Abstract
An implicit assumption underlying previous interactivity studies is that every time people use a communication medium (e.g., website) or device (e.g., smartphone), they perceive its interactivity through analyzing it from scratch trait-by-trait. As psychologists have long shown, however, we quite often skip such an intensive analysis, and rely on our expectations or schematic knowledge to perceive/evaluate an object. This study is designed to develop a measure of individuals’ expectation of interactivity toward a medium, called expected interactivity (EI). After specifying three conceptual dimensions underlying EI – sensory, semantic, and behavioral dimensions – scales for capturing them are developed, refined, and validated through multiple studies. Implications for future interactivity research are discussed.
Keywords
For the past few decades, interactivity has been arguably the single most popular term used to characterize emerging telecommunication technologies, such as the Internet, across various disciplines, including communication, business, and information science. In spite of the widespread usage of the term, however, interactivity still remains a nebulous concept as scholars and practitioners have reached little consensus with regard to its role in communication (Rafaeli and Ariel, 2007; Walther et al., 2005). While some earlier studies found positive effects of interactivity on users’ various psychological responses, including attitude, involvement, and arousal (e.g., Cho and Leckenby, 1999; Coyle and Thorson, 2001; McMillan et al., 2003; Wu, 1999), others found negative (e.g., Bezjian-Avery et al., 1998) or non-linear effects with decreasing marginal returns, namely plateau effects (e.g., Fortin and Dholakia, 2005).
In order to explain what causes such variations, researchers have attempted to identify factors that moderate the effects of interactivity on communication outcomes, which include individuals’ cognitive capacity (Vorderer et al., 2001), need for cognition (NFC) (Jee and Lee, 2002; Macias, 2003; Sicilia et al., 2005), need for emotion (Macias, 2003), and product/task involvement and experience (Liu and Shrum, 2009), among many others. Furthermore, Sohn et al. (2007) found that the effects of interactivity tended to vary depending on the extent of interactivity that people anticipated they would experience in an interaction situation. In their experiment, participants who expected a higher degree of interactivity toward a product-related website responded positively to the increase of the website’s interactivity, while those with a lower expectation responded negatively.
They called this particular type of expectation expected interactivity (EI hereafter) and defined it as “the extent of interactivity that a person expects to experience during a prospective interaction with a message vehicle, such as a website” (Sohn et al., 2007: 110). Note here that EI is a category-level expectation, which relies not on the specific qualities of the object (e.g., website) to interact, but on the actor’s knowledge regarding a class of objects to which the focal object belongs (i.e., how interactive automobile brand websites usually are). This finding reminds us of an important, but often overlooked, notion that when interacting with an object, people consider not only the object’s specific qualities, but also the holistic information of the categories it belongs to, which is called schema.
As psychologists have long demonstrated since Gestalt psychology in the early 20th century, people do not necessarily perceive and evaluate things, one by one, in an algebraic manner, but rather bring their own schematic knowledge to everyday perceptions to simplify an otherwise very complex world (Fiske and Taylor, 1991). When meeting a stranger, for example, we tend to compare the target’s attributes to what we already know about the categories it belongs to (e.g., occupation, gender, ethnic background, etc.) and make a judgment of the person (e.g., “I don’t trust him, because I think most politicians are liars”). This cognitive process in which schematic knowledge translates into evaluation or attitude, called “schema-triggered affect” (Fiske, 1982), may also occur when individuals are facing a communication medium such as a website. For instance, a person may hold EI toward a website or a mobile application based on his/her prior interaction experiences with similar items in the same product category (e.g., computers) or content genres (e.g., news).
The role of schematic knowledge, such as EI, in the context of interactive communication has not received much scholarly attention due partly to the lack of adequate measurement instruments. Instead of measuring directly individuals’ EI, thus, Sohn et al. (2007) presented two experimental stimuli, automobile and furniture websites, toward which subjects were expected to have a high or low EI respectively. Then, it was examined how the effects of interactivity on subjects varied between the two different levels of EI. Such an experimental approach was proven quite useful for testing the internal validity of a hypothesized causal relationship, but may be limited in testing its external validity in a broader context encompassing various media and issue categories.
The purpose of this study is, therefore, to develop a scale for measuring individuals’ EI toward any category of media or content using a rigorous scale development procedure (Churchill, 1979; DeVellis, 1991; Furr and Bacharach, 2008). Measuring people’s EI toward a variety of interaction situations or issue categories would enable a more comprehensive and systematic understanding of the interactive communication process in which prior experiences and expectations are intertwined with newly formed experiences. In the following sections, EI as a psychological construct is defined and its conceptual structure is discussed based on previous literature. Then, a measurement instrument reflecting the conceptual structure is developed, and a series of psychometric tests are conducted to affirm the reliability and validity of the measurement instrument.
Conceptualization of expected interactivity
Types of expectation
Expectation is an essential element of life in an environment where uncertainty looms; the behaviors of all living organisms are largely based on a projection of the future outcomes to be brought by them – if a behavior is expected to lead to a positive outcome (e.g., turning left would lead to prey), it is likely to be enacted, while otherwise it is not. Since Tolman (1932) defined it broadly as a subjective belief in the probability that a certain kind of behavior leads to a particular outcome, social scientists have employed expectation as a key concept in diverse areas that include organizational behavior (Vroom, 1964), the formation and change of attitude (Fishbein and Ajzen, 1975), and judgment and decision-making (Kahneman et al., 1982), among many others.
The ordinary meaning of expectation is straightforward and unequivocal, but in academia, it has been distinguished into two kinds, respectively named as contingent and intentional expectation (Van Raaij, 1998). This division is made based on whether expectation is formed through the mode of problem solving or repetition. Contingent expectation refers to “future contingencies over which a person has no control” (Katona, 1975: 402). Like predicting weather, gambling outcomes, or the occurrence of a natural disaster (e.g., earthquake), people can have expectations about something that they cannot control or affect. Such probabilistic projections may not always be met by reality, and thus are revised immediately once they are proven wrong.
Expectation in this sense, a subjective estimation of an event’s likelihood, combined with the value of the event, forms expected value, and rational decision-makers are assumed to maximize it (Schoemaker, 1982). Applying this approach to the consumer behavior context, some have theorized consumer satisfaction as a process of expectation disconfirmation; positive disconfirmation (i.e., better than expected) is assumed to lead to satisfaction, while negative disconfirmation (i.e., no better than expected) to dissatisfaction and the subsequent revision of the former expectation (Cardozo, 1965; Oliver, 1980; Oliver and Winer, 1987).
In contrast, intentional expectation refers to one’s belief formed through repeated learning, a series of trial–error experiences like operant conditioning (Olshavsky and Miller, 1972; Tolman, 1932). This learned belief, which is “at least partially under one’s own control” (Van Raaij, 1998: 402), tends to remain as is for a long time and not be modified easily by a few disconfirming events, which make it more resistant to situational turbulences. If a person has had many unpleasant experiences with car dealers in the past, for example, meeting one or two good people with the same occupation will not substantially change his/her expectation toward the entire occupation category (the opposite can also be the case, of course). Since people generally tend not to change this type of expectation for a single incidence of disconfirmation, the consistency principle (Festinger, 1957) seems more appropriate for explaining how it works than the aforementioned value-maximization principle. That is, individuals have a tendency of remaining consistent with their existing beliefs or expectations, which determines the way they respond to expectation-consistent or inconsistent information.
Sohn et al. (2007) have found that EI is a type of intentional rather than contingent expectation, which means that it is a belief learned through one’s repeated interaction experiences with similar kinds of communication media. For instance, a person’s EI toward a commercial website may be dependent on his/her prior interaction experiences with other similar websites within the same product category (e.g., computers). The notion implies that a person’s EI toward a category of communication media may not change drastically depending on situational contingencies, which makes it a relatively stable characteristic that is meaningful to measure. This by no means suggests that EI is a constant or dispositional characteristic of individuals. Instead, the EI people hold toward a category of media may change, but this happens relatively slowly as the experiences violating their EIs accumulate; for instance, a website with interactive features once regarded as very interactive years ago (i.e., high EI) may not be perceived as being as interactive as before (i.e., low EI), since technologies continue to evolve, and the features on the website may be too common or outdated to be perceived to be considered as novel and interactive (Voorveld et al., 2011).
Dimensions of expected interactivity
Measuring a construct requires specifying its domain, boundary, and structure, which involves clarifying which and how many dimensions there should be. Since EI is not a whole new construct, but derived from an existing one (i.e., interactivity), we first reviewed a variety of conceptual dimensions of interactivity that previous studies have suggested, which include speed and control (Coyle and Thorson, 2001; Steuer, 1992), reciprocity (or bi-directionality; Downes and McMillan, 2000; McMillan, 2002) and synchronicity (or time; Liu, 2003; Liu and Shrum, 2009; McMillan and Hwang, 2002), responsiveness (Wu, 1999), personalization (Zack, 1993), connectedness (or hypertextuality; Ha and James, 1998; Sundar et al., 2003), choice complexity (Heeter, 2000), and role-exchange (Fortin and Dholakia, 2005), among many others.
Although all these dimensions may seem relevant to some extent, defining interactivity as consisting of these dimensions is problematic, because “it remains unclear how and to what extent these dimensions overlap, and more seriously, whether these dimensions indeed constitute interactivity with nothing else left out” (Sohn, 2011: 1333). Reciprocity and control are, for example, not separate, but related, the former as a cause and the latter as its outcome. That is, it is reciprocity that allows actors to have control over the process of interaction, but not the other way around (Johnson et al., 2006). Furthermore, some of the dimensions were derived from the technological aspect of a medium like the Internet, while others are derived from the perceptual aspect of user experiences, which obscures the ontological root of the dimensions. With interactivity being defined as consisting of such dimensions, its conceptual boundary and inner structure have remained unclear, which made researchers decide somewhat arbitrarily which and how many dimensions should be considered for measuring it (Rafaeli and Ariel, 2007; Walther et al., 2005).
Sohn (2011) recently suggested a comprehensive framework in which interactivity is defined as an experiential construct having three facets or dimensions – sensory, semantic, and behavioral. With the dimensions derived from the general structure of interaction experience, not from technological functions or features of specific communication media, he argued, the concept could have “a clearer, more intuitively understandable boundary and inner structure” (Sohn, 2011: 1324). In other words, technological functions/features deemed important to some may not be so to others, which makes it difficult to judge which and how many of them to include for defining the concept; the three dimensions underlying interaction experience are comprehensive enough and relatively invariant so as to be applicable to a wide variety of interaction forms and situations.
Furthermore, whereas previous conceptualizations of interactivity mostly concentrated on behavioral affordance (Norman, 1988) – the extent to which a medium has technological features allowing its users to do such things as modifying the messages/contents (i.e., control), giving feedback (e.g., bidirectionality), and so forth – the new framework includes sensory and semantic level of interactivity as well. Several studies have already shown that one may perceive a communication situation as interactive not only through behavioral engagement with the interaction partner, but also merely through sensory experience (Coyle and Thorson, 2001; Downes and McMillan, 2000; Morrison, 1998; Quiring and Schweiger, 2008) and/or rhetorical communication with it (Warnick et al., 2005). Sundar and Kim (2005), for example, found that subjects in their experiment perceived animated banner ads as more interactive than static ones even when no behavioral engagement was allowed. This suggests that the mere presence of an animated object may make one feel the situation is more interactive than without it. Also, Warnick et al. (2005) and Trammell et al. (2006) found that rhetorical or semantic features and strategies were as effective as a medium’s technological features in elevating the perceived interactivity among website visitors.
EI is considered a mirror image or “essentially the reverse of perceived interactivity” (Rafaeli and Ariel, 2007: 82), and we find it reasonable to define EI as consisting of the aforementioned three dimensions of perceived interactivity. Assuming the structural symmetry between EI and perceived interactivity is advantageous in that it would allow us to examine the dimension-to-dimension correspondence, people who have “a greater EI on a sensory level may not think a medium with a higher interactivity on behavioral level is interactive enough” (Sohn, 2011: 1332). Examining how the dimensions of the two constructs are interrelated would in turn enable a more systematic understanding of how perceived interactivity and EI are formed and modified in a circular manner (see Figure 1). Given the discussion, therefore, EI can be broadly defined as follows: Expected interactivity refers to a person’s learned belief of the extent to which s/he can experience interactivity at a sensory, semantic, and/or behavioral dimension during a prospective interaction with a particular category of interaction partners.
Interaction as a circular process.
Study 1: Generating the initial item pool
Following the psychometric procedure of scale development (Churchill, 1979; DeVellis, 1991; Furr and Bacharach, 2008), first of all, a pool of unique terms and concepts, related to what users expect to experience in mediated interactions, was generated by using a multi-method approach (McMillan and Hwang, 2002). The initial pool of key terms and concepts was first constructed based on an extensive review of literature on interactivity and then interviews with experts (e.g., researchers and practitioners in interactive communication), from which several broad open-ended questions with regard to individuals’ experiences of and expectations toward interactive media as a whole (e.g., the Internet) were derived.
Using the open-ended questions developed, an in-depth interview with users of the Internet (n = 22) was conducted not only for generating more relevant terms to be used for developing measurement items, but also for ascertaining the face validity of the tripartite conceptual model of EI, proposed in the preceding section. In the interview, the interviewees were asked to write on a space provided (1) as many features/attributes as possible they think necessary at a minimum for a website to be interactive; (2) as many features/attributes as possible they usually expect to see or experience before visiting a website; and (3) what product categories, they think, are most (or least) likely to have interactive websites. The last question was asked to collect preliminary information of respondents’ category-level expectations of interactivity, which would be used for studies at later stages.
The keywords the interviewees linked most frequently to interaction experience and EI include animation, attractive design, ease of navigation, sensation, user-friendly, personalization, feedback possibility, customizability, multimedia, colorful, eye-catching, user controllability, personalized messages (e.g., “good bye” or “thank you”), among many others. These keywords indicate that individuals indeed associate interactivity not only with behavioral affordance (e.g., user control, feedback possibility, customizability), but also with sensory stimulation (e.g., eye-catching, animation, colorful, sensation, attractive design) and/or semantic level of interaction (e.g., personalized messages, user-friendly). These preliminary results appear to correspond to the tripartite framework suggested by Sohn (2011), which indicates that using the framework for conceptualizing EI might be an adequate approach.
From the pool of key terms and concepts related to interaction experience, the most relevant and frequently mentioned ones were identified and collected, which includes responsive, diverse, dynamic, static, vivid, colorful, relevant, dull, surprising, inanimate, exciting, and complex, among many others. Note here that we focused on the general adjectives the interviewees used to describe their experiences and expectations to any specific medium, while excluding any terms related to technological features. With the selected terms and concepts, initial question items for measuring EI were constructed using seven-point semantic differentials anchored with “strongly agree” and “strongly disagree.” In order to ensure the face validity of the items, several researchers who are familiar with interactivity research were invited to review the initial pool of question items, and the items with the possibility of causing misunderstanding and/or confusion were removed or revised based on their comments and suggestions. Through this process, 50 measurement items in total were finalized for the next phase of the study.
Study 2: Item purification
In order to test the internal consistency among the 50 initial question items, an online survey was conducted. Students at a large state university were recruited, and participants received an extra course credit for participating in the survey. In total, 141 (n = 141) respondents participated in the survey. After a brief description of the study and an instruction for the survey, respondents were asked to imagine any website that was “highly interactive in their point of view,” and to rate their level of EI toward it on a series of scale items provided. The reason why we did not specify websites to rate in this study was to identify the measurement items that would be consistent across website categories individuals considered, since the measurement outcomes that are consistent only with some specific categories, but not with others, would be problematic, and limit the measurement items’ range of applicability.
After collecting and cleaning the data, first of all, two measures for testing internal consistency, item-to-total correlation and inter-item (pairwise) correlation, were calculated and examined. The criteria for item selection, following Nunnally (1967), were (1) the item-to-total correlation should be higher than .5; and (2) the inter-item correlation, meaning the pairwise correlations across dimensions, should not exceed .3. All the items that did not meet any of the two criteria were removed, and this process left the items correlated more strongly with items in the same dimension than with items in other dimensions. Further, an exploratory factor analysis (EFA) using the varimax rotation method was conducted to examine the underlying factor structure and to see if there was any new dimension within each dimension that was conceptually meaningful. Any item with a factor loading below .5 was also removed through this process. Through the procedure described above, a total of 16 items were retained out of 50, five for sensory (Cronbach’s alpha = .82) and another five for semantic dimension (α = .91), and the remaining six for behavioral dimension (α = .86).
Study 3: Latent structure analysis and validity assessment
Another online survey was carried out to examine whether the scale adequately reflected the proposed dimensional structure of the construct (i.e., whether the items purported to load on a factor actually loaded on it). Students at a large state university were recruited, and participants received an extra course credit for participating in the survey. In total, 237 (n = 237) respondents participated in the survey. Respondents were randomly assigned to each of the two versions of the questionnaire and asked to rate their EI toward a product category given, either automobile or furniture, on the 16 items retained from the previous stage. This was done to check the criterion validity of the scale – whether the scale correctly differentiates the positions of known groups (i.e., automobile versus furniture groups) in terms of EI.
The same procedure as was done in Study 2 was repeated to check the test–retest reliability of the items, and four items failing to satisfy the minimum requirement were removed, which left 12 items (see Table 1). In order to see if the latent dimensions of EI (i.e., sensory, semantic, and behavioral) satisfied the minimum psychometric requirements, each dimension was separately examined through the confirmatory factor analysis. The standardized factor loadings of the observed variables ranged from .61 to .89 (see Table 1). To check more clearly the reliability and validity of the latent constructs, the composite reliability (CR) and average variance extracted (AVE) of each dimension were calculated following the formula suggested by Fornell and Larcker (1981). CR indicates the extent to which a set of observed variables share common attributes, while AVE means the extent to which a latent factor explains the variance in the observed variables. A high score in both measures indicates the internal consistency and convergent validity of a latent dimension.
Expected interactivity scale items.
These items were reverse coded.
As shown in Table 1, all three dimensions’ CR is higher than .6, and their AVE is also greater than .5, which is the minimum requirement for convergent validity (Fornell and Larcker, 1981). For testing the discriminant validity of the latent dimensions, the squared correlation for each pair of dimensions (φ) was calculated and compared with each dimension’s AVE. It was found that all three dimensions’ AVE was greater than φ. This means that the commonality among the observed variables was greater within than between the latent dimensions, which shows they possess discriminant validity. If the measurement outcomes possess discriminant validity, a constrained model with more dimensions should have more explanatory power of given data than the other models with fewer dimensions (Anderson and Gerbing, 1988). In order to test whether the proposed three-factor correlated model of EI better explains the data than alternative ones, a series of confirmatory factor analyses were conducted, and the fit indices of the models were compared.
A critical assumption underlying the structural equation modeling (SEM) is multivariate normality, which means that all observed variables and their combinations in the model should be normally distributed (Joreskog, 1973). To see whether the assumption was met, the observed variables’ skewness and kurtosis were examined using the AMOS program: univariate skewness of the observed variables ranged from –.28 to .59, and univariate kurtosis ranged from –1.0 to –.18. A conventional method for judging non-normality is that the absolute value of a critical ratio larger than 2.0 indicates a substantial deviation from normality. Unfortunately, many observed variables in this study had critical ratios exceeding 2.0, which means the multivariate normality assumption was not met. It has been reported that the violation of the normality assumption might not only inflate the likelihood of the ratio statistic χ2 under the maximum likelihood estimation, but also underestimate some fit indices, such as the Comparative Fit Index (CFI) and Tucker–Lewis Index (TLI), which would increase Type II error (West et al., 1995). In order to correct this problem, the Bollen–Stine bootstrapping, which adjusts the critical value p of χ2 for the non-normal variables, was employed. Thus, note that parameter estimates and the p-values of χ2 in the following are adjusted by the bootstrapping.
As shown in Table 2, the proposed three-factor correlated model of EI was found to be the best among the seven models. The model’s χ2 was 91.45 (df = 52), which was not statistically significant (p = .085). The relative χ2 (the chi-square divided by the degree of freedom) was 1.76, which also satisfied the suggested minimum requirement (i.e., χ2/df < 2). In addition, the model comparison shows that the three-factor correlated model made a significant improvement in χ2 from the other models with fewer constraints. In particular, the improvement in χ2 made by the three-factor correlated model was 214.19 (p < .005) over the model with semantic and behavioral dimensions combined, 175.12 (p < .005) over the model with sensory and behavioral dimensions combined, and 169.17 (p < .005) over the model with sensory and semantic dimensions combined, respectively. This suggests that the proposed three-dimensional structure of EI represents the data better than alternative structures (see Figure 2). Among the proposed model’s fit indices, furthermore, the CFI, TLI, and Normed Fit Index (NFI) were above .90, and the Root Mean Square Error of Approximation (RMSEA) was below .08, which overall satisfies minimum requirements.
Confirmatory factor analysis results of competing models.
Note: S: sensory dimension; SM: semantic dimension; B: behavioral dimension; CFI: Comparative Fit Index; TLI: Tucker–Lewis Index; NFI: Normed Fit Index; RMSEA: Root Mean Square Error of Approximation.
Chi-square difference over the null model.
Chi-square difference over the one-factor model.
Chi-square difference over the two-factor models as ordered.
Chi-square difference over the three-factor uncorrelated model.
p < .005.
Three-factor correlated model of expected interactivity.
Criterion validity and relationships with other variables
Sohn et al. (2007) measured subjects’ EI for manipulation check with a single seven-point bipolar item anchored with interactive/not interactive, and found that they tend to have a higher EI toward the automobile category than the furniture product category. In order to test whether the EI scale developed differentiates the two product categories in the same way as done in the previous study, respondents’ EI toward the automobile and furniture categories, which purport to be high and low in EI, respectively, were measured using the scale. As expected, it was found that respondents had a higher EI toward automobiles (Mautomobile = 3.59) than furniture (Mfurniture = 3.25), and this difference was statistically significant, t (237) = 2.50, p < .05, which affirms that the current EI measure has criterion validity.
One may argue that EI is merely a reflection of, and therefore proportional to, individuals’ experience and skill level with the Internet – the more experienced a person is with the Internet, the higher EI s/he would have toward it. Also, it may be said that EI may not be a meaningfully distinctive construct, but merely a mirror image of individuals’ personality characteristics, such as NFC – the greater a person’s NFC, the higher EI s/he has toward a medium or issue domain. In order to test this, the bivariate correlations between EI and the characteristics mentioned above were examined by measuring Internet skill (Potosky, 2007) and NFC (Caccioppo and Petty, 1982). No statistically significant correlations were found between EI and Internet skill/experience level, r = .02, p = ns, and NFC, r = –.06, p = ns. On the other hand, the correlation between EI and respondents’ involvement with the product categories, measured using the scale developed by Zaichkowsky (1985), was statistically significant, r = .32, p < .01. This suggests that EI is independent of individuals’ level of skills and experiences or personality characteristics, but reflects their situational involvement with the domain of interest.
Discussion and implications
Our perception of something current is often based on the existing knowledge of similar kinds. This schematic knowledge or schema, defined as “an organized pattern of expectations toward a stimulus domain” (Folkes and Kiesler, 1998: 285), often serves as a temporal basis on which we perceive and evaluate things. The results of this study suggest that this is also the case when individuals perceive and interact with communication media – individuals have varying degrees of EI across issue domains. This implies that we do not always perceive and evaluate a medium or device for interactive communication by decomposing it into multiple features or traits for evaluation, but often rely on a broader experiential or schematic dimension of knowledge stored in our memory.
Given the notion, we may locate three distinct, but related, concepts – actual, expected, and perceived interactivity – in a circular process as illustrated in Figure 1. Actual interactivity refers to a medium’s potential interactivity that is realized only when “users engage perceptually and/or physically with the features” (Sohn, 2011: 1322). EI becomes the stage for such an actual user–medium interaction through which perceived interactivity is formed. In turn, perceived interactivity becomes an ingredient for the formation/modification of EI. This illuminates that expected, perceived, and actual interactivity are not separate, but merely temporal states inseparably interlocked in the circular process.
Viewing interaction as such a circular, generative process enables us to better explain how individuals’ dispositional characteristics, as well as situational factors, are intertwined in the causal process of interactive communication. Several empirical studies, for example, have reported that involvement and personality characteristics, such as the NFC, play an important moderating role in determining the effects of interactivity (Cauberghe and Pelsmacker, 2008; Jee and Lee, 2002; Liu and Shrum, 2009; Sicilia et al., 2005), but what actually happens in the intermediate process between the moderators and the outcomes (i.e., why and how they moderate the effects of interactivity) can be only speculated.
Measuring EI with the scale developed here would enable us to fill the gap beyond speculation by testing directly the process in which EI bridges various personal and situational antecedents with the outcomes of interaction. For example, the level of involvement may moderate interactivity effects through individuals’ EI (i.e., mediated moderation) – people with a greater involvement with an issue domain tend to hold a higher EI and be more ready to engage in interactive communication than those with a lower involvement. This speculation seems to explain the positive correlation between EI and product involvement, found in this study, but is subject to a more rigorous empirical test in the future. Likewise, examining other personal or situational factors in relation to EI and perceived interactivity would reveal a clearer picture of how people experience interactive communication.
In a practical sense, being able to measure EI would also help us plan, design, and optimize user interfaces/user experiences (UI/UX) in a variety of contexts. As blindly opting for technological fads (i.e., increasing interactivity by including more features) without considering what the target audiences expect may result in negative returns on investment, mapping the prospective users’ EI in various issue domains before actually designing the interfaces may be a sensible way to avoid potential pitfalls. This practical benefit is not limited to designing interfaces for human–computer interaction (HCI), but may also span to computer-mediated social interaction, because EI captures not only what we expect of user interfaces, but also the (perhaps interpersonal) interaction possibilities lying beyond them. For example, the users of social media like Facebook and Twitter engage in social interactions with numerous others everyday through the designed user interfaces. In this case, one’s EI toward an online social networking site can be a function not only of the characteristics of the user interfaces, but also of the various properties of his or her own social networks (e.g., the size/range of the network, the person’s location in the network, etc.).
As user interfaces of online social interactions become more diverse and sophisticated, the interactive communication experiences individuals have would become increasingly complex. Measuring EI with the scale developed here can be a good starting point for disentangling the complex relationships between user interfaces and user experiences/expectations in the ever-widening networked communication environment. In this vein, further research attempts should be made to illuminate various issues, which includes (1) how different people’s EI is toward non-human entities than toward humans; (2) how EI toward online social networks varies depending on the structural properties of the networks; and (3) how EI toward online social networks is influenced by the characteristics of the user interfaces applied. Answering these questions would not only facilitate our understanding of the mediated social experience, but also provides practical insights and guidelines for designing a better communication environment.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.