Technology Induction in Education During COVID-19 is Recreation or a Curse?: Integration of Technological and Behavioral Factors from the Students’ Perspective

The term compatibility explains the extent to which an individual experiences innovation consistent with their current needs, experiences, values, and the degree to which new technology is seen to be in line with prevailing needs (Hair et al., 2017; Isaac et al., 2019). The way that a student’s perceived technological compatibility is perceived to be in line with their needs significantly impacts the online learning experience. (Cheng, 2015; Isaac et al., 2016; Islam & Azad, 2015). Some studies provide evidence that compatibility notably affects perceived usefulness (Ozturk et al., 2016; Wu & Wang, 2005). Previous studies on online and internet learning contexts have shown interest in this impact on online learning (Isaac et al., 2016; Islam & Azad, 2015; Premkumar, 2003).

Perceived usefulness (PU) describes individuals’ believe that the available technology or learning is useful and improves their performance (Chang & Tung, 2008; Lee, 2006; Venkatesh & Davis, 2000). Many studies proved perceived usefulness as the strongest antecedent of online learning (Chang & Tung, 2008; Gefen & Straub, 2003; Venkatesh & Davis, 2000). When a person perceives that online technology and e-learning will enhance his learning experience, it will significantly impact learning readiness that ultimately leads individuals’ perception towards online learning (Hsu & Lu, 2004; Lee, 2006; Ong et al., 2004). The current study measures the impact of the perceived usefulness of technologic induction in online learning.

The perceived ease of use (PEU) describes how much a person feels the particular technology is user-friendly, easy to use, and free from effort (Chang & Tung, 2008; Chin & Todd, 1995; Doll et al., 1998). Previous studies on perceived ease of use and perceived usefulness empirically proved that if students feel at ease and comfortable while using technology, it will also impact their perception of its usefulness (Chang & Tung, 2008; Gefen et al., 2003; Venkatesh & Davis, 2000). Other studies demonstrate a clear understanding of the effect perceived ease of use has on a student’s online learning interaction and experiences. When students feel at ease while using the technology or taking online classes, it will have a direct and positive impact on their online learning experiences (Chang & Tung, 2008; Davis et al., 1992; Gefen & Straub, 1997).

As human characteristics and personalities vary significantly, computer self-efficacy plays a significant role in technology adoption (Kreth et al., 2019; Mahdavian et al., 2016; Prior et al., 2016). Self-efficacy is defined as a personality trait related to how an individual believes that he or she may complete the task or will experience the learning online course (Kreth et al., 2019; Yu et al., 2013). Computer self-efficacy is describes the level of confidence a person has that he or she can perform computer-related tasks easily and can accomplish learning by using the computer (Mahdavian et al., 2016; Prior et al., 2016). Previously, self-efficacy had been extensively tested in the learning and teaching setting (Isaac et al., 2019). Several studies have demonstrated that if an individual has high technology or computer self-efficacy, then he or she might be more inclined towards e-learning (Kreth et al., 2019; You et al., 2012). From the perspective of online learning, students’ strong belief in their computer self-efficacy and low resistence will positively affect their online learning and they will experience it as recreation (Isaac et al., 2019; Yu et al., 2013).

Relative advantage is defined as a new technology, practice, or innovation that is supposed to have a potential advantage and be a better tool in comparison to old practices (Hsbollah, 2009a; Winters & Martins, 2004). Relative advantage can also describe the degree to which innovation and new technology are perceived to be superior to current technology, and thus its adoption will be beneficial (Bennett & Bennett, 2003). Several studies have demonstrated that relative advantage is an antecedent of technology acceptance and adoption, which in our case is online learning (Hsbollah, 2009b). Kendall et al. (2001), found that relative advantage has emerged as the most influential factor that will affect the willingness of learners and adopters to accept the new technology. In this way, it is hypothesized that relative advantage in the context of e-learning or online learning will enhance the learning experience in instructional technology (Hsbollah, 2009a; D. I. MacKenzie & Kendall, 2002).

Technological advancement and development of rapid innovation made providing education much easier than it would have been, especially during the COVID-19 pandemic (Dhawan, 2020, McBrien et al., 2009). During COVID-19, this tool became the only way to continue the educational process while maintaining physical distance without wasting students’ time and causing academic losses (Basilaia & Kvavadze, 2020; Dhawan, 2020). To address this challenge of ‘Stay at Home’ learning, researchers, learners, educationists, students, teachers, and higher education committees advocated that transforming traditional learning into online learning systems to continue to educate the students (Dhawan, 2020). Online learning includes the terms e-learning, blended learning, open learning, computer-mediated learning, web-based learning, m-learning with common characteristics to use the computer, networking with an internet connection or mobile devices, or networking that can be possible anywhere, anytime, and in any means by using mobiles and laptops (Cojocariu & Boghian, 2014; Dhawan, 2020). Online learning tools may be more effective than traditional tools, because it is a more interactive, innovative, and student-centric approach. In these environments, students can easily interact with teachers, and students may participate more actively (Singh & Thurman, 2019). Online learning is designed in a way that students can attend online classrooms, virtual recorded lectures, interact in real-time, and give feedback during class (Currie-Mueller & Littlefield, 2018).

Online learning can be defined as a synchronous or asynchronous environment depending on whether students are using mobile devices or laptops with internet access. Students are independent in these environments to learn and correspond with other students and teachers (Singh & Thurman, 2019). The synchronous environment provides real-time experience, live lectures, classes, webinars, workshops, video conferencing, and other online learning systems in social interaction, whereas asynchronous learning is not well structured (McBrien et al., 2009). Because of the need to stop the spread of a deadly virus, COVID-19, the need for synchronous learning or online learning emerged in which online classes, webinars, online meetings and discussion, internet connection quality, mobile, and laptop availability, instant feedback, all become the crucial needs of the time (Basilaia & Kvavadze, 2020; Dhawan, 2020, Hastie et al., 2010; Szeto, 2014). In keeping with the above-discussed literature, we formed the following hypotheses:

H₁: Compatibility has a direct influence on perceived usefulness.

The instrument of this study was adopted from existing studies. All 20 items used in this study were answered on the 5 point-Likert scale with 1 being “strongly disagree” and 5 representing “strongly agree”. The questionnaire was divided into three parts: Section 1 included demographics information like age, gender, education, university, and a filter regarding taking online classes during COVID-19. The second section consisted of items related to independent and dependent variables. The construct of compatibility was adopted from IDT and defined by Wu and Wang (2005) as containing three items. In our study, the variables of perceived usefulness and perceived ease of use, containing four items each, were taken from TAM and adapted from Vankatesh and Davis (1996, 2000). The construct of computer self-efficacy was adapted from Vijayasarathy (2004) and contained two items. From an extensive literature review, it is evident that there is no scale so far developed; however, we adapted a scale from a study by Karahanna (2002), who developed three items of relative advantage after face-to-face interviews for the study. The dependent variable is adapted from Davraj (2002) and Harnandez (2009) and consists of two items. The items of the dependent variable used for this study were re-worded where needed. To check the reliability, a pilot study was conducted which shows strong internal consistency.

Descriptive statistics showed that 204 (52.71%) out of 387 respondents were male and 47.29% were female; 94.57% were between the ages of 20 and 30 years, 4.91% were between the age of 31 and 40 years, and 0.52% were above the age of 40. Among the respondents, 79.07% were enrolled in a bachelor’s degree program (16 years), 17.05% were enrolled in master’s degree programs (18 years), and 3.88% were enrolled in Ph.D. programs. The largest number of respondents (82) was from Bahria University, 75 from SZABST, 75 from Textile Institute of Pakistan, 29 from the University of Education, 25 from COMSATs, 23 from IBA Sukkur, 22 from DOW, 15 from the University of Agriculture, 13 from National University of Modern Languages, 10 from GIFT, 8 from Institute of Business Management, 7 from University of Punjab, and 5 from Indus University,.

Structural equation modelling (SEM) was employed because it can handle complex models and run completely at the same time (Schumacker & Lomax, 2012; Tabachnick et al., 2007). According to Anderson and Gerbing (1988), SEM works on two levels: measurement and structural models. Details of these two have been provided in the results and interpretation section. Smart PLS 3.2.9 was utilized for data analysis using the PLS-SEM approach. This software was used for two reasons: first, this study developed a novel framework by integrating three theories. This study was exploratory based on developing theory (Hair et al., 2017). Second, data is not normally distributed in the survey research. Using the PLS-SEM approach through Smart PLS does not require data normality assumptions to be fulfilled (Chin et al., 2003).

Common method bias (CMB) can be a potential threat for self-reported research surveys (Podsakoff, 2003) as data were collected from a single source (S. B. MacKenzie & Podsakoff, 2012). Harman’s single factor was employed through exploratory factor analysis (EFA) to detect CMB. It was indicated that the first factor described a 36.319% variance which, being less than 50%, confirmed CMB was not a problem in our study (Babin et al., 2016).

Figure 1 shows the findings of validity and reliability using the measurement model. The reliability of the constructs was assessed through CA and CR. According to Hair et al. (2017), values for CA and CR should be greater than 0.70. All values of CA and CR were greater than 0.70, ensuring the reliability of the constructs as indicated by Table 1.

Figure 1

For assessing convergent validity (CV) was used. For this purpose, outer loadings and average variance extracted (AVE) were evaluated. According to Hair et al. (2017), AVE should be greater than 0.50. Outer loadings above 0.70 are considered satisfactory, values between 0.40 and 0.70 should be removed if the value of AVE is lower than 0.50, values for Cronbach alpha (CA) and Composite reliability (CR) are less than 0.70. In our study, all outer loadings were greater than 0.70 except PEU4 (0.486), but it was not removed because criteria for validity and reliability were met as shown in Table 1. For assessing discriminant validity (DV), the heterotrait–monotrait (HTMT) correlation ratio was utilized. According to Henseler et al. (2015), all values in the correlation table should be less than 0.90 for DV to exist. As shown in Table 2, all values were below 0.90, thus showing the presence of DV.

Figure 2 shows the relationships among the constructs and the model’s explanatory and predictive power. The bootstrapping procedure was performed using a resample of 5000 as suggested by Hair et al. (2017) and Ramayah et al. (2018) for evaluating the structural model. Multi-collinearity was assessed through the variance inflation factor (VIF) values as suggested by Hair et al. (2017). The VIF value should be less than five, as indicated in Table 3 showing the absence of multi-collinearity. For evaluating the explanatory power of the model, the R² value was determined for perceived usefulness (PU) and online learning experience (OLE). Cohen et al. (1988) categorized the R² values of 0.02, 0.13, and 0.26 for endogenous constructs as weak, moderate, and substantial respectively. In our study, both values were above 0.26, indicating a substantial effect of exogenous constructs on endogenous constructs (PU and OLE). Effect values of individual exogenous on endogenous constructs were reported through (f²) values. f² values of 0.02, 0.15, and 0.35 show small, medium, and large effect values, respectively (Cohen, 1988). The effect values of C, CSE, and RA were less than 0.02, showing almost no effect. PU, PEU on OLE, and PEU on PU had moderate effect, C on PU had a large effect value, as shown in Table 3. A blindfolding procedure was utilized for assessing the predictive relevance (Q²) of the model. According to Richter et al. (2016), predictive relevance should be greater than zero, (Q²) for PU was 0.504, and OLE was 0.449, which demonstrates the presence of predictive relevance.

Figure 2

Figure 3

For this study, all hypotheses were developed based on direct relationships among the constructs. The first hypothesis of the study showed a direct relationship between C and OLE. This relationship was found to be positive and significant as t value was more than 1.96, LL and UL do not straddle between zero as shown in Table 3 (beta = 0.119; t value = 2.395; LL = 0.024; UL = 0.217). Similarly, the second hypothesis indicated the impact of C on PU and was also found to be strongly positively significant (beta = 0.529; t value = 13.364; LL = 0.452; UL = 0.606). The third hypothesis consisted of the influence of CSE on OLE and was not supported because the t value was less than 1.96, two tail, LL, and UL straddle zero (beta= 0.055; t value = 1.200; LL = −0.033; UL = 0.145). The fourth hypothesis determined positive association between PU and OLE and was supported (beta = 0.442; t value = 7.641; LL = −0.331, UL = 0.557). The fifth hypothesis indicated the impact of PEU on OLE and was also found to be positively significant (beta = 0.297; t value = 6.097; LL = 0.198; UL = 0.389). The sixth hypothesis showed the influence of PEU on PU and was also supported (beta = 0.346; t value = 7.420; LL = 0.246; UL = 0.431). Lastly, the seventh hypothesis, which showed a direct relationship between RA and OLE, was not supported (beta = 0.022; t value = 0.425; LL = −0.076; UL = 0.130).

Priority map analysis or importance-performance map analysis (IPMA) was performed as given in Table 4 and represents a contrast of performance between average of latent variable scores (LVS) and importance, which is the total effect of all predictors for a target construct (Ringle & Sarstedt, 2016). A one-unit point increase in the performance of the predicting construct will increase the performance of the target construct, by the total effect size (i.e. importance) of the same predicting construct (Ringle & Sarstedt, 2016; Schloderer et al., 2014). In our study, OLE was the target construct whereas C, CSE, PU, PEU, and RA (5) were direct predictors. The main aim of the IPMA was to identify those constructs which have relatively high importance but low performance. The highest value of importance (total effect) was 0.444 for PEU. It means that if PEU is increased by 1 unit, it will increase OLE by 0.444 of the university students. Similarly, RA is the lowest performer among all the constructs; improving RA will increase the overall OLE of university students.

This study, like others, has some limitations. A small sample size is used in this study to generalize the online learning experience of students throughout Pakistan. Secondly, technological know-how varies across Pakistan, which inhibits the generalization of results across Pakistan.

Future studies can be conducted by linking online learning intention and behavior. Moreover, further study may be carried out by using different sampling techniques, such as probability sampling that could be helpful in the generalization of results. This study is conducted in the Pakistani context which is not much advanced in technological and innovation know-how. Further research can be conducted in technologically advanced countries.