between factors: the weighted association between factors of the research question must be indicated by allocating them rankings in magnitude; (3) understanding of the characteris-tics of each factor arising out of the research question followed by supplementary conclusions after analysis (Wu, 2008). Below is a detailed description of the DEMATEL’s structural frameworks and calculation procedures.

Step 1: find the average matrix. Assume the questionnaire has n criteria and Q test subjects. A pairwise comparison between cri-teria must be made to establish the measurement scale. The Likert-scale is often utilized which ranges from 0 to 4, repre-senting ‘‘no influence’’, ‘‘little influence’’, ‘‘medium influence’’, ‘‘strong influence’’, and ‘‘very strong influence’’, respectively. Thus, the influence matrix of the qth respondent between total criteria n is given as:

Zq¼ zqij

h i

nn ð1Þ

Furthermore, the total average inﬂuenced value from all respon-dents when considering the score from criteria i(ai) to j(aj) is given

as: zij¼ PQ q¼1z q ij Q ð2Þ

Hence, the resulting total average matrix is given as:

ð3Þ

Step 2: calculate the normalized initial direct-relation matrix. The normalized initial direct-relation matrix is obtained by normal-izing the average matrix Z in the following way:

S ¼ Max max 16i6n Xn j¼1 zij;max 16j6n Xn i¼1 zij " # ð4Þ Znor¼ z11=S z1j=S z1n=S .. . .. . .. . zi1=S . . . zij=S zin=S .. . .. . .. . zn1=S znj=S znn=S 2 6 6 6 6 6 6 6 6 4 3 7 7 7 7 7 7 7 7 5 ð5Þ

where Eq.(4)represents the maximum values out of the sums of all the rows, and the sums of all the columns. Eq.(5)represents the normalized initial direct-relation matrix.

Step 3: compute the total relation matrix. After step 2, the total relation matrix can be obtained by using simple numerical cal-culation. The total relation matrix, T, is given by:

T ¼ Znorþ ðZnorÞ2þ þ ðZnorÞp¼ Znor ðI ZnorÞ1

¼ ½tij_nn; p ! 1 ð6Þ

where p represents the power. Hence when p tends to inﬁnity, the matrix T will converge. Furthermore, I is the identity matrix. The to-tals for each row and each column in Eq. (5)can be obtained as follows: ri¼ Xn j¼1 tij; i ¼ 1; 2; . . . ; n ð7Þ cj¼ Xn i¼1 tij; j ¼ 1; 2; . . . ; n ð8Þ

where rirepresents the direct inﬂuence value which is given by the

factor ai; cjrepresents the indirect inﬂuence value which is given by

the factor aj. Similarly, if the lth factor is used, the direct–indirect

value is displayed as:

ril¼ tl1þ tl2þ þ tlðn1Þþ tln ð9Þ

cjl¼ t1lþ t2lþ þ tðn1Þlþ tnl ð10Þ

Hence, through the evaluation of Eqs.(9) and (10), the total relation matrix T can be represented inTable 1. FromTable 1, if the value of ril cjlis positive and has greater value, it means that the criteria al

is of the positively-affected type and has more impact on other criteria.

Step 4: set a threshold value and obtain the impact-relation map. Finally, in order to explain the structural relation between the factors, it is necessary to decide a threshold value to remove the some unsuitable effects from consideration in matrix T. At this stage, experts will discuss how to decide each factor’s threshold to make the rational decisions.

In the following section, more in-depth discussion involving ﬁndings from case analyses of practical examples of blog design are included. Key determinant factors are identiﬁed from the crite-ria, followed by a DEMATEL analysis, tracing the directions of exist-ing causal relationships displayed by these factors to identify the exact elements of blog design that correspond with the main con-cerns of bloggers and blog users.

4. Empirical study and results

In this section, the design evaluation criteria for each factor were identiﬁed. There are three subsections in this section.

(4)

Section4.1describes the questionnaires and the expert interview. Section4.2 describes the results of factor analysis, including the key factors. Section4.3describes the results of the DEMATEL anal-ysis, including the inﬂuence matrix and the impact-relation maps.

4.1. Materials

To develop the design evaluation criteria, eight blog experts were consulted. The experts were all experienced blog users, all of whom read blogs regularly and have managed a blog for at least

three years. After conducting the literature review and the expert interview, 23 criteria were identiﬁed (Table 4).

In the ﬁrst phase of this study, a factor analysis was conducted to obtain independent criteria clusters. The questionnaires in sur-vey were distributed to 202 undergraduate students from National Chiao Tung University who were experienced in blogging. The respondents’ data were analyzed using the windows software package SPSS version 12. Following the factor analysis, the factors were then named by the researcher of this study.

In the second phase, DEMATEL analysis was used to identify the interrelationships between the criteria. The results from the ﬁrst phase of the study were incorporated into a questionnaire, which was then given to 30 experts. These experts included experienced bloggers and website-related scholars. The experts ranked each cri-terion using the Likert 5 point scale.

4.2. Results of factor analysis

Before the questionnaires were given out, a reliability analysis was carried out. The ﬁndings of this analysis indicated a Cronbach’s

a

value 0.939, higher than 0.8, and a standardized element

a

value of 0.943 (Table 2) which demonstrated questionnaire reliability to be signiﬁcant and effective. In order to understand whether respondents responses in the questionnaire were suitable for fac-tor analysis, the KMO value and

a

value were calculated. The values were found to be 0.918 and 2990.373, respectively (Table 3), indi-cating that the data was suitable.

This research applied factor rotation and principal component analysis to extract factors in ways that ensured better interpreta-tion for extrapolainterpreta-tion and hence derive appropriate subjective findings. According toHair et al. (1998), when the sample size is between 200 and 300, the loading should be 0.384. Through con-ducting statistic analysis on the figures extracted, five factors were then identified (Table 4).Table 4shows the criteria under each fac-tor and offers explanations for the magnitudes of accumulated variances as a whole. These five factors were named as follows: vi-sual clarity (F1), interface and usability (F2), content and search-ability (F3), programming (F4), and socisearch-ability (F5).

The following subsection will discuss the observations derived from the impact-relation maps based on the DEMATEL analysis, where the criteria under each factor and their factorial interrela-tionships were examined.

4.3. Results of the DEMATEL analysis

The purpose of this study is to find the relationships between these design criteria. After identifying the five factors, DEMATEL analysis was used to calculate the weighted significance of each criterion under each factor based on the interrelationships shown on the impact-relation maps.

Table 1

The total relation matrix (T).

Table 2

Reliability analysis results.

Source of variance Sum of sq. d.f. Mean square F-test Probability Between people 2878.965 192 14.995 Within people 4726.087 4246 1.113 Between measures 883.653 22 40.166 Residual 3842.434 4224 0.910 27.717 0.000 Total 7605.052 4438 1.714 Grand mean 4.5443 Alpha 0.939 Standardized element Alpha 0.943 Table 3

K.M.O and Bartlett test.

Kaiser–Meyer–Olkin 0.918

Bartlett test Approach Chi. 2990.373

d.f. 253

Signiﬁcance 0.000

Table 4

Factor analysis results: factors and criteria.

Factors Criteria Eigenvalue Percentage of

variance

Cumulative % Visual clarity (F1) Color arrangement (C1), Simplicity of layout (C2), Font arrangement (C3), Text and ground contrast

(C4), Stylistic consistency (C5), Readability (C6), No advertisement banner (C7)

10.479 45.563 45.563 Interface and

usability (F2)

Aesthetic layout (C8), Multiple layout style choice (C9), Ease of management (C10), Ease of registration (C11), Storage capacity (C12), System stability (C13), Friendliness to beginners (C14)

1.834 7.972 53.535 Content and

searchability (F3)

Specialized ﬁeld or professional content (C15), Content richness (C16), Fluency of writing (C17), Customized website address (C18), Ease of searching (C19)

1.451 6.307 59.842

Programming (F4)

Open CSS (C20), Open JavaScipt (C21) 1.125 4.893 64.736

Sociability (F5) Popularity of the blog platform (C22), Friends’ Hyperlinks (C23) 1.063 4.62 69.356 Extraction methods: principal component analysis. Rotation method: Varimax with Kaiser normalization.

(5)

4.3.1. Total relation matrix and causal inﬂuence table

Based on the responses of 30 experts, the total relation matrixes were calculated (Tables 5–9). Take factor one (F1) for example ( Ta-ble 5), color arrangement (C1) will directly impact simplicity of layout (C2) with an impact level of 0.994; color arrangement (C1) will directly impact font arrangement (C3) with an impact le-vel of 0.839. Besides this, color arrangement (C1) will impact itself with an impact level of 0.783.

The ril+ cjland ril cjlvalues can be calculated from the total

relation matrix (Tables 10–14). The ril+ cjl value indicates how

important a criterion is, while the ril cjlvalue indicates the size

of the direct impact of this criterion on other criteria. However, if this value is negative and large, it implies that this criterion is highly inﬂuenced by other criteria (Lee & Kim, 2009).

4.3.2. Impact-relation maps

This research proceeded further by carried thorough discussion with eight experts where appropriate threshold values were de-rived in accordance with the identiﬁed ﬁve factors. The threshold values for visual clarity (F1), interface and usability (F2), content

and searchability (F3), programming (F4), and sociability (F5) were 0.73, 0.55, 0.80, 12.81 and 42.92, respectively. Only values above these thresholds were considered, otherwise the relationships would be too complex. The threshold values have not standard set-tings. The thresholds are decided through expert options in this ﬁeld.

After deciding the threshold values, the DEMATEL impact-relation maps could be obtained (Fig. 1). By drawing the impact-relation maps for the criteria of each factor, the complicated causal relationships could be visualized. These are summarized below:

Visual clarity (F1): color arrangement (C1) (ril cjl= 0.084), font

arrangement (C3) (ril cjl= 0.247), text and ground contrast (C4)

(ril cjl= 0.394), and no advertisement banner (C7) (ril cjl=

0.284) were the positively-affected criteria of factor F1. Simplicity of layout (C2) (ril cjl= 0.016), style consistency (C5) (ril cjl=

0.064), and readability (C6) (ril cjl= 0.93) were the

nega-Table 5

The total relation matrix for the factor of visual clarity (F1).

F1 C1 C2 C3 C4 C5 C6 C7 C1 0.783 0.994 0.839 0.888 0.729 1.004 0.639 C2 0.981 0.88 0.898 0.924 0.766 1.049 0.69 C3 0.871 0.946 0.679 0.825 0.695 0.953 0.574 C4 0.923 0.986 0.847 0.728 0.728 0.996 0.608 C5 0.713 0.765 0.659 0.666 0.48 0.773 0.484 C6 0.844 0.893 0.79 0.802 0.669 0.775 0.588 C7 0.677 0.738 0.586 0.589 0.537 0.739 0.392 Table 6

The total relation matrix for the factor of interface and usability (F2).

F2 C8 C9 C10 C11 C12 C13 C14 C8 0.366 0.557 0.507 0.461 0.422 0.46 0.6 C9 0.562 0.446 0.571 0.517 0.473 0.512 0.687 C10 0.504 0.548 0.441 0.5 0.458 0.501 0.69 C11 0.446 0.487 0.521 0.377 0.417 0.472 0.665 C12 0.45 0.489 0.509 0.457 0.335 0.5 0.592 C13 0.502 0.544 0.561 0.525 0.51 0.409 0.67 C14 0.529 0.6 0.649 0.597 0.511 0.568 0.583 Table 7

The total relation matrix for the factor of content and searchability (F3).

F3 C15 C16 C17 C18 C19 C15 1.055 1.403 1.324 0.827 1.171 C16 1.244 1.11 1.269 0.807 1.148 C17 1.224 1.339 1.017 0.779 1.109 C18 0.646 0.733 0.649 0.391 0.73 C19 0.741 0.793 0.725 0.571 0.595 Table 8

The total relation matrix for the factor of programming (F4).

F4 C20 C21

C20 12.556 12.556

C21 13.556 12.556

Table 9

The total relation matrix for the factor of sociability (F5).

F5 C22 C23

C22 42.667 42.667

C23 43.667 42.667

Table 10

The causal inﬂuence levels for the visual clarity factor (F1).

Criteria ril+ cjl ril cjl

Color arrangement (C1) 11.668 0.084

Simplicity of layout (C2) 12.385 0.016

Font arrangement (C3) 10.84 0.247

Text and ground contrast (C4) 11.237 0.394

Style consistency (C5) 9.142 0.064

Readability (C6) 11.652 0.93

No advertisement banner (C7) 8.233 0.284

Table 11

The causal inﬂuence levels for the interface and usability factor (F2).

Aesthetic layout (C8) 6.734 0.013

Multiple layout style Choice (C9) 7.437 0.097

Ease of management (C10) 7.401 0.115 Ease of registration (C11) 6.819 0.051 Storage capacity (C12) 6.458 0.205 System stability (C13) 7.143 0.3 Friendliness to beginners (C14) 8.526 0.449 Table 12

The causal inﬂuence levels for the content and searchability factor (F3).

Specialized ﬁeld or professional content (C15) 10.688 0.87

Content richness (C16) 10.955 0.2

Fluency of writing (C17) 10.451 0.482

Customized website address (C18) 6.524 0.225

Ease of searching (C19) 8.178 1.328

Table 13

The causal inﬂuence levels for the programming factor (F4).

Open CSS (C20) 51.222 1

Open JavaScipt (C21) 51.222 1

Table 14

The causal inﬂuence levels for the sociability factor (F5).

Popularity of the Blog Platform (C22) 171.667 1

(6)

tively-affected criteria. The key criterion of factor F1, i.e. that crite-rion which inﬂuences the other criteria most and so be taken into consideration, was found to be ‘‘text and ground contrast (C4).’’

Interface and usability (F2): Aesthetic layout (C8) (ril cjl

= 0.013), multiple layout style choice (C9) (ril cjl= 0.097), storage

capacity (C12) (ril cjl= 0.205), and system stability (C13) (ril cjl

= 0.3) were the positively-affected criteria of factor F2. Ease of management (C10) (ril cjl= 0.115), ease of registration (C11)

(ril cjl= 0.051), and friendliness to beginners (C14) (ril cjl

= 0.449) were the negatively-affected criteria. The key criterion of factor F2 was found to be ‘‘system stability (C13).’’

Content and searchability (F3): specialized ﬁeld or professional content (C15) (ril cjl= 0.87), content richness (C16) (ril cjl= 0.2),

and ﬂuency of writing (C17) (ril cjl= 0.482) were the

positively-affected criteria of factor F3. Customized website ad-dress (C18) (ril cjl= 0.225) and ease of searching (C19)

(ril cjl= 1.328) were the negatively-affected criteria. The key

criterion of factor 3 was found to be ‘‘specialized ﬁeld or profes-sional content (C15).’’

Programming (F4): open JavaScipt (C21) (ril cjl= 1) was a

pos-itively-affected criterion, and open CSS (C20) (ril cjl= 1) was a

negatively-affected criterion. The key criterion of factor 4 was found to be ‘‘open JavaScipt (C21).’’

Sociability (F5): friends’ hyperlinks (C23) (ril cjl= 1) was a

pos-itively-affected criterion, and popularity of the blog platform (C22) (ril cjl= 1) was a negatively-affected criterion. The key criterion

of factor 5 was found to be ‘‘friends’ hyperlinks (C23).’’

5. Discussion and conclusion 5.1. Discussion

Based on the results of the empirical study, the researchers make the following suggestions regarding blog design: firstly, there are five important factors that influence blog design: visual clarity (F1), interface and usability (F2), content and searchability (F3), programming (F4), and sociability (F5).

Secondly, there are some important messages and implications to be drawn from the impact-relation maps (Fig. 1). In considering the first factor, if a designer wishes to enhance the visual clarity of a blog, then the criteria of color arrangement (C1), font arrange-ment (C3), and text and ground contrast (C4) should be the most important considerations, because they influence the other criteria most and are influenced by the other criteria least. These three cri-teria demonstrate the importance of readability, where readability is defined as the degree of ease and comfort in interpreting the

(7)

out and reading the content of a blog. Moving onto the second fac-tor, if a blog developer wants to enhance the interface and usability of a blog, then system stability (C13) and multiple layout style choice (C9) are the most influential criteria and should be consid-ered first. In other words, improving these two criteria will result in improving the overall quality of the interface and its usability. For the third factor, if we want to see improvements in the quality of the content and searchability aspects of a blog, then specialized field or professional content (C15) is the most important criterion to work on; the quality of a blog’s content is the main reason read-ers keep returning to read it. When considering the fourth factor, some advanced bloggers wish to customize the appearance of their blog using code, then the criterion open JavaScipt (C21) is impor-tant. Finally, for the fifth factor, friends’ hyperlink (C23) is the key criterion that influencing the popularity of the blog platform (C22).

5.2. Conclusion

Nowadays blogs are a common medium of daily communica-tion. The factors inﬂuencing blog interface design are various and complicated. However, there is still insufﬁcient evaluation of blog interface designs. The evaluation of blog interface design criteria is a MCDM problem, which requires considering many complex and interactive factors.

Based on several aspects of blog design evaluation, this study presents a hybrid MCDM evaluation model for blog design criteria. In accordance with the numerous criteria for blog design evalua-tion, this study combined factor analysis and the DEMATEL method to analyze and illustrate the interactive relationships and impact levels between criteria. By incorporating the opinions of experts, the DEMATEL method was used to systematically evaluate the cri-teria for blog design. When designing a blog, the designer is no longer restricted to his or her won subjective judgment, but can use a statistical method to obtain a more rational reference point. This can greatly contribute to the improvement of blog designs. In addition, the key factors in blog design evaluation were derived using factor analysis. The application of the hybrid MCDM model not only provides designers with a better design strategy but also helps to avoid some unnecessary failures.

Although there are numerous excellent studies devoted to blog design, few of these can explain the interaction between evalua-tion criteria in a systematic way. According to the results of empir-ical studies, a hybrid model should prove useful in evaluating blog design and visualizing the interrelationships between the criteria. Consequently, the author hopes that this study will make a useful contribution to better understanding blog design.

Acknowledgement

The author would like to thank Mr. Yi-Xian Lee’s and Ms. Jo-Tzu Chu’s support in data collection. This study was supported in a re-search grant from the National Science Council, Taiwan.

References

Blood, R. (2000). Weblogs: A history and perspective. <http://

www.rebeccablood.net/essays/weblog_history.html>.

Chen, J. K., & Chen, I. S. (2010). Using a novel conjunctive MCDM approach based on DEMATEL, fuzzy ANP, and TOPSIS as an innovation support system for Taiwanese higher education. Expert Systems with Applications, 37(3), 1981–1990.

Du, H. S., & Wagner, C. (2006). Weblog success: Exploring the role of technology. International Journal of Human–Computer Studies, 64(9), 789–798.

Guan, S. S., & Liu, Y. Y. (2007). Evaluation from the perspective of usability on wretch website’s blog interface design. Journal of Science and Technology, 16(1), 61–74.

Guo, Y., & Salvendy, G. (2009). Factor structure of content preparation for e-business web sites: Results of a survey of 428 industrial employees in P.R. China. Behavior and Information Technology, 28(1), 73–86.

Hair, J. F., Jr., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis (5th ed.). Upper Saddle River, NJ: Prentice Hall.

Horton, R. P., Buck, T., Waterson, P. E., & Clegg, C. W. (2001). Explaining intranet use with the technology acceptance model. Journal of Information Technology, 16, 237–249.

Hsu, C. L., & Lin, J. C. C. (2008). Acceptance of blog usage: The roles of technology acceptance, social inﬂuence and knowledge sharing motivation. Information and Management, 45(1), 65–74.

Kaiser, H. F. (1958). The varimax criterion for analytic rotation in factor analysis. Psychometrika, 23, 187–200.

Lee, S., & Kim, B. G. (2009). Factors affecting the usage of intranet: A conﬁrmatory study. Computers in Human Behavior, 25(1), 191–201.

Lee, Y. C., Li, M. L., Yen, T. M., & Huang, T. H. (2010). Analysis of adopting an integrated decision making trial and evaluation laboratory on a technology acceptance model. Expert Systems with Applications, 37(2), 1745–1754. Lin, C. L., & Tzeng, G. H. (2009). A value-created system of science (technology) park

by using DEMATEL. Expert Systems with Applications, 36(6), 9683–9697. Liou, J. J. H., Tzeng, G. H., & Chang, H. C. (2007). Airline safety measurement using a

hybrid model. Journal of Air Transport Management, 13(4), 243–249. Mullet, K., & Sano, D. (1995). Designing visual interfaces. America: A Prentice-Hall. Nielsen, J., & Tahir, M. (2002). Homepage usability: 50 websites deconstructed.

Indianapolis, IN: New Riders.

Preece, J., Rogers, Y., & Sharp, H. (2006). Interaction design: Beyond human–computer interaction. New York, NY: John Wiley & Sons.

Tractinsky, N., Cokhavi, A., Kirschenbaum, M., & Sharﬁ, T. (2006). Evaluating the consistency of immediate aesthetic perceptions of web pages. International Journal of Human–Computer Studies, 64(11), 1071–1083.

Tseng, M. L. (2009). Using the extension of DEMATEL to integrate hotel service quality perceptions into a cause-effect model in uncertainty. Expert Systems with Applications, 36(5), 9015–9023.

Tzeng, G. H., Chiang, C. H., & Li, C. W. (2007). Evaluating intertwined effects in e-learning programs: A novel hybrid MCDM model based on factor analysis and DEMATEL. Expert Systems with Applications, 32(4), 1028–1044.

Wu, W. W. (2008). Choosing knowledge management strategies by using a combined ANP and DEMATEL approach. Expert Systems with Applications, 35(3), 828–835.