Page - 253 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

4.1.4 Diffusion of Innovations (DoI) Model According to Rogers [21], the DoI model explains how new ideas are adopted and spread. Its five predictive factors are defined in terms of which an innovation: 1) Relative Advantage: improves status, prevents aversiveness, or is incentivized; 2) Compatibility: aligns to values, experiences, needs, or accepted innovations; 3) Complexity: has a steep learning curve or is difficult to understand; 4) Trialability: is possible for users to sample before investing fully; and 5) Observability: results in outcomes that are easy to visualize or describe. According to the Diffusion of Innovations (DoI) model, the more advantageous, compatible, trialable, and observable users believe an innovation is, the more likely they are to adopt it. Conversely, the more difficult it is to understand it, the less likely users are to adopt it. The DoI gave rise to many other models, as illustrated in [24]. S.K. Sharma, Joshi, and H. Sharma [24] took a multi-analytical approach to predicting the educational use of Facebook. The five variables derived from preceding models all originate from DoI: Social Influence (UTAUT), Perceived Usefulness (TAM), Perceived Enjoyment (UTAUT2 [24]), and Resource Sharing and Collaboration ([22] and [23]). Based on survey responses from 215 college students, all variables influenced Facebook use in structural equation and neural network models. 4.2 Effects of Variables on Endorsement and Use Across Statistical Models The number of variables differed across models: Akman and Turhan [18] tested six, Saini and Abraham [11] used 13, Premadasa and colleagues [19] used 14, Sharma and colleagues [24] tested five, and Balakrishnan [20] tested six constructs with 13 untested variables. The models overlapped slightly in terms of significant effects. Table 3 lists the direct and indirect effects of these variables on endorsement and use. Considering the four models that tested all variables measured by the several studies described above ([11][18][19][24]), some patterns emerge. First, a total of 20 variables had direct and indirect effects on endorsement and/or usage across studies. Second, Perceived Usefulness was the only factor with significant effects in all four models. Third, Social Influence, Collaboration, and Resource Sharing were supported by at least three models. Fourth, the models in two aforementioned studies [11] [19] supported 10 variables not examined in other models, although their comparable findings lent some generality of effects. Fifth, three studies [11], [18], and [19] tested factors unique to each of those studies, making the generality of Motivational Influence, Security Awareness, Ethics Awareness, Mobility, and Interactivity unclear. Based on these findings, a future model might start with these four factors most supported by current research: Perceived Usefulness, Social Influence, Collaboration, and Resource Sharing. Future models should also measure actual use directly, such as rate of posts per week, time students spent logged in, or the number of resources shared. All models analyzed here used survey questions to measure actual usage. It would be time-consuming but worthwhile to determine how survey responses compare to observable behavior, as a significant mismatch would compromise the validity of any findings. E.Damianoetal. /Bridging theDivide: TheCurrentStatusofWeb2.0 inHigherEducation 253