Abstract
This paper argues for enhanced consideration of third variables in interactivity research and proposes a “mediated moderation” model to bring increased sophistication to bear on the study of information technology effects. Interactivity, a central phenomenon in new media research, is an elusive concept that has enduringly intrigued and confused scholars. Extant conceptualizations have produced incomplete causal models and have generally ignored the effect of third variables. We conceptualize interactivity as technological attributes of mediated environments that enable reciprocal communication or information exchange, which afford interaction between communication technology and users, or between users through technology. Specifying roles for mediator and moderator variables, this paper proposes a model that incorporates interactive attributes, user perceptions (mediators such as perceived interactivity), individual differences (moderators such as Internet self-efficacy), and media effects measures to systematically examine the definition, process, and consequences of interactivity on users. Lastly, statistical procedures for testing mediated moderation are described.
Notes
∗: Statistical significance
As CitationBucy (2004a) observes, the literature is rife with exhaustive reviews and definitional reformulations of the interactivity concept (see CitationJensen, 1998; CitationKiousis, 2002; CitationMcMillan, 2002a); hence, we take a more focused approach here.
In addition, scholars have employed manipulation checks of interactivity to verify that technological attributes are experienced as intended. For instance, scale items have been used to ask participants how interactive they considered the visited website to be, confirming whether the level of interactivity was reliably manipulated (e.g., CitationBucy, 2004b; CitationMacias, 2003; CitationSundar et al., 2003).
Perceived interactivity was operationalized as the degree to which participants indicated that the communication event shown in the clip was spontaneous, two-way in nature, or occurring in real-time; that is, they or someone like them could have expected an individualized response had they been present (CitationBucy & Newhagen, 1999).
On the other hand, CitationSpencer, Zann, and Fong (2005) argue that, under certain conditions, experiments are more effective than mediational analyses in examining psychological processes. In particular, when it is easy to manipulate and measure a proposed psychological process, a series of “causal chain” experiments is preferable. On the other hand, when measurement of a psychological process is easy but manipulation of it is difficult, designs that rely on mediational analyses are preferred (CitationSpencer, Zanna, & Fong, 2005).
Parasocial interaction elicited by a noninteractive medium is a typical example. When watching a political talk show, for instance, audience members may “call in” to express their opinions using the telephone but the perception of the experience depends on television, which does not afford a talk-back function. Although nonverbal communication may evoke the illusion of intimacy (or perceived interactivity) and promote further contact through other media, it is not interactivity per se.
CitationPotter and Tomasello (2003) suggest that there are four types of mediators: demographics, personality traits, viewer or user states, and audience interpretations of treatment material.
Centering the variable is necessary because moderated multiple regression (MMR) utilizes the interaction term, which is computed from existing variables in the regression equation. However, this creates problems with multicollinearity, which inflates standard errors of regression coefficients of first order terms (the independent and moderator variables), which in turn can widen confidence intervals and produce low statistical power (CitationCohen, 1978). In addition, correlated variables make regression coefficients unstable and change with different samples. This leads to the difficulty of separating and explaining the impact of each variable. Although variable centering minimizes multicollinearity, its effectiveness is still disputable because centering does not affect any values of interest; that is, the regression coefficient, standard error, simple slope, t -test, and p -value of second order terms (the interaction term) are identical regardless of whether the continuous variable is centered (see CitationAiken & West, 1991; CitationKromrey & Foster-Johnson, 1998). Variable centering does not improve the power of MMR for detecting moderation effects but should be conducted with tests of mediated moderation so as to distinguish the impact of each variable and obtain meaningful regression results.
When the regression equation involves interaction terms, we suggest reporting unstandardized regression coefficients (b) rather than standardized regression coefficients (β) in the results. The computation of standardized regression coefficients requires that the independent and moderator variables be standardized before being entered into the equation, producing an interaction term that is the product of the standardized variables but not the standardized product of the raw variables (CitationFriedrich, 1982).