Adding Innovation Diffusion Theory to the Technology Acceptance ...

Lee, Y.-H., Hsieh, Y.-C., & Hsu, C.-N. (2011). Adding Innovation Diffusion Theory to the Technology Acceptance Model: Supporting Employees' Intentions to use E-Learning Systems. Educational Technology & Society, 14 (4), 124–137.

Adding Innovation Diffusion Theory to the Technology Acceptance Model: Supporting Employees’ Intentions to use E-Learning Systems Yi-Hsuan Lee1*, Yi-Chuan Hsieh2 and Chia-Ning Hsu1

1

Department of Business Administration, National Central University, Zhongli, Taiwan // 2Department of Applied Foreign Languages, Ching Yun University, Zhongli, Taiwan // [email protected] // [email protected] // [email protected] *corresponding author ABSTRACT This study intends to investigate factors affecting business employees’ behavioral intentions to use the elearning system. Combining the innovation diffusion theory (IDT) with the technology acceptance model (TAM), the present study proposes an extended technology acceptance model. The proposed model was tested with data collected from 552 business employees using the e-learning system in Taiwan. The results show that five perceptions of innovation characteristics significantly influenced employees’ e-learning system behavioral intention. The effects of the compatibility, complexity, relative advantage, and trialability on the perceived usefulness are significant. In addition, the effective of the complexity, relative advantage, trialability, and complexity on the perceived ease of use have a significant influence. Empirical results also provide strong support for the integrative approach. The findings suggest an extended model of TAM for the acceptance of the e-learning system, which can help organization decision makers in planning, evaluating and executing the use of e-learning systems.

Keywords E-learning system, Technology Acceptance Model (TAM), Innovation Diffusion, Eheory (IDT), Employee training, Structural equation modeling, System adoption, End-users' perception

Introduction To maintain competitiveness and keep a highly-trained and educated workforce, organizations have invested considerable amount of time and resources in e-learning as a supplement to traditional types of training, because it can be simultaneously implemented company –wide, achieve immediacy, consistency and convenience, and is associated with higher profits and lower turnover, thus playing a significant role in training and development (DeRouin, Fritzche & Salas, 2005). Many studies have discussed the benefits of e-learning applications (Ong, Lai, & Wang, 2004; Piccoli, Ahmad, & Ives, 2001). But, despite increased usage, underutilization remains a problem (Moore & Benbasat, 1991; Johansen & Swigart, 1996; Ong et al., 2004). Therefore, if learners fail to use-learning systems, the benefits of such systems will not be achievable (Pituch & Lee, 2006; McFarland & Hamilton, 2006). Researchers and practitioners alike strive to find answers to the problem by investigating individuals’ decisions on whether or not to adopt e-learning systems that appear to promise substantial benefits (McFarland & Hamilton, 2006; Xu & Yuan, 2009; Venkatesh, Morris, Davis, & Davis, 2003). To this end, studies of user perceptions and of understanding factors involved in promoting effective use of these systems (Mun & Hwang, 2003) have become increasingly essential to improve understanding and prediction of acceptance and utilization (Lau & Woods, 2008). Prior empirical studies strived to explicate the determinants and mechanisms of users’ adoption decisions on the basis of the technology acceptance model (TAM) (Davis, Bagozzi, & Warshaw, 1989; Taylor & Todd, 1995; Venkatesh & Davis, 2000) with the conviction that the adoption process influences successful use of particular technology systems (Karahanna, Straub, & Chervany, 1999; Liao, Palvia, & Chen, 2009). This study contributes to the TAM literature by examining the relationships between the innovation diffusion theory and TAM variables in the same model. We propose to examine the effects of motivational determinants on TAM constructs using IDT as a background theory. Thus, we employed five factors: relative advantage, compatibility, complexity, trialability and observability as determinants of perceived usefulness (PU), perceived ease of use (PEU) and behavioral intention to use (BI). This empirical study could be useful for developing and testing theories related to e-learning system acceptance, as well as to practitioners for understanding strategies for designing and promoting e-learning systems.

ISSN 1436-4522 (online) and 1176-3647 (print). © International Forum of Educational Technology & Society (IFETS). The authors and the forum jointly retain the copyright of the articles. Permission to make digital or hard copies of part or all 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 the full citation on the first page. Copyrights for components of this work owned by others than IFETS must be honoured. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from the editors at [email protected].

124

E-learning and TAM The TAM has been widely used as the theoretical basis for many empirical studies of user technology acceptance and has partially contributed to understanding users’ acceptance of information systems (IS)/information technology (IT) (Taylor & Todd, 1995; Venkatesh & Davis, 2000). Our research shows that many studies focus on the acceptance by students in educational institutions (Chang & Tung, 2008; Pituch & Lee, 2006), but acceptance within organizations is rarely covered, and very few studies have adopted the TAM as a model for explaining the use of an e-learning system designed and provided by organizations. TAM could be useful in predicting end-users’ acceptance of an e-learning system in organizations (Davis et al., 1989; Arbaugh, 2002; Wu, Tsai, Chen, & Wu, 2006); additionally, existing antecedents of the technology acceptance intention in the TAM model do not sufficiently reflect the e-learning system end users’ acceptance within organizations (Ong et al., 2004; Lau & Woods, 2008). In our model, employees’ PU of the e-learning systems is defined as the perception of degrees of improvement in learning because of adoption of such a system. PEU of the e-learning systems is the users’ perception of the ease of adopting e-learning systems. We made assumptions that the more end-users who perceive usefulness of the elearning systems within an organization, the more positive their acceptance of e-learning systems, consequently increasing their chances for future usage of the e-learning systems (Arbaugh & Duray, 2002; Pituch & Lee, 2006). Furthermore, technology acceptance is determined by behavioral intention to use (Ajzen & Fishbein, 1980). Therefore, within an organizational context adoption of an e-learning system is a positive function of the intention (BI) to accept the systems.

Theoretical background Although much research supports the TAM as an excellent model to explain the acceptance of IS/IT, it is questionable whether the model can be applied to analyze every instance of IS/IT adoption and implementation. Many empirical studies recommend integrating TAM with other theories (e.g. IDT, or DeLone & McLean’s IS success model) to cope with rapid changes in IS/IT, and improve specificity and explanatory power (Carter & Be´langer, 2005; Legris, Ingham, & Colerette, 2003). TAM and IDT are similar in some constructs and complement each another to examine the adoption of IS/IT. Researchers indicate that the constructs employed in TAM are fundamentally a subset of perceived innovation characteristics; thus, the integration of these two theories could provide an even stronger model than either standing alone (Wu & Wang, 2005; Chen, Gillenson, & Sherrell, 2002). Past studies integrated the two theories, providing good results (Sigala, Airey, Jones, & Lockwood, 2000; Chen et al, 2002). This study employs two major theoretical paradigms—the TAM (Gefen, 2004; Talyor & Todd, 1995; Davis et al., 1989) and IDT (Roger, 1995; Moore & Benbasat, 1991). After reviewing literature on technology acceptance, we synthesized the major theories and empirical research, then proposed a model that blended key constructs involved in e-learning system acceptance and intention to use the e-learning systems. Five constructs of innovative characteristics, PEU, and usefulness and intention to use the e-learning system, were taken from the TAM and IDT. With appropriate modifications, our proposed model could successfully be generalized to acceptance within an organizational context.

The Technology Acceptance Model (TAM) The TAM was derived to apply to any specific domain of human–computer interactions (Davis et al., 1989). The TAM asserts that two salient beliefs —PU and PEU—determine technology acceptance and are the key antecedents of behavioral intentions to use information technology. The first belief, PU was the degree to which an individual believes that a particular system would enhance job performance within an organizational context (Davis et al., 1989). PEU, the second key belief, was the degree to which an individual believes that using a particular system would be free of effort (Davis et al., 1989). In addition, the model indicated that system usage was indirectly affected by both PEU and PU. 125

Many researchers have conducted empirical studies to examine the explanatory power of the TAM, which produced relatively consistent results on the acceptance behavior of IT end users (Igbaria, Zinatelli, Cragg, & Cavaye, 1997; Venkatesh & Davis, 2000; Horton, Buck, Waterson, & Clegg, 2001). Researchers have agreed that TAM is valid in predicting the individual acceptance of numerous systems (Chin & Todd, 1995; Segars & Grover, 1993). In summary, TAM provided an explanation of the determinants of technology acceptance that enables explanation of user behavior across a wide scope of end-user information technologies and user populations (Davis et al, 1989).

Innovation Diffusion Theory (IDT) Research on the diffusion of innovation has been widely applied in disciplines such as education, sociology, communication, agriculture, marketing, and information technology, etc (Rogers, 1995; Karahanna, et al., 1999; Agarwal, Sambamurthy, & Stair, 2000). An innovation is “an idea, practice, or object that is perceived as new by an individual or another unit of adoption” (Rogers, 1995, p. 11). Diffusion, on the other hand, is “the process by which an innovation is communicated through certain channels over time among the members of a social system” (Rogers, 1995, p. 5). Therefore, the IDT theory argues that “potential users make decisions to adopt or reject an innovation based on beliefs that they form about the innovation” (Agarwal, 2000, p. 90). IDT includes five significant innovation characteristics: relative advantage, compatibility, complexity, and trialability and observability. Relative advantage is defined as the degree to which an innovation is considered as being better than the idea it replaced. This construct is found to be one of the best predictors of the adoption of an innovation. Compatibility refers to the degree to which innovation is regarded as being consistent with the potential end-users’ existing values, prior experiences, and needs. Complexity is the end-users’ perceived level of difficulty in understanding innovations and their ease of use. Trialability refers to the degree to which innovations can be tested on a limited basis. Observability is the degree to which the results of innovations can be visible by other people. These characteristics are used to explain end-user adoption of innovations and the decision-making process. Theoretically, the diffusion of an innovation perspective does not have any explicit relation with the TAM, but both share some key constructs. It was found that the relative advantage construct in IDT is similar to the notion of the PU in TAM, and the complexity construct in IDT captures the PEU in the technology acceptance model, although the sign is the opposite (Moore & Benbasat, 1991). Additionally, in terms of the complexity construct, TAM and IDT propose that the formation of users’ intention is partially determined by how difficult the innovation is to understand or use (Davis, et al., 1989; Rogers, 1995). In other words, the less complex something is to use, the more likely an individual is to accept it. Compatibility is associated with the fit of a technology with prior experiences, while the ability to try and observe are associated with the availability of opportunities for relevant experiences. These constructs relate to prior technology experience or opportunities for experiencing the technology under consideration. Compatibility, and the ability to try and observe can be treated as external variables, which directly affect the constructs in the technology acceptance model. After the initial adoption, the effects of these three constructs could be diminished with continuous experience and reduced over time (Karahanna et al., 1999). Thus far, numerous studies successfully integrated IDT into TAM to investigate users’ technology acceptance behavior (Hardgrave, Davis, & Riemenschneider, 2003; Wu & Wang, 2005; Chang & Tung, 2008). Few have attempted to examine all IDT characteristics with the integration of TAM. In this research, we improve TAM by combining IDT characteristics, adding compatibility, complexity, relative advantage, and the ability to try and observe as additional research constructs to increase the credibility and effectiveness of the study.

Research model and hypotheses We propose an integrated theoretical framework, which blends TAM and IDT theories. The research model holds that the five innovative characteristics (compatibility, complexity, relative advantage, ability to try and observe) exert an important effect on the employees’ PU, PEU and intention to use e-learning systems. We thus tested the validity and applicability of the proposed model based on the following hypotheses.

126

Compatibility Agarwal and Prasad (1999) asserted a positive relationship between an individual’s prior compatible experiences and the new information technology acceptance. They found that the extent of prior experience with similar technologies was positively associated with an ease of use belief about an information technology innovation. Moreover, Chau and Hu (2001) reported that the effect of compatibility was found to be significant only in relation to PU. Later, Wu and Wang (2005) and Chang and Tung (2008a) confirmed that compatibility had a significant positive and direct effect on PU and the behavioral intention. Likewise, prior studies have investigated compatibility from different aspects, resulting in support for its impact on PU, PEU and intention to use (Hardgrave et al., 2003). Based upon the preceding research, the following hypotheses were proposed: H1-1: Compatibility had a positive effect on PU of the e-learning system. H1-2: Compatibility had a positive effect on PEU of the e-learning system. H1-3: Compatibility had a positive effect on behavioral intention to use the e-learning system.

Complexity Empirical studies provided evidence indicating that complexity had a significantly negative effect on the intention to use (Shih, 2007; Lee, 2007). Additionally, a negative relationship between complexity and PU was also revealed in a study conducted by Hardgrave, et al. (2003). Similarly, empirical research has also shown that the more complex the end users perceived the e-learning system as being, the lower the users’ intention to use the system (Lin, 2006). Thus, based on the aforementioned studies, we proposed the following hypotheses: H2-1: Complexity negatively affected PU of the e-learning system. H2-2: Complexity negatively affected PEU of the e-learning system. H2-3: Complexity negatively affected behavioral intention to use the e-learning system.

Relative advantages Research consistently found that the perceived relative advantages positively affected the users’ intention to use the system across different participants (Shih, 2007; Lee, 2007). However, in TAM and IDT research, the relationships among relative advantages, PU, and PEU had seldom been studied with the only one study revealed that when the users perceived higher relative advantages, they perceived a higher level of usefulness of the systems. Accordingly, we hypothesized: H3-1: The relative advantages had a positive effect on PU of the e-learning system. H3-2: The relative advantages had a positive effect on PEU of the e-learning system. H3-3: The relative advantages had a positive effect on behavioral intentions to use the e-learning system.

Observability Using different methodologies and involving different participants from many fields, some studies found that observability had a positive impact on the users’ attitude toward the system and intention to use the system (Lee, 2007). Also in line with previous studies combining TAM and IDT, when the employees perceived the systems as being easier to be observed or described, they tended to perceive the systems more useful and easier to use (Huang 2004; Yang, 2007). Therefore, we proposed that observability would have a positive effect on PU, PEU, and behavioral intention to use the e-learning system. The following hypotheses tested these assumptions: H4-1: Observability had a positive effect on PU of the e-learning system. H4-2: Observability had a positive effect on PEU of the e-learning system. H4-3: Observability had a positive effect on behavioral intention to use the e-learning system.

127

Trialability Some studies have empirically tested in understanding the association between trialability and the intention to use the system (Lee, 2007). They found that trialability had a positive effect on the intention to use the system. However, limited research has been conducted to investigate the relationship among trialability, PU, PEU, and behavioral intentions to use the systems. There was only one research reported that when the users perceived higher trialability, they perceived higher levels of usefulness, and ease of use of the system (Yang, 2007). Accordingly, we tested the following hypotheses: H5-1: Trialability had a positive effect on PU of the e-learning system. H5-2: Trialability had a positive effect on PEU of the e-learning system. H5-3: Trialability had a positive effect on behavioral intention to use the e-learning system.

PEU PEU is the degree to which an individual believes that using a particular system would be free of effort (Davis et al., 1989). Information system researchers have indicated that PEU has a positive effect on the end-users’ behavioral intention and PU to use the systems (Chin & Todd, 1995). Thus, we hypothesized: H6-1: PEU had a positive effect on the PU of the e-learning system.

PU PU is the degree to which an individual believes that a particular system would enhance his or her job performance within an organizational context (Davis et al., 1989). Information system researchers have investigated TAM, and asserted that PU was valid in predicting the individual’s acceptance of various systems (Venkatesh & Davis, 2000). Previous studies discovered that PU positively affected the users’ behavioral intention to use systems (Chin & Todd, 1995). Therefore, we hypothesized: H6-2: PU will have a positive effect on the behavioral intention to use the e-learning system. Demographics Gender Female Male Age ＜29 30-39 40-49 ＞50 Education High school College/University degree Master degree Doctoral degree Experience with computers ＜1 year 1 to 3 years 3 to 6 years 6 to 9 years ＞9 years

Table 1: Demographics of the respondents Number 260 292

% 47.1 52.9

320 155 49 28

58.0 28.1 8.9 5.1

13 308 224 7

2.4 55.8 40.6 1.3

120 173 83 54 122

21.7 31.3 15.0 9.8 22.1

128

Research methodology The subjects and the procedure This study utilized a web-based and mailed survey to collect data for quantitative testing of the research model. Because of the lack of a reliable sampling frame, it proved difficult to conduct a random sampling for all the endusers in the organizations using e-learning systems in Taiwan. Thus, in this study we adopted a non-random sampling technique (i.e. convenience sampling) to collect the sample data. To generalize results, we gathered sample data from the five largest e-learning systems using industries (Chan, 2005), including manufacturing, finance, marketing and service, information technology, and government agencies in Taiwan, and randomly selected 15 firms that provide an e-learning training system for employees (three in each industry). Of the 736 mailed and electronic questionnaires, 566 were completed and returned. Sample demographic information is depicted in Table 1.

Measures To ensure content validity of the scales, the items chosen for the constructs were adapted from previous research to ensure content validity. The questionnaire consisted of three parts. The first part was based on nominal scales and the rest are 5-point Likert scales. Part 1 of the questionnaire was based on IDT including compatibility (CPA), complexity (CPL), relative advantages (ADV), observability (OB), and trialability (TRI). The above items were adapted from the previous studies (Davis et al., 1989; Moore & Benbasat, 1991; Taylor & Todd, 1995; Karahanna et al., 1999), containing 18 items. Part 2 of the questionnaire was based on the constructs of PU, PEU, BI in the TAM model and was adapted from the measurement defined by Davis et al. (1989) and Venkatesh & Davis (2000), containing 12 items for the above constructs. Part 3 of the questionnaire was to collect the interviewees’ basic demographic data, such as gender, educational level, work experience, prior experience using computers, etc.

H1-1

CPA

H1-2 H1-3

CPL

H2-1 H2-2

ADV

PU H6-2

H2-3

H3-1 H3-3

BI

H3-2 H4-1

OB

H4-

H4-2 H5-1 H5-3

TRI

H6-1

PEU

H5-2 Figure 1. Proposed research model.

129

Results Instrument validation Two confirmatory factor analyses (CFA) were computed using AMOS 6.0 to test the measurement models. The model-fit measures were used to assess the model’s overall goodness of fit (χ2 /df, GFI, NFI, CFI, RMSEA) and values all exceeded their respective common acceptance levels (Hair, Black, Babin, Anderson, & Tatham, 2006). This showed that the measurement model exhibited a fairly good fit with the collected data (Table 2). Goodness-of-fit measure χ2/df GFI AGFI NFI CFI RMSEA Constructs/Factors CPA

CPL ADV

OB TRI PU PEU BI

Table 2: Fit indices for endogenous and exogenous measurement models Recommended value Endogenous measurement Exogenous measurement model model 1.764 1.977 ≦3.00 0.979 0.958 ≧0.90 0.960 0.936 ≧0.90 0.967 0.967 ≧0.90 0.983 0.983 ≧0.90 0.037 0.042 ≦0.05 Indicators CPA1 CPA2 CPA3 CPA4 CPL1 CPL2 CPL3 ADV1 ADV2 ADV3 ADV4 ADV5 OB1 OB2 OB3 TRI1 TRI2 TRI3 PU1 PU2 PU3 PEU2 PEU3 PEU4 BI1 BI2 BI3 BI4 BI5

Table 3: Convergent validity Standardized Reliability loadings (>0.707) (R2) (>0.50) .807 .652 .720 .518 .791 .626 .739 .547 .854 .730 .918 .842 .848 .719 .777 .604 .812 .660 .876 .768 .905 .819 .854 .729 .744 .554 .953 .908 .740 .547 .790 .624 .838 .703 .720 .518 .847 .717 .870 .757 .717 .514 .769 .591 .766 .587 .703 .494 .675 .456 .773 .598 .935 .874 .879 .773 .854 .729

Composite reliability (>0.70) .849

Average variance extracted (>0.50) .585

.906

.764

.926

.716

.857

.670

.827

.615

.854

.663

.841

.570

.915

.686

Convergent validity of scale items was estimated by reliability, composite reliability, and average variance extracted (Fornell & Larcker, 1981). The standardized CFA loadings for all scale items exceeded the minimum loading criterion of 0.70, and the composite reliabilities of all factors also exceeded the recommended 0.70 level. In addition, 130

the average variance-extracted values were all above the threshold value of 0.50 (Hair, et al., 2006). Hence all three conditions for convergent validity were met for the four measurement models (See Table 3). Discriminant validity was obtained by comparing the shared variance between factors with the average variance extracted from the individual factors (Fornell & Larcker, 1981). This analysis showed that the shared variances between factors were less than the average variance extracted for the individual factors. Hence, discriminant validity was assured (see Table 4). To sum up, the four measurement models reached satisfactory levels of reliability, convergent validity and discriminant validity. Construct

Table 4: Discriminant validity Interconstruct correlations PU PEU CPA CPL ADV

BI OB TRI BI 0.828 PU 0.353 0.814 PEU 0.286 0.229 0.755 CPA 0.466 0.401 0.253 0.765 CPL 0.180 0.068 0.572 0.210 0.874 ADV 0.368 0.375 0.269 0.624 0.138 0.846 OB 0.138 0.052 0.094 0.123 0.061 0.138 0.819 TRI 0.228 0.095 0.271 0.240 0.240 0.185 0.203 0.784 Note. Diagonals represent the square root of average variance extracted, and the other matrix entries are the factor correlation.

Structural model estimation and hypotheses testing Descriptive statistics The means and standard deviations for all constructs were determined and were displayed in Table 5. The highest mean of 3.56 was for the trialability, while the lowest mean for complexity was 2.30 on a scale of 1 to 5. The means for PU, PEU and behavioral intention were 3.79, 3.73, and 3.62, respectively. Construct (# Items) BI (six items) PU (five items) PEU (four items) CPA (four items) CPL (three items) RA (five items) OB (three items) TRI (three items)

Table 5: Descriptive statistics Mean 3.62 3.79 3.73 3.54 2.30 3.46 3.39 3.56

Standard deviation .774 .708 .709 .808 .769 .794 .930 .794

Structural equation modeling (SEM) SEM was performed to test the fit between the research model (Figure 1) and the obtained data. This technique was chosen for its ability to simultaneously examine a series of dependence relationships, especially when there were direct and indirect effects among the constructs within the model (Hair, et al., 2006). The first step in interpreting SEM results includes reviewing fit indices, which provide evidence on how well the fit is between the data and the proposed structural model. If the model fits the data well enough, a second step involves reviewing the feasibility of each path in the model by examining whether the weights are statistically significant and practically significant. Practical significance is evaluated on the basis of whether the effect size estimation (the R2) regarding a given path in the models is large enough.

131

In this study, Amos 6.0 was employed and the SEM estimation procedure was a maximum likelihood estimation. A similar set of fit indices was used to examine the structural model. Comparison of all fit indices with their corresponding recommended values provided evidence of a good model fit (χ2/df = 1.42, GFI = 0.95, AGFI = 0.93, CFI = 0.99, RMR = 0.02, and RMSEA = 0.03). The next step in the data analysis was to examine the significance and strength of hypothesized relationships in the research model. The results of the analysis of the structural model, including path coefficients, path significances, and variance explained (R2 values) for each dependent variable presented in Figure 2. Figure 2 showed the resulting path coefficients of the proposed research model. Overall, fourteen out of seventeen hypotheses were supported by the data. Three endogenous variables were tested in the model. The results showed that PU significantly influenced BI (β= 0.267, P