Toward a bi-dimensional measurement of computer experience.

1. INTRODUCTION

There is a great deal of interest in an individual's computer experience (CE) because there is a strong linkage between past experience with computers and ensuing attitude and consequent behavior in the computing environment (Beckers and Schmidt, 2001; Polzikova, 1995; Maurer, 1994; Chou, 2001; Compeau and Higgins, 1995). If this phenomenon were better understood, educational institutions and employers would be able to more efficiently deploy their computing resources (Taylor and Todd, 1995; Smith et al, 1999). Also, many would benefit from being able to measure an individual's CE because some aspect of CE (typically its time-based, objective component) has been shown to have a significant, positive correlation with constructs such as computer self-efficacy (Harrison and Rainer, 1992; Hill et al., 1987; Igbaria and Iivari, 1995; Potosky, 2002; Hasan, 2003), and a negative correlation with constructs such as computer anxiety (Durndell, 2002; Cohen, 1989; Maurer, 1994).

Because of its significance, CE has been included in several studies involving the constructs of computer self-efficacy (Harrison and Rainer, 1992; Hill et al., 1987; Igbaria and Iivari, 1995; Potosky, 2002; Hasan, 2003), computer attitudes (Smith, Caputi, and Rawstorne, 2000), and computer anxiety (Durndell, 2002; Cohen, 1989; Maurer, 1994). However, this research was performed without a consistent definition of CE, a theoretical distinction between CE and other similar attitudinal constructs, or a validated measurement instrument. This has led to conflicting results and conclusions (Potosky and Bobko, 1998; Kay, 1992; Smith, Caputi, Crittendon, Jayasuriya, and Rawstorne, 1999).

1.1 Problem

Most previous studies have used a time-based measure of CE, while not including the potential impact of thoughts and feelings acquired from the interactions with a computer (e.g., Gardner et al., 1993; Wise, Barnes, Harvey, and Plake, 1993; Comber et al, 1997; Arthur and Olson, 1991; Mahar et al, 1997). This approach to studying CE leaves out a very important element, the quality of those experiences (Arthur and Olson, 1991; Bradley and Russell, 1997; Karsten and Roth, 1998). Researchers have proposed that the conflicting CE research is the product of CE being measured solely by its quantitative (time based) component, without considering the qualitative component of the individual's thoughts and feelings (Rawstorne, Caputi, and Smith, 1998; Smith et al., 1999; Smith, Caputi, and Rawstorne, 2000; Smith, Caputi and Rawstorne, 2004).

Smith et al. (1999) attempted to ground CE research by theoretically defining an individual's CE as the combination of two separate and distinct components, namely, a quantitative Objective Computer Experience (OCE) and qualitative Subjective Computer Experience (SCE). Smith et al. (1999, p. 228) defined OCE as the "totality of externally observable, direct and/or indirect human-computer interaction which transpires across time." Smith et al. (2004, p. 3) defined SCE as "a private psychological state, reflecting the thoughts and feelings a person ascribes to some previous or existing computing event." Measurement of both OCE and SCE was discussed (Smith et al., 1999) but an instrument was presented only for SCE (Smith et al., 2000; Smith et al., 2004). Also, due to the instrument's construction (discussed later), there seems to be a theoretical disconnect between the definition of SCE and the proposed measure of SCE (Smith et al., 2004).

To further complicate the situation, only one attempt have been made to theoretically and statistically differentiate this construct (and its measurements) from similar constructs whose factors could easily overlap one another (Smith et al., 2000). By definition, constructs must be differentiated from similar psychometric constructs to increase the possibility of unidimensionality (Miller and Rainer, 1995; Straub, 1989) and decrease the chance of conceptual overlap (Hattie, 1985; Straub, 1989). Smith et al. (2000) attempted to differentiate their proposed measure of CE (specifically SCE) from computer attitudes, but were not completely successful. The sub-factors of their measure of SCE did not exhibit discriminant validity from the Computer Attitude Scale developed by Lloyd and Gressard (1984) or the Computer Attitude Scale developed by Dambrot, Watkins-Malek, Silling, Marshall and Garver (1985). Theoretically and statistically, this lack of discriminant validity is a problem for researchers wishing to use CE as a construct in statistical research models.