|
Article Excerpt
Abstract
User beliefs and attitudes are key perceptions driving information technology usage. These perceptions, however, may change with time as users gain first-hand experience with IT usage, which, in turn, may change their subsequent IT usage behavior. This paper elaborates how users' beliefs and attitudes change during the course of their IT usage, defines emergent constructs driving such change, and proposes a temporal model of belief and attitude change by drawing on expectation-disconfirmation theory and the extant IT usage literature. Student data from two longitudinal studies in end-user computing (computer-based training system usage) and system development (rapid application development software usage) contexts provided empirical support for the hypothesized model, demonstrated its generalizability across technologies and usage contexts, and allowed us to probe context-specific differences. Content analysis of qualitative data validated some of our quantitative results. We report that emergent factors such as disconfirmation and satisfaction are critical to understanding changes in IT users' beliefs and attitudes and recommend that they be included in future process models of IT usage.
Keywords: Information systems, usage, acceptance, attitude, belief, perceived usefulness, expectation disconfirmation theory
**********
Introduction
Change is an inevitable and inalienable part of human life. We continually adjust, revise, and even reverse our personal beliefs, our opinions of others, our views of social institutions, and our own behaviors as we learn more about our social environments and our own behaviors. Likewise, our beliefs, attitude, intention, and usage of information technology (IT) innovations also change over time as we experience IT usage first-hand and learn from such use. In 1990, Melone (1990) stated, "For the most part, the IS literature is silent on how users form initial attitudes about technologies and how these attitudes are modified over time" (emphasis added). Since then, although a growing body of IT usage research has examined formation of initial beliefs and attitudes, to date, very little research has been directed at explicating why and how beliefs and attitudes change over time. Explaining temporal changes in users' beliefs and attitude toward IT usage is the goal of this study.
We focus on user beliefs (specifically, perceived usefulness) and attitude because prior studies of IT usage, predominantly based on the technology acceptance model (TAM) and similar models, have established these perceptions as the key determinants of both initial IT usage (acceptance) and long-term usage (continuance) intention and behavior (Bhattacherjee 2001; Davis et al. 1989). Any change in beliefs or attitudes will likely have a corresponding impact on, and may even reverse, users' continuance intention and behavior. Such reversal in one's IT usage behavior from initial acceptance to later discontinuance, which Bhattacherjee (2001) termed the "acceptance-discontinuance anomaly," may undermine organizational efforts directed at exploiting the full potential of IT as a means of enhancing employee productivity, efficiency, and effectiveness in the workplace.
Prior research (e.g., Sjazna and Scamell 1993; Venkatesh and Morris 2000) provides preliminary empirical evidence that user beliefs and attitudes do change over time, although to date no study has examined or validated potential reasons for such change. Our study builds on these studies by theorizing and empirically validating the causative drivers and emergent mechanisms driving temporal changes in user beliefs and attitude toward IT usage. In doing so, it extends the traditional static models of IT usage by bringing in a dynamic perspective, explicates emergent constructs that can explain temporal patterns in IT usage, and also resolves some of the prior empirical inconsistencies in the referent literature.
The three research questions of interest to this study are:
(1) Do IT users' beliefs and attitude toward IT usage change over time as they experience IT usage first-hand?
(2) What emergent factors, if any, drive this change and why?
(3) To what extent are these effects generalizable across technology and IT usage contexts?
To address these questions, we draw upon expectation-disconfirmation theory (EDT) (Oliver 1980) and prior TAM research to propose a two-stage model of belief and attitude change. Empirical data collected using student subjects employing longitudinal studies of computer-based tutorial (CBT) and rapid application development (RAD) system usage validated the hypothesized model across two different technologies and usage contexts. Observed differences between the two studies allowed us to probe into and examine the implications of specific contextual differences. Qualitative analysis of respondents' comments from the CBT study helped triangulate and validate some of our quantitative results.
The rest of the paper proceeds as follows. The next section integrates EDT with the IT usage literature to build a theoretical model of belief and attitude change. The third section describes the two empirical studies that test the proposed model. The fourth section describes instrument construction and validation. The fifth section empirically tests the hypothesized model across the CBT and RAD studies, performs post hoc analysis, and examines qualitative data from the CBT study. The final section discusses the theoretical and practical contributions of the study's findings and presents avenues for future research.
Theory and Research Model
Expectation Disconfirmation Theory
Expectation-disconfirmation theory (EDT) (Oliver 1980), an extension of cognitive dissonance theory (CDT) in the social psychology literature, has been used by researchers to understand consumer satisfaction, repurchase intentions, and complaining behaviors in contexts ranging from automobile repurchase (e.g., Oliver 1993), camcorder repurchase (e.g., Spreng et al. 1996), restaurant services (e.g., Swan and Trawick 1981), business professional services (e.g., Patterson et al. 1997), and, most recently, IT usage (e.g., Bhattacherjee 2001). CDT was formulated by Festinger (1957) to explain how discrepancies (dissonance) between one's cognition and reality change the person's subsequent cognition and/or behavior. Cognition, in this context, refers to one's beliefs, affect, opinion, values, and knowledge about one's environment, while behavior refers to actions initiated in response to this cognition and/or personal evaluation of that behavior (Festinger 1957).
In IT usage contexts, CDT suggests that users' pre-usage cognitions (e.g., beliefs, attitude) are generally based on second-hand information, such as vendor claims or industry reports, communicated via interpersonal or mass media channels. Such communicated information may be exaggerated (by vendors) or unrealistic, resulting in cognitions that are less reliable or stable. Over time, as users gain first-hand experience with IT usage, they evaluate the extent to which their initial cognition is consonant or dissonant with actual experience, and revise their cognition and/or behavior to achieve greater consonance. Cognitions are generally more easily changed than behaviors, especially under circumstances where users lack complete volition over their behavior (e.g., at the workplace). Over time, user cognitions reach a steady-state equilibrium, as they become more realistic and entrenched in observed behaviors.
EDT expands on CDT to depict a process model of individual behavior whereby users form an initial pre-usage expectation (belief) about a product, experience its usage over time, and then form post-usage perceptions of the product. The dissonance between users' original expectations and observed performance is captured in the disconfirmation construct. Disconfirmation may be positive or negative depending on whether the observed performance is above or below initial expectations, and is viewed as a deviation from the initial expectation (as the baseline or reference level). Disconfirmation and initial expectation jointly determine user satisfaction or dissatisfaction with the product, which then determines continued product usage or non-usage. Disconfirmation (a belief) and satisfaction (an affect) are, therefore, the two emergent constructs in EDT hypothesized to change subsequent user behavior.
The relationship between expectations and disconfirmation (and hence satisfaction) is somewhat complex (Yi 1990). Most researchers expect these two constructs to be negatively related, since high expectations are more likely to be negatively disconfirmed and low expectations are positively disconfirmed (Yi 1990). Empirically, however, this effect appears to be mixed (e.g., Bearden and Teel 1983; Churchill and Suprenant 1982). Some contend that this ambiguity is an artifact of multiple operationalizations of the disconfirmation construct, measured alternatively as the difference score between expected and realized levels of the overall product or of predefined product attributes, and perceived post hoc difference between expectations and performance (Yi 1990). Spreng et al. (1996) suggested that one's disconfirmation should be evaluated with respect to her desired product attributes, rather than expected product attributes. Patterson et al. (1997) observed that product expectation is an inadequate construct and should be expanded to include fairness of expectations as well.
Although most empirical EDT research directly linked disconfirmation and/or satisfaction to subsequent user intention (e.g., Bhattacherjee 2001; Patterson et al. 1997; Spreng et al. 1996), Oliver (1980), in his original conceptualization of EDT, described a mediated model where the impact of disconfirmation and satisfaction on later intention was mediated by later belief and attitude. In other words, EDT can be viewed as a two-stage model where later-stage expectation (belief) and attitude at time [t.sub.2] is caused by initial-stage expectation and attitude at time [t.sub.1] and also disconfirmation and satisfaction realized at time [t.sub.2]. While disconfirmation and satisfaction capture the cognitive effects of the interim usage experience, initial-stage expectation and attitude may also have a residual effect on the formation of later-stage expectation and attitude by serving as the baseline against which disconfirmation and satisfaction are assessed.
Oliver's (1980) two-stage EDT model was empirically validated in the marketing literature by Bearden and Teel (1983), using a two-period longitudinal study, and by Boulding et al. (1993) and Olson and Dover (1979), using three-period studies. (2) These three-period studies, employing data from one pre-usage and two post-usage time periods, demonstrate that the effect of disconfirmation on later-stage expectation or intention tend to stabilize or "wear off" over time as expectation stabilizes and becomes more consonant with actual experience.
CDT/EDT, as a theoretical referent, is just beginning to gain prominence in the IS usage literature. Szajna and Scamell (1993) designed a laboratory experiment where user expectations were manipulated to be unrealistically high, unrealistically low, or realistic, and found that (1) user expectations change over time as unrealistically high or low expectations tend to wear off over time and regress toward realistic levels, and (2) user satisfaction was significantly different among the realistic, unrealistically high, and unrealistically low groups, although there were no substantive difference in users' decision performance. The authors explained the observed expectation change in terms of cognitive dissonance, but did not operationalize or measure dissonance, and were therefore unable to validate the causative mechanism driving this change. However, they recommended that future studies should investigate the dissonance (disconfirmation) construct in order to better explicate why and how expectations change across time.
In a cross-sectional field survey of online banking, Bhattacherjee (2001) demonstrated that EDT-based constructs such as disconfirmation and satisfaction can successfully explain the continuance intention among online banking users. However, this study did not measure later-stage beliefs and attitudes, which may be immediate antecedents of continuance intention. Although this study suggested that these emergent constructs may explain temporal reversal in IT usage behavior from acceptance to discontinuance, it did not empirically examine such temporal change. In a two-period laboratory experiment, McKinney et al. (2002) used EDT to explain student subjects' satisfaction with online retailers, separately focusing on users' disconfirmation with an online retailing site and with the quality of information presented on that site.
While the above studies established the validity of CDT or EDT in IT usage contexts (McKinney et al. 2002), provided preliminary evidence of temporal changes in user beliefs (Szajna and Scammell 1993), and noted that continuance intention is impacted by emergent constructs such as disconfirmation and satisfaction (Bhattacherjee 2001), no study has yet examined the process by which user beliefs (expectations) or attitude regarding IT usage change over time from the pre-usage stage to usage stage, or the role of emergent constructs in driving that change. Our study addresses this gap in the literature by proposing an EDT-based process model of belief and attitude change.
Prior Research on IT Usage
A vast body of research, based on technology acceptance model (TAM), theory of planned behavior (TPB), and related theories, has examined the effects of user beliefs and attitude on IT usage intention and behavior (e.g., Ajzen and Fishbein 1977; Davis et al. 1989; Taylor and Todd 1995b; Venkatesh and Davis 2000). This research demonstrates that perceived usefulness (the extent to which users believe that system usage will enhance their job performance) is the primary belief driving IT usage intention, whose effect on the dependent variable is partially mediated by attitude (personal affect toward IT usage). In other words, perceived usefulness and attitude are both important predictors of IT usage intention.
Recent longitudinal studies of IT usage suggest interesting temporal patterns in the causal associations predicted from TAM. Following a three-period study of technology use in the workplace, Venkatesh and Morris (2000) found that perceived usefulness has a strong persistent effect on user intention, ease of use has a smaller effect, and both effects are moderated by users' gender. However, since the goal of their study was to explore gender differences in technology usage patterns, the authors did not delve into the temporal drivers of these beliefs. In another study, Szajna (1996) found perceived usefulness to be a strong and consistent predictor of usage intentions across time, but found ease of use to have a declining effect, eventually becoming...
|
