Sage Journals: Discover world-class research

Abstract

Computer self-efficacy (CSE) continues to be an important construct in research and application. Two measures of CSE, the Brief Inventory of Technology Self-Efficacy (BITS) and the Brief Inventory of Technology Self-Efficacy – Short Form (BITS-SF) were recently developed to correct for issues in other available measures. The BITS and BITS-SF were originally written in English, and their psychometric properties assessed in samples from the United States. The current two studies translated the BITS and BITS-SF into simplified Chinese (Mainland China) and traditional Chinese (Taiwan) and assessed their psychometric properties. In Study 1, 207 adults in Mainland China completed the simplified Chinese BITS and BITS-SF, as well as measures given to assess convergent, discriminant, and concurrent validity. In Study 2, 273 adults in Taiwan did the same, except that they completed the traditional Chinese BITS and BITS-SF. In both studies, the translated BITS showed evidence of a three-factor correlated structure, and the translated BITS-SF yielded several underlying classes consistent with theory and scoring interpretation. Additionally, the translated measures’ scores showed solid evidence of convergent, discriminant, and concurrent validity. The results replicate the findings using the original BITS and BITS-SF and extend them to simplified Chinese and traditional Chinese translated versions. These versions are recommended for use in research and applied settings to assess CSE and are available for use. Both the original and translated measures are available for download at www.bitssurvey.com.

Keywords

Brief Inventory of Technology Self-Efficacy (BITS)Chinese Taiwanese translation validation

Computer skills are considered necessary for successful education and employment (Peng, 2017; Stošić, 2015). Consequently, there has been an interest in factors contributing to employees’ and students’ adoption and use of technology (Chibisa et al., 2021; Ferdousi, 2019). Of these factors, one of the most emphasized in current research is computer self-efficacy (CSE; Gupta & Bostrom, 2019). CSE is a person’s confidence in their ability to successfully engage with computers and similar technology (Compeau & Higgins, 1995). Over the past decades, a robust literature base has emerged regarding the positive predictive role of CSE on the use of technology (e.g., Chibisa et al., 2021; Compeau & Higgins, 1995). As a result, numerous measures of CSE have been developed (see Gupta & Bostrom, 2019, for a review). However, many of these measures currently suffer from conceptual or methodological issues (Howard, 2014; Marakas et al., 2007). Two CSE measures were recently developed to correct for these issues: the Brief Inventory of Technology Self-Efficacy (BITS) and the Brief Inventory of Technology Self-Efficacy – Short Form (BITS-SF; Weigold A. & Weigold I.K., 2021). Both measures are available for download at www.bitssurvey.com. Given the importance of CSE in technology research (Gupta & Bostrom, 2019), it is necessary that current measures are available for use with diverse populations. Consequently, in the current studies, we translated the BITS and BITS-SF into simplified Chinese (Mainland China) and traditional Chinese (Taiwan). This is also the second set of studies reporting on the psychometric properties of the BITS and BITS-SF, providing both a replication and extension of the original studies.

Computer Self-Efficacy

Self-efficacy is an individual’s judgment of their ability to engage in behaviors that will lead to desired outcomes. Self-efficacy influences aspects of performance, including a person’s cognitions, emotions, and behaviors. Even when a person has sufficient ability to successfully engage in a behavior, they may not do so if they have low self-efficacy (Bandura, 1982). Bandura (2006) also emphasized that self-efficacy is domain-specific such that a person might have high self-efficacy in one area (e.g., writing a paper) but not another (e.g., using computers); consequently, each self-efficacy measure should only assess a specific domain, such as CSE.

CSE research has shown extensive support for CSE’s role in computer and technology use (e.g., Chibisa et al., 2021; Hsia et al., 2014; Thongsri et al., 2020). For example, CSE mediated the relations between several predictor variables and computer use in a sample of pre-service teachers in South Africa (Chibisa et al., 2021). Additionally, CSE was a significant positive predictor of intentions toward e-learning for Chinese college students (Thongsri et al., 2020) and both perceived ease of e-learning and intentions to engage in e-learning for Taiwanese employees of high-tech companies (Hsia et al., 2014). Researchers have also recommended that CSE be considered when designing computer trainings or implementing new technology (Ferdousi, 2019; see Hsia et al., 2014), as simple exposure and access to computers and the Internet are not sufficient for skillful use (Peng, 2017).

Over the last decades, many CSE measures have been developed; however, most of them have been critiqued for issues such as outdated items, problematic operationalization, criterion contamination, and limited range (Howard, 2014; Marakas et al., 2007). Scales often contain outdated items due to the rapid development of technology; however, simply removing outdated items when administering measures can lead to construct underrepresentation (Marakas et al., 2007). Additionally, measures may not be accurate representations of the CSE construct. For example, the Computer User Self-Efficacy Scale by Cassidy and Eachus (2002) purports to assess self-efficacy, although its instructions ask participants to respond to “statements concerning how you might feel about computers” (p. 148), rather than questions about confidence (Howard, 2014). Next, many scales assess other constructs in addition to CSE. For example, the authors of the Computer Self-Efficacy Measure, which is one of the most often-used CSE scales (Howard, 2014), have stated that scores on this measure can be influenced by respondents’ ability to think about hypothetical situations and learning self-efficacy (Compeau & Higgins, 1995). Finally, almost no CSE measures assess confidence across different computer skill levels (Weigold A. & Weigold I.K., 2021). One exception is the Computer Self-Efficacy Scale (Murphy et al., 1989); however, this scale contains many outdated items (Marakas et al., 2007). Due to these issues, researchers have recently developed two measures, the BITS and BITS-SF, designed to provide those in research and applied settings with current, psychometrically strong CSE scales (Weigold & Weigold, 2021).

Brief Inventory of Technology Self-Efficacy

The BITS and BITS-SF were developed and assessed using seven studies (Weigold & Weigold, 2021). The authors’ stated goals were to create one full-length assessment (BITS) and one screening tool (BITS-SF) that evaluated confidence for novice, advanced, and expert computer skills and could be used in research and applied settings. The underlying theory was the five-stage model of adult skill acquisition (Dreyfus, 2004). In this model, learners pass through several stages: novice, advanced beginner, competence, proficiency, and expert. Weigold A. & Weigold I.K. ’s (2021) scales emphasize the novice, competence (which they called advanced), and expert stages. Novice computer skills were defined as those “required for modern daily computer use,” advanced as those needing some knowledge of appropriate use, and expert as those requiring specialized training (Weigold A. & Weigold I.K., 2021, p. 4). Within each level, Weigold and Weigold included skills from six computer use categories: hardware, software, Internet, operating system, networking, and troubleshooting. The final BITS is comprised of 18 items, with six items each (one per computer use category) on the novice, advanced, and expert subscales. Higher scores indicate higher levels of self-efficacy for computer skills of the corresponding level. The BITS-SF was developed based on the strongest items of the BITS and consists of six items, two from each of the subscales on the BITS and one from each of the computer use categories. The BITS-SF yields a total score ranging from 0 (negligible CSE) to 6 (expert skill level CSE).

The BITS and BITS-SF items were developed based on a survey of computer experts regarding what skills novice, advanced, and expert computer users should have and a series of analyses using several independent samples (Weigold A. & Weigold I.K., 2021). Confirmatory factor analyses provided support for the three-factor correlated structure of the BITS, which was a better fit than the single-factor (ΔAIC = 1,275.82) and three-factor uncorrelated (ΔAIC = 145.62) structures. The three factors also showed strong evidence of internal consistency, with Cronbach’s alpha levels of .86 (novice), .87 (advanced), and .90 (expert). A test of multigroup invariance comparing men and women supported configural (item associations), metric (item loadings), and scalar (item intercepts) invariance for the novice and advanced subscales, and configural and metric invariance for the expert subscale. Relatedly, a latent class analysis of the BITS-SF indicated there were three underlying classes corresponding to self-efficacy for novice, advanced, and expert computer skills. Items indicating novice self-efficacy had a high likelihood of being answered affirmatively by participants in all classes; items suggesting advanced self-efficacy had a high likelihood of being answered positively by those in the advanced and expert classes; and items indicating expert self-efficacy had a moderate-to-high likelihood of being answered affirmatively by those in the expert class.

Both the BITS subscale and BITS-SF total scores generally correlated positively and significantly with 10 measures of convergent validity (e.g., measures of CSE) and generally had small correlations with 11 measures of discriminant validity (e.g., measures of personality and self-esteem). These scores also showed evidence of concurrent validity related to self-identified computer skill level (i.e., those reporting higher skill levels also had significantly higher scores on the BITS and BITS-SF than those indicating lower skill levels) and major in college (i.e., students majoring in engineering and computer science had significantly higher scores on the BITS and BITS-SF than students majoring in education and nursing). Finally, the BITS subscale scores showed strong test-retest reliability for 2- and 8-week periods (ICC range = .80 for novice to .93 for expert, both at eight weeks; Weigold A. & Weigold I.K., 2021).

Taken together, the BITS and BITS-SF evidenced solid psychometric properties related to internal consistency; test-retest reliability; convergent, discriminant, and concurrent validity; and either the factor structure (BITS) or underlying class structure (BITS-SF). However, the development article (Weigold A. & Weigold I.K., 2021) is currently the only one to report on these properties. Additionally, the BITS and BITS-SF were written in English and assessed using United States college students and Mechanical Turk workers, and it is unclear to what extent these measures can be used with other samples. Finally, although the BITS-SF total score potentially includes 0, indicating negligible CSE (Weigold, 2021), no participant in the original latent class analysis scored below 1 (Weigold A. & Weigold I.K., 2021). As a result, further studies are needed to determine the degree to which negligible CSE, as defined by the BITS-SF, exists in current populations; this might have implications for how the BITS-SF is scored. Consequently, in the current studies, we translated the BITS and BITS-SF into simplified and traditional Chinese and examined their psychometric properties in adults residing in Mainland China and Taiwan, respectively.

Relevance of CSE to Mainland China and Taiwan

Both Mainland China and Taiwan have introduced technology-related initiatives with implications for research, application, and education. For example, Mainland China recently began the 15-year China Brain Project (officially titled “Brain Science and Brain-Inspired Intelligence”) in which brain research focuses both on brain-related diseases and “brain-inspired intelligence technology,” such as artificial intelligence (Poo et al., 2016, p. 591). Mainland China has also worked toward the incorporation of information and communication technology (ICT) into education since the 1980s (Zeng, 2022). The most recent five-year plan related to the ICT for Education initiative began in 2021 and focuses on digital technologies (Cyberspace Administration of China, 2021). The goals of these plans are to provide infrastructure and related assets to facilitate students’ learning of digital technologies, and they include providing resources to millions of teachers and lessening the ICT gap between urban and rural schools (Bajpai et al., 2019).

Taiwan has also introduced initiatives related to technology. For example, Taiwan has recently made large investments into artificial intelligence and medical devices, with a focus on becoming the “Silicon Valley of Asia” and a “Hub of Biotech and Medical R&D in Asia” (Haldorai et al., 2020, p. 569). The country has also invested in e-learning and digital technologies since 2002, with the first five-year plan showing positive results for the business sector and e-learning (Chang et al., 2009). More recently, Taiwan has focused on digital education by allocating money for tablets, high-speed Internet, and smart classrooms in schools, as well as worked to reduce the urban-rural education divide through technology allocation (Hsin-fang & Tzu-hsuan, 2022).

Given the initiatives in Mainland China and Taiwan, there is a need for current measures of CSE that can be used to best tailor computer instruction in educational and business settings. For example, Tsai (2019) found that using a specific intervention to facilitate learning basic computer programming skills in Taiwanese college students was most helpful for those with low and moderate self-efficacy. Consequently, the current study focused on translating the BITS and BITS-SF into simplified and traditional Chinese and assessing their psychometric properties to make them widely available for use with Chinese-reading populations.

The Current Studies

To address the need for updated and current CSE measures, as well as replicate and extend the findings of the original BITS and BITS-SF studies (Weigold A. & Weigold I.K., 2021), we conducted two studies. The first focused on the translation of the BITS and BITS-SF into simplified Chinese (BITS-SC and BITS-SF-SC) and the examination of the translated measures’ psychometric properties in adults living in Mainland China. The second was similar to the first, except the measures were translated into traditional Chinese (BITS-TC and BITS-SF-TC) and assessed using adults in Taiwan.

We made similar hypotheses for both studies, which were based on the results described in the BITS and BITS-SF development article (Weigold A. & Weigold I.K., 2021). First, we expected the translated BITS to show evidence of a three-factor correlated structure corresponding to the novice, advanced, and expert subscales (Hypothesis 1a). We also expected this model to be a better fit than either single-factor or three-factor uncorrelated models (Hypothesis 1b).

Second, we hypothesized that the translated BITS’ subscale and translated BITS-SF’s total scores would show evidence of convergent validity by having positive, moderate-to-strong correlations with other measures of computer self-efficacy (Hypothesis 2a), as well as discriminant validity by yielding small correlations with self-esteem (Hypothesis 2b). We also expected the measures to show evidence of concurrent validity such that respondents with higher self-identified computer skill levels and occupations requiring specialized computer training would score higher on the translated BITS’ subscale and translated BITS-SF’s total scores than those with lower self-identified computer skill levels (Hypothesis 2c) and occupations requiring less computer training (Hypothesis 2d).

Finally, we hypothesized that the translated BITS-SF would show evidence of either three or four underlying classes (Hypothesis 3a). The original BITS-SF latent class analysis yielded three underlying classes corresponding to self-efficacy for novice, advanced, and expert levels. The authors reported that scores of 1 and 2 indicate novice skill level CSE, 3 indicates novice-to-advanced skill level CSE, 4 indicates advanced skill level CSE, 5 indicates advanced-to-expert skill level CSE, and 6 indicates expert skill level CSE (Weigold A. & Weigold I.K., 2021); this pattern corresponds to the five theoretical stages of the model of skill acquisition (Dreyfus, 2004). However, it is also possible that a fourth class could emerge corresponding to negligible self-efficacy. This would be consistent with the scoring and interpretation of the BITS-SF (Weigold, 2021). We also expected that the translated BITS-SF’s underlying classes would show evidence of convergent validity such that scores on the translated BITS and additional measures of CSE would be higher as class increased from negligible/novice to expert CSE.

Study 1: Simplified Chinese

Method

Participants

Participants were 207 adults (18 years and older) living in Mainland China. They identified primarily as women (n = 123, 59.4%), followed by men (n = 84, 40.6%). Ages ranged from 19 to 75 (M = 42.86, SD = 9.15). Half of participants reported their occupation as teachers (n = 111, 53.6%), whereas others were corporate workers (n = 28, 13.5%), IT engineers (n = 19, 9.2%), public servants (n = 8, 3.9%), workers (n = 7, 3.4%), farmers (n = 6, 2.9%), unemployed (n = 20, 9.7%), or other (n = 8, 4.4%).

Measures

Brief Inventory of Technology Self-Efficacy – Simplified Chinese. We translated the BITS-SC, BITS-SF-SC, and their instructions and scoring guides from English to simplified Chinese. Various measure adaptation approaches exist, with two of the most common being the use of translation teams, such as the one specified by the Translation, Review, Adjudication, Pretesting, and Documentation (TRAPD) model, and forward and backward translation (Harkness et al., 2010; International Test Commission, 2017). Although backward translation can detect some mistakes, its use has long been considered insufficient for the appropriate cross-cultural adaptation of measures due to the multitude of errors that it does not identify (Behr & Braun, 2022). The preferred method is team approaches involving appropriate experts, as they require collaboration and can better detect nuanced errors (Behr & Braun, 2022; Harkness et al., 2010). However, backward translation can be combined with other methods to provide an additional check (International Test Commission, 2017). Consequently, we used forward and backward translation embedded within a translation team approach.

For the first step, we employed a professional translator to perform forward translations. We contacted this translator through a large online marketplace in the United States that specifically advertises translation services. This person was a native writer of simplified Chinese and specialized in both English to simplified Chinese translations and ICT content. After agreeing to provide the service, the translator completed the translations of the BITS and BITS-SF into simplified Chinese, then sent us the translated files.

Upon receiving the translations, we reviewed both the English and simplified Chinese versions as a five-member team. One team member has a doctorate in education, expertise in CSE, and lives in Mainland China. Two other team members are international doctoral students studying psychology in the United States whose home countries are Taiwan and Singapore. All three are fluent in both simplified Chinese and English; they also have relevant knowledge of self-report surveys, although they do not have prior translation experience. The other two team members have doctorates in psychology, live in the United States, are fluent in English, and are experts in CSE and measure development. These two team members have prior experience in both developing and translating measures, and one has worked as a professional translator.

Following the suggestions of the International Testing Commission (2017), the stated goal for the team was to critically examine the translations “consider[ing] linguistic, psychological, and cultural differences” so that the final measures would be accurate, read fluently, and appear indigenous (p. 11). First, the three team members fluent in both English and simplified Chinese, who were also familiar with the target cultures, separately reviewed both the original English and simplified Chinese measures. Each of these team members provided comments about the translation to the full five-person team, such as suggesting different word choices and flagging inaccurately translated terms. The full team then discussed and reached consensus on all comments, as well as agreed on tentative simplified Chinese versions of the BITS and BITS-SF.

After reaching a consensus, we sent the simplified Chinese versions to a different professional translator, also contacted through the translation marketplace, for backward translation. This person had English as their native language, as well as specialized in translations from simplified Chinese to English and ICT content. The translator completed backward translations of the measures into English and sent the files to the authors. The full team compared the original and backward-translated BITS and BITS-SF, and the three team members fluent in both English and simplified Chinese also compared the simplified Chinese BITS and BITS-SF to the backward-translated versions. The backward translations were very close to the original English versions. As answering questions about translations was included in the translation service, we contacted the translator who had performed the backward translation with minor questions about wording choice and received feedback. The full team then discussed and reconciled any discrepancies between the original and backward-translated versions, and there was full consensus on the final BITS-SC and BITS-SC-SF.

The BITS-SC used the same six-point Likert scale response options as the English version, which range from 1 (毫无信心/Not at all confident) to 6 (信心十足/Completely confident). Cronbach’s alpha levels were strong: novice α = .87, advanced α = .91, and expert α = .95. Relatedly, the BITS-SF-SC used the same dichotomous response options as the English version: 是/Yes and 否/No.

Computer Self-Efficacy. We used two measures to assess CSE. The first was the simplified Chinese version of the Information Retrieval on the Web Self-Efficacy Scale (Ch-IROWSE; Rodon & Chevalier, 2017). Originally written in French (Rodon & Meyer, 2018), the Ch-IROWSE consists of eight items assessing confidence finding information on the Internet. Respondents use a four-point Likert scale ranging from 1 (否/False) to 4 (是/True). A sample item is, “上网查询时, 如遇到困难, 我相信能找到新的方法继续在网上查找”/“If I come across any difficulty finding information on Internet, I am sure I would be able to carry out a new research strategy” (Rodon & Chevalier, 2017, p. 4). Items were averaged, with higher scores indicating higher self-efficacy. The Ch-IROWSE has shown evidence of a single-factor structure, and the total score has shown evidence of convergent, discriminant, and concurrent validity (Rodon & Chevalier, 2017). Cronbach’s alpha for the current study was .94.

The second measure was a four-item ICT self-efficacy scale written in simplified Chinese (Dong et al., 2020). This measure assesses CSE for general computer tasks. Respondents use a five-point Likert scale ranging from 1 (强烈反对/Strongly disagree) to 5 (非常同意/Strongly agree). A sample item is, “当我在使用信息与通讯技术遇到问题时, 我总能找到几种解决方法”/“When I am confronted with a problem when using ICT, I can usually find several solutions” (Dong et al., 2020, p. 155). Items were averaged, with higher scores indicating higher CSE. Support has been shown for the unidimensional structure, and the measure’s score has shown evidence of concurrent and discriminant validity (Dong et al., 2020). Cronbach’s alpha for the current study was .92.

Self-Esteem. Self-esteem was assessed using the simplified Chinese translation of the Rosenberg Self-Esteem Scale (RSES) – Positive Version (Chen et al., 2015). Originally written in English (Rosenberg, 1965), the RSES is a ten-item measure of individual self-esteem consisting of both positively and negatively worded items. The RSES – Positive Version consists of the same 10 items, except they are all positively worded (Greenberger et al., 2003). Participants respond to items using a four-point Likert scale ranging from 1 (强烈反对/Strongly disagree) to 4 (非常同意/Strongly agree). A sample item is, “总的来说, 我对自己是满意的”/“On the whole, I am satisfied with myself” (Chen et al., 2015, p. 534). Items were averaged, with higher scores indicating higher self-esteem. Support has been found for the unidimensional factor structure, with a test-retest coefficient of .76 (Chen et al., 2015). Cronbach’s alpha for the current study was .94.

Procedure

The study was approved by the Institutional Review Boards (IRBs) at the first two authors’ institutions. Participants were recruited through WeChat, a popular social media platform and messaging app in Mainland China. Data were collected using the Wenjuanxing survey platform, which encrypts data and can be integrated with WeChat; this approach is recommended for collecting data in Mainland China (Mei & Brown, 2018). Interested participants were presented with a QR code. Upon scanning the code, they were sent to the survey, which consisted of a study introduction, demographics form, and the measures of interest. There were no incentives for participating.

Results

BITS-SC Confirmatory Factor Analysis. We conducted confirmatory factor analyses (CFAs) to examine the three solutions tested in the development of the original BITS (Weigold A. & Weigold I.K., 2021). The first was a single-factor model in which all items loaded onto one overall factor. The second was a three-factor uncorrelated model in which each item loaded onto one factor (novice, advanced, or expert); factors were not allowed to covary. The final was a three-factor correlated model that was identical to the second model except that factor covariances were freely estimated. Consistent with the original BITS analyses (Weigold A. & Weigold I.K., 2021), we expected the three-factor correlated model to provide the best fit.

To assess model fit, we used several recommended indices: the robust χ², the Comparative Fit Index (CFI), the Root Mean Square Error of Approximation (RMSEA), and the Standardized Root Mean Square Residual (SRMR). The robust χ² should be nonsignificant, although this is unlikely to occur in practice. CFIs >.900 and >.950 show evidence of adequate and strong fit, respectively. RMSEAs <.100, <.080, and <.060 indicate marginal, adequate, and strong fit, respectively (see Byrne, 2012, for a review). SRMRs <.100 suggest adequate fit (Kline, 2016), and SRMRs <.050 indicate strong fit (Byrne, 2012). We compared the three models using the Akaike Information Criterion (AIC), which is appropriate for both non-nested and nested models. Lower AICs indicate better fit, such that an AIC difference of >10 shows no support for the model with the higher AIC (Burnham & Anderson, 2004).

Models were run using Mplus v. 8.0 (Muthén & Muthén, 1998–2017). The single-factor model showed poor fit: robust χ²₍₁₃₅₎ = 717.58, p < .001; CFI = .727; RMSEA = .144 (90% CI = .134, .155); and SRMR = .146. The three-factor uncorrelated model also evidenced poor fit: robust χ²₍₁₃₅₎ = 630.03, p < .001; CFI = .768; RMSEA = .133 (90% CI = .123, .144); and SRMR = .321. Finally, the three-factor correlated model showed marginal to adequate fit: robust χ²₍₁₃₂₎ = 340.33, p < .001; CFI = .902; RMSEA = .087 (90% CI = .076, .099); and SRMR = .065. The AIC for the three-factor correlated model (11,274.36) was much lower than the AICs for the single-factor (ΔAIC = 568.78) and three-factor uncorrelated (ΔAIC = 322.87) models. Together, these findings suggest that the three-factor correlated model is the best fit for the data. Standardized parameter estimates and descriptive information for this model are shown in Table 1.

Table 1.

Item Loadings and Descriptive Information for the BITS Confirmatory Factor Analyses for Simplified Chinese/Traditional Chinese Translations.

Item	Novice	Advanced	Expert	M (SD) Simplified	M (SD) Traditional
Novice: Operating system	.79 / .77			5.53 (1.11)	5.62 (0.77)
Novice: Software	.73 / .71			5.18 (1.49)	5.68 (0.72)
Novice: Hardware	.76 / .69			5.44 (1.26)	5.62 (0.86)
Novice: Internet	.75 / .84			5.64 (0.93)	5.75 (0.55)
Novice: Networking	.74 / .68			5.24 (1.38)	5.54 (0.80)
Novice: Troubleshooting	.67 / .73			5.69 (0.94)	5.71 (0.77)
Advanced: Internet		.71 / .66		3.80 (1.90)	3.44 (1.57)
Advanced: Software		.78 / .69		3.89 (1.75)	4.15 (1.41)
Advanced: Networking		.70 / .69		4.36 (1.79)	3.33 (1.68)
Advanced: Operating system		.77 / .66		4.40 (1.78)	4.31 (1.48)
Advanced: Hardware		.89 / .78		3.14 (2.05)	3.03 (1.81)
Advanced: Troubleshooting		.89 / .81		3.09 (1.83)	3.02 (1.43)
Expert: Software			.87 / .86	2.62 (1.84)	2.00 (1.41)
Expert: Troubleshooting			.86 / .91	2.70 (1.84)	1.91 (1.32)
Expert: Operating system			.84 / .67	2.95 (1.85)	2.67 (1.65)
Expert: Internet			.90 / .77	2.63 (1.76)	2.27 (1.45)
Expert: Hardware			.84 / .52	2.76 (1.81)	3.06 (1.70)
Expert: Networking			.86 / .76	2.29 (1.68)	1.88 (1.24)
Factor Intercorrelations
F1	—			5.45 (0.94)	5.65 (0.58)
F2	.52 / .44	—		3.78 (1.54)	3.55 (1.20)
F3	.29 / .19	.82 / .71	—	2.66 (1.59)	2.30 (1.16)
Cronbach’s α	.87 / .87	.91 / .86	.95 / .87
McDonald’s ω	.88 / .87	.91 / .86	.95 / .86

Note. Means were calculated by summing the items on each factor and dividing them by the number of items. Factor intercorrelations were calculated using Pearson’s r.

BITS-SC and BITS-SF-SC Validity Evidence. We next examined aspects of convergent, discriminant, and concurrent validity. Regarding convergent validity, we correlated the BITS-SC subscale and BITS-SF-SC total scores with measures of Internet information retrieval self-efficacy and ICT self-efficacy. As expected, all correlations were positive, moderate or large, and significant at p < .001 (r range = .38 for novice with Internet information retrieval self-efficacy to .74 for advanced with ICT self-efficacy). For discriminant validity, we correlated the BITS-SC subscale and BITS-SF-SC total scores with self-esteem. All correlations were positive, small or moderate, and significant at p < .001 (r range = .25 for novice to .38 for advanced). Although significant, these correlations were similar to those for self-esteem and both Internet information retrieval self-efficacy and ICT self-efficacy. All correlations are found in Table 2.

Table 2.

Correlations of the BITS Subscale and BITS-SF Total Scores with Measures Used to Assess Aspects of Validity.

Simplified Chinese
Measure	Novice	Advanced	Expert	BITS-SF	Inf Ret SE	CSE
Inf Ret SE	.38**	.58**	.43**	.55**	—
CSE	.48**	.74**	.63**	.63**	.59**	—
Self-esteem	.25**	.38**	.33**	.32**	.37**	.34**

Traditional Chinese
Measure	Novice	Advanced	Expert	BITS-SF	Comp SE	ICT SE
CSE	.49**	.53**	.36**	.31**	—
ICT SE	.40**	.48**	.35**	.32**	.51**	—
Self-esteem	.18**	.21**	.13*	.05	.13*	.21**

Note. Inf Ret SE = information retrieval on the Web self-efficacy, CSE = computer self-efficacy, ICT SE = information and computer technologies self-efficacy.

^*p ≤ .05, ^**p ≤ .01.

Next, we examined two aspects of concurrent validity: self-identified computer skill level and current occupation. For computer skill level, participants self-identified as novice (n = 94, 45.4%), novice-to-advanced (n = 65, 31.4%), advanced (n = 34, 16.4%), advanced-to-expert (n = 11, 5.3%), and expert (n = 3, 1.4%); the latter two categories were combined into one. We then compared the four groups on the three BITS-SC subscale scores using MANOVA and the BITS-SF-SC total score using ANOVA. The BITS-SC MANOVA and its corresponding univariate ANOVAs were all significant at p < .001, as was the BITS-SF-SC ANOVA. In most cases, scores on the BITS-SC subscales and BITS-SF-SC increased significantly as self-identified computer skill increased. Detailed results are shown in Table 3.

Table 3.

BITS and BITS-SF Scores for Computer Skill Level and Occupation.

Simplified Chinese
Comparison	df	F	p	η²_P	M (SD)	M (SD)	M (SD)	M (SD)
Computer skill level					Novice	Novice/Adv	Adv	Adv/Exp
BITS MANOVA	9, 609	12.01	<.001	.151
BITS novice	3, 203	9.19	<.001	.120	5.12 (1.06)^ab	5.67 (0.77)^a	5.93 (0.29)^b	5.57 (1.02)
BITS advanced	3, 203	43.83	<.001	.393	2.85 (1.34)^abc	4.03 (1.17)^abc	5.26 (0.91)^ab	5.31 (1.11)^ac
BITS expert	3, 203	24.11	<.001	.263	1.98 (1.30)^abc	2.64 (1.44)^abc	3.83 (1.49)^ab	4.48 (1.33)^ac
BITS-SF ANOVA	3, 203	33.38	<.001	.330	3.17 (1.28)^abc	4.25 (1.17)^ab	5.24 (0.96)^ab	5.36 (1.39)^c
Occupation					Teacher	Corporate	IT engineer	Public/farm/worker
BITS MANOVA	9, 525	5.33	<.001	.084
BITS novice	3, 175	1.94	.124	.032	5.39 (1.00)^a	5.64 (0.79)	5.85 (0.61)^a	5.27 (0.94)
BITS advanced	3, 175	9.81	<.001	.144	3.55 (1.59)^a	3.70 (1.34)^b	5.53 (0.77)^abc	3.85 (1.41)^c
BITS expert	3, 175	15.34	<.001	.208	2.47 (1.51)^a	2.11 (1.33)^b	4.78 (0.98)^abc	2.96 (1.74)^c
BITS-SF ANOVA	3, 175	7.49	<.001	.114	3.79 (1.49)^a	4.04 (1.29)^b	5.47 (1.02)^abc	4.14 (1.59)^c

Traditional Chinese
Comparison	df	F	p	η²_P	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Computer skill level					Novice	Novice/Adv	Adv	Adv/Exp
BITS MANOVA	9, 804	17.39	<.001	.163
BITS novice	3, 268	15.41	<.001	.147	5.17 (0.97)^abc	5.71 (0.45)^ab	5.80 (0.34)^ac	6.00 (0.00)^abc	—
BITS advanced	3, 268	51.51	<.001	.366	2.38 (0.86)^a	3.41 (1.02)^a	4.32 (0.91)^a	5.50 (0.80)^a	—
BITS expert	3, 268	37.68	<.001	.297	1.63 (0.63)^a	2.08 (0.96)^a	2.84 (1.13)^a	4.65 (1.27)^a	—
BITS-SF ANOVA	3, 268	16.50	<.001	.156	3.27 (1.17)^a	3.91 (1.12)^a	4.45 (1.04)^a	5.45 (1.20)^a	—
Occupation					Student	Service worker	Teacher	Public servant	IT engineer
BITS MANOVA	12, 588	3.48	<.001	.066
BITS novice	4, 196	1.56	.186	.031	5.71 (0.46)	5.57 (0.63)	5.76 (0.48)	5.54 (0.66)	5.89 (0.30)
BITS advanced	4, 196	7.89	<.001	.139	3.44 (1.06)^a	3.37 (0.93)^b	3.84 (1.29)^c	3.40 (1.17)^d	5.24 (0.87)^abcd
BITS expert	4, 196	7.75	<.001	.137	2.29 (1.03)^a	2.21 (0.97)^b	2.35 (1.24)^c	2.10 (1.16)^d	4.14 (1.46)^abcd
BITS-SF ANOVA	4, 196	5.18	<.001	.096	3.89 (1.29)^a	4.03 (1.01)^b	4.21 (1.19)^c	3.82 (1.10)^d	5.55 (0.82)^abcd

Note. Adv/Exp = advanced-to-expert, Novice/Adv = novice-to-advanced, Adv = advanced, Adv/Exp = advanced-to-expert (also includes expert for the simplified Chinese sample), Corporate = corporate employees, Public/farm/worker = combined group of public servants, farmers, and workers. Means in each row with identical subscripts differ significantly at p ≤ .05; although subscripts are reported for the occupation BITS novice subscales, they should not be interpreted, as the univariate ANOVA was not significant.

For occupation, we compared teachers (n = 111, 53.6%), corporate employees (n = 28, 13.5%), IT engineers (n = 19, 9.2%), and a combined group of public servants, workers, and farmers (n = 21, 10.2%). We ran one MANOVA on the BITS-SC subscale scores and one ANOVA on the BITS-SF-SC total score. All analyses were significant at p < .001, except the follow-up novice ANOVA (p = .124). For the advanced and expert subscales of the BITS-SC, as well as the BITS-SF-SC, IT engineers scored significantly higher than teachers, corporate employees, and public servants/workers/farmers. Detailed results are shown in Table 3. These findings support the convergent, discriminant, and concurrent validity of the BITS-SC’s subscale and BITS-SF-SC’s total scores.

BITS-SF-SC Latent Class Analysis. We used latent class analysis (LCA) to determine the number of underlying classes of the BITS-SF-SC. We expected analyses to show evidence of either three or four latent classes. We examined a series of one, two, three, four, and five possible classes. To compare classes, we primarily used the recommended bootstrap likelihood ratio test (BLRT), which estimates if a certain number of classes provide better fit than one fewer. Additionally, we reported the AIC and adjusted Bayesian Information Criterion (BIC_adj); in both cases, smaller numbers indicate better fit (Nylund et al., 2007).

Class comparisons yielded a best-fitting four-class solution. Four classes provided significantly better fit than three (BLRT p = .003) and did not show significantly worse fit than five (BLRT p = 1.000). Additionally, both the four-factor solution’s AIC (967.04) and BIC_adj (971.48) were lower than those for the other solutions, with the next lowest being for the three-factor solution (AIC = 973.09; BIC_adj = 976.37). The four underlying classes corresponded to self-efficacy for novice (n = 76, 36.7%), advanced (n = 62, 30.0%), and expert (n = 64, 30.9%) levels, as well as negligible self-efficacy (n = 5, 2.4%). Participants in the negligible class had a low probability of responding to any item with Yes (p range = .000 to .256). Those in the novice class had a high probability of responding to both novice items affirmatively (p = 1.000 for both), a low-to-moderate probability of doing so for the advanced items (p = .202 and .375), and a low probability of doing so for the expert items (p = .020 and .099). Participants in the advanced class had a high probability of responding Yes to the novice (p = .948 and 1.000) and advanced (p = .975 and 1.000) items, as well as a low probability of doing so for the expert items (p = .000 and .282). Finally, those in the expert class had a high probability of responding Yes to the novice and advanced items (p = 1.000 for all) and a moderate-to-high probability of doing so for the expert items (p = .717 and .958).

There were different ranges of scores within each class: negligible 0 to 1, novice 2 to 4, advanced 3 to 5, and expert 5 to 6. However, only six (7.9%) of the novice users scored a 4 and two (3.2%) of the advanced users scored a 3; consequently, ranges for the classes were negligible 0 to 1, novice 2 to 3, advanced 4 to 5, and expert 5 to 6, with a score of 5 representing self-efficacy for advanced-to-expert levels.

We next compared the four classes on the BITS-SC subscales and the convergent validity self-efficacy measures using two MANOVAs. All main and univariate follow-up analyses were significant at p < .001. Regarding pairwise comparisons, mean scores on the outcome variables mostly significantly increased as latent class increased from negligible to expert. Results of these analyses are shown in Table 4. Together, these findings provide convergent validity evidence for the four underlying classes of the BITS-SF-SC.

Table 4.

Mean Differences Across BITS-SF Classes for Related Variables.

Simplified Chinese
Comparison	df	F	p	η²_P	Negligible M (SD)	Novice M (SD)	Advanced M (SD)	Expert M (SD)
BITS MANOVA	9, 609	29.62	<.001	.304
BITS novice	3, 203	19.97	<.001	.228	3.33 (1.54)	5.10 (1.08)^ab	5.74 (0.57)^a	5.72 (0.65)^b
BITS advanced	3, 203	76.25	<.001	.530	2.23 (1.85)	2.41 (0.97)^ab	4.17 (1.00)^ac	5.04 (1.16)^bc
BITS expert	3, 203	56.25	<.001	.454	2.37 (2.18)	1.63 (0.81)^ab	2.31 (1.18)^ac	4.20 (1.43)^bc
Efficacy MANOVA	6, 406	18.88	<.001	.218
Inf Retr SE	3, 203	38.66	<.001	.364	1.55 (0.87)^ab	2.84 (1.06)^cd	3.83 (0.90)^ace	4.37 (0.86)^bde
CSE	3, 203	29.37	<.001	.303	1.68 (0.62)^abc	2.90 (0.69)^abc	3.40 (0.53)^ab	3.62 (0.54)^ac

Traditional Chinese
Comparison	df	F	p	η²_P	Novice M (SD)	Advanced M (SD)	Expert M (SD)
BITS MANOVA	6, 538	25.19	<.001	.219
BITS novice	2, 270	12.15	<.001	.083	5.15 (0.93)^ab	5.69 (0.53)^a	5.79 (0.33)^b
BITS advanced	2, 270	40.96	<.001	.233	2.35 (0.98)^a	3.47 (1.05)^a	4.70 (1.15)^a
BITS expert	2, 270	64.03	<.001	.322	1.78 (0.94 )^a	2.06 (0.88)^b	3.87 (1.29)^ab
Efficacy MANOVA	4, 536	7.20	<.001	.051
Computer SE	2, 268	12.05	<.001	.083	6.54 (1.96)^a	7.54 (1.41)^b	8.35 (1.30)^ab
ICT SE	2, 268	9.76	<.001	.068	4.16 (1.91)^a	4.89 (1.56)^b	5.83 (1.20)^ab

Note. Inf Retr SE = Internet information retrieval self-efficacy, Computer SE = computer self-efficacy, ICT SE = ICT self-efficacy. Means in each row with identical subscripts differ significantly at p ≤ .05.

Discussion

This study examined the psychometric properties of the translated simplified Chinese versions of the BITS and BITS-SF. The BITS-SC showed evidence of a three-factor correlated structure (novice, advanced, and expert) that mirrored the results of the original BITS (Weigold A. & Weigold I. K., 2021). The BITS-SF-SC evidenced four underlying classes (negligible, novice, advanced, and expert self-efficacy). Although the original BITS-SF only showed the latter three classes (Weigold A. & Weigold I. K., 2021), the addition of a negligible category is theoretically plausible. One reason this category may have emerged is that participants primarily accessed the study using a QR code, therefore likely completing the study on a smartphone. Conversely, the original BITS study recruited participants using a Web link (Weigold A. & Weigold I. K., 2021). It should be noted, however, that there were only five participants showing negligible self-efficacy.

We also found good evidence of convergent, discriminant, and concurrent validity. Both the three BITS-SC subscale and BITS-SF-SC total scores correlated as expected with Internet information retrieval self-efficacy and ICT self-efficacy. These scales also correlated significantly with self-esteem; however, the correlations were smaller and similar to those of the convergent validity scales with self-esteem. Finally, participants with higher self-identified computer skill levels had higher BITS-SC subscale and BITS-SF-SC total scores than those with lower computer skill levels, as did IT engineers compared to occupations in which strong computer skills are not as critical. The only exception was that the novice subscale scores were high across occupations and did not significantly differ across groups, indicating that self-efficacy for novice computer skills is generally high for those in the occupations assessed. These findings all align with those of the original BITS and BITS-SF (Weigold A. & Weigold I. K., 2021).

Taken together, our results suggest that the BITS-SC and BITS-SF-SC are viable for use with adults in Mainland China. Given the country’s strong focus on ICT infrastructure, education, and teacher training (Bajpai et al., 2019; Zeng, 2022), these measures provide a useful tool for assessing CSE at several levels to assist in identifying training needs, changes in confidence as a result of trainings, and differences associated with the ICT gap between urban and rural systems.

Study 2: Traditional Chinese

Method

Participants

Participants were 273 adults (20 years and older) located in Taiwan. They identified primarily as women (n = 185, 67.8%), followed by men (n = 84, 30.8%) and identity not listed (n = 3, 1.1%). Their ages ranged from 20 to 67 (M = 30.01, SD = 10.57). A third of participants reported that they were students (n = 100, 36.6%), followed by service workers (n = 35, 12.8%), teachers (n = 33, 12.1%), public servants (n = 22, 8.1%), IT engineers (n = 11, 4.0%), health professionals (n = 10, 3.7%), unemployed (n = 8, 2.9%), and other (n = 54, 19.8%).

Measures

Brief Inventory of Technology Self-Efficacy – Traditional Chinese. We used the same translation procedures to develop the traditional Chinese versions of the BITS (BITS-TC) and BITS-SF (BITS-SF-TC) as we used to develop their simplified Chinese counterparts. For the forward translations, we employed a professional translator from the same marketplace who was a native writer of traditional Chinese and specialized in both English to traditional Chinese translations and ICT content. Upon receiving the translations, the same three team members as previously (who all were fluent in both traditional Chinese and English) separately reviewed the English and traditional Chinese versions, then sent their comments to the full team. Through discussion, the team reached consensus regarding all changes. After reaching a consensus, we sent the traditional Chinese versions to a different professional translator for backward translation who had English as their native language and specialized in both traditional Chinese to English translations and ICT content. Similar to the simplified Chinese versions, the backward translations were very close to the original English versions. The five team members discussed and reconciled any discrepancies after eliciting feedback from the professional translator.

The BITS-TC again used the six-point Likert scale response options, which ranged from 1 (毫無信心/Not at all confident) to 6 (信心十足/Completely confident). Cronbach’s alpha levels were all good: Novice α = .87, advanced α = .86, and expert α = .87. The BITS-SF-TC again used the dichotomous response options of 是/Yes and 否/No.

Computer Self-Efficacy. We used two measures to assess aspects of CSE. The first was the traditional Chinese translation (Pu & Hung, 2011) of the Computer Self-Efficacy Measure originally written in English by Compeau and Higgins (1995). Respondents answer 10 questions regarding their self-efficacy using an unnamed computer software with which they are unfamiliar. Participants provide two answers. The first is dichotomous (是/Yes or 否/No), whereas the second uses a ten-point Likert-type scale with responses ranging from 1 (完全沒有信心/Not at all confident) to 10 (完全有信心/Totally confident). We only used the Likert-type scale responses for this study. A sample item is, “如果有人指導我使用電腦, 我就能夠使用它”/“I could complete the job using the software package if someone else had helped me get started” (Pu & Hung, 2011, p. 10). Items were averaged, with higher numbers indicating higher CSE. This measure’s scores have shown adequate evidence of content, criterion, and construct validity (Pu & Hung, 2011). Cronbach’s alpha for the current study was .93.

The second measure was a 3-item scale written in traditional Chinese to assess ICT self-efficacy (Hsu et al., 2014). Participants respond to items using a seven-point Likert scale ranging from 1 (非常不同意/Strongly disagree) to 7 (非常同意/Strongly agree). A sample item is “我自己能夠順暢地使用ICT來進行工作”/“I can smoothly use ICT accomplishing works by myself” (Hsu et al., 2014, pp. 75 and 115). Items were averaged, with higher scores indicating higher ICT self-efficacy. The three items loaded together onto one factor when examined as part of a larger set of items. This factor has shown evidence of convergent validity (Hsu et al., 2014). Cronbach’s alpha for the current study was .96.

Self-Esteem. Similar to Study 1, self-esteem was measured using the traditional Chinese translation (Pai, 2015) of the ten-item Rosenberg Self-Esteem Scale (Rosenberg, 1965). Items were reverse coded as appropriate and averaged, with higher scores indicating higher self-esteem. This measure’s scores have shown evidence of concurrent validity (Ling & Huang, 2010). Cronbach’s alpha for the current study was .89.

Procedures

The study was approved by the IRBs at the first two authors’ institutions, and National Taiwan University accepted the United States IRB approval. Participants were primarily recruited through the National Taiwan Normal University’s Dcard, an anonymous online communication platform open to those with an email account associated with the university. Data were also collected using social media, such as through Facebook groups for graduate students completing survey research in Taiwan. Participants clicked on a Web link, which allowed them to access the survey introduction, demographics form, and measures. Upon completion, participants had the option of being included in a raffle for one of six $500 New Taiwan dollars (approximately $17.21 USD) gift cards to 7–11 in Taiwan.

Results

BITS-TC Confirmatory Factor Analysis. We examined the factor structure of the BITS-TC using the same models as in Study 1: a single-factor model, an uncorrelated three-factor model, and a correlated three-factor model. We again used the robust χ², CFI, RMSEA, and SRMR to determine model fit, as well as the AIC to compare models.

Models showed a similar pattern of fit to those in Study 1. The single-factor model was a poor fit: robust χ²₍₁₃₅₎ = 1005.06, p < .001; CFI = .625; RMSEA = .154 (90% CI = .145 to .163); and SRMR = .148. The three-factor uncorrelated model also evidenced poor fit: robust χ²₍₁₃₅₎ = 601.26, p < .001; CFI = .799; RMSEA = .112 (90% CI = .103 to .122); and SRMR = .235. Finally, the three-factor correlated model showed generally adequate fit: robust χ²₍₁₃₂₎ = 327.00, p < .001; CFI = .916; RMSEA = .074 (90% CI = .064 to .084); and SRMR = .057. The three-factor correlated model had a meaningfully lower AIC (13,149.35) than both the single-factor (ΔAIC = 811.39) and three-factor uncorrelated (ΔAIC = 244.61) models. Taken together, the three-factor correlated model appeared to be the best fit. Standardized parameter estimates and descriptive information for this model are shown in Table 1.

BITS-TC and BITS-SF-TC Validity Analyses. We examined aspects of convergent, discriminant, and concurrent validity. First, to assess convergent validity, we correlated the BITS-TC subscale and BITS-SF-TC total scores with measures of CSE and ICT self-efficacy. As expected, all correlations were positive, moderate, and significant at p < .01 (r range = .35 for expert and ICT self-efficacy to .53 for advanced and CSE). We assessed discriminant validity by correlating the BITS-TC subscale and BITS-SF-TC total scores with a measure of self-esteem. Correlations were positive and small, ranging from not significant at p < .05 to significant at p < .01 (r range = .05 for expert and .21 for advanced). All correlations are shown in Table 2.

We next assessed the two aspects of concurrent validity examined in Study 1: self-identified computer skill level and occupation. For computer skill level, participants self-identified as novice (n = 44, 16.1%), novice-to-advanced (n = 153, 56.0%), advanced (n = 64, 23.4%), and advanced-to-expert (n = 11, 4.0%); none self-identified as expert. We compared these groups on the BITS-TC subscale scores using MANOVA and the BITS-SF-TC total score using ANOVA. All main analyses were significant at p < .001. Regarding post hoc group analyses, participants self-identifying at lower computer skill levels mostly scored significantly lower on the BITS-TC and BITS-SF-TC than those self-identifying at higher levels. Detailed results are found in Table 3.

We next examined whether scores on the BITS-TC and BITS-SF-TC differed by occupation. As there were not enough participants in each occupation, we only included students (n = 100, 36.6%), service workers (n = 35, 12.8%), teachers (n = 33, 12.1%), public servants (n = 22, 8.1%), and IT engineers (n = 11, 4.0). Most main analyses were significant at p < .001, except for the novice subscale univariate ANOVA (p = .186). For the significant ANOVAs, the IT engineers had significantly higher scores than all other groups. Detailed results are shown in Table 3. Taken together, these findings show support for the convergent, discriminant, and concurrent validity of the BITS-TC’s subscale and BITS-SF-TC’s total scores.

BITS-SF-TC Latent Class Analysis. We conducted LCA to determine the number of underlying classes of the BITS-SF-TC. As in Study 1, we expected to find evidence for either three latent classes (self-efficacy for novice, advanced, and expert computer skills) or four (the addition of negligible self-efficacy). We examined one, two, three, and four possible latent classes, comparing them using the BLRT, AIC, and BIC_adj (Nylund et al., 2007).

Comparisons yielded a best-fitting three-class underlying solution. Three classes were a significantly better fit than two (BLRT p < .001) and not a significantly worse fit than four (BLRT p = .500). Additionally, the three-class AIC (1,205.11) and BIC_adj (1213.88) were both lower than those of the other solutions, with the next-lowest being the four-class AIC (1,211.47) and two-class BIC_adj (1,217.31). The three classes corresponded to self-efficacy for novice (n = 27, 9.9%), advanced (n = 206, 75.5%), and expert (n = 40, 14.7%) computer skills. Those in the novice class had a high probability of responding Yes to both novice items (p = .868 and .939), a low-to-moderate probability of doing so for the advanced items (p = .000 and .372), and a low probability of doing so for the expert items (p = .111 and .123). Participants in the advanced class had a high probability of replying affirmatively to the novice items (p = 1.000 and .991), a moderate-to-high probability of doing so for the advanced items (p = .414 and .970), and a low probability of doing so for the expert items (p = .093 and .296). Finally, those in the expert class had a high probability of replying Yes to the novice (p = 1.000 for both) and advanced (p = .999 and 1.000) items, and a moderate-to-high probability of doing so for the expert items (p = .687 and .995).

There were different score ranges in each class: novice 0 to 4, advanced 3 to 5, and expert 6. However, only one person in the novice class (4.0%) had a score of 4, indicating the actual ranges were novice 0 to 3, advanced 3 to 5, and expert 6; scores of 3 corresponded to self-efficacy for novice-to-advanced computer skills. Only one participant had a score of 0.

Finally, we used two MANOVAs to compare the three underlying classes on the three BITS-TC subscales and the CSE and ICT self-efficacy measures. All multivariate and follow-up univariate analyses were significant at p < .001. Group means generally significantly differed, with those in the novice class scoring the lowest and those in the expert class scoring the highest. Detailed results of these analyses are shown in Table 4. Taken together, these findings provide evidence of convergent validity for the three underlying classes of the BITS-SF-TC.

Discussion

This study examined the psychometric properties of the traditional Chinese versions of the BITS and BITS-SF. As expected, the three-factor correlated model (novice, advanced, and expert) was the best-fitting structure for the BITS, compared to single-factor and three-factor uncorrelated models. Relatedly, the LCA of the BITS-SF-TC showed evidence of three underlying classes corresponding to self-efficacy for novice, advanced, and expert computer skills. Both results replicate those found for the original English versions of the BITS and BITS-SF (Weigold A. & Weigold I. K., 2021).

The results also showed good evidence for aspects of convergent, discriminant, and concurrent validity for the BITS-TC subscale and BITS-SF-TC total scores. All scores had positive, moderate correlations with measures of CSE and ICT self-efficacy, as well as small correlations with self-esteem. These scores also significantly differed across self-identified computer skill levels such that those with higher skill levels generally showed higher levels of CSE. Additionally, IT engineers had significantly higher self-efficacy for advanced and expert computer skills, as well as for the BITS-SF-TC, compared to students, service workers, teachers, and public servants. However, the differences across occupations in self-efficacy for novice computer skills was not significant, suggesting that people in these fields generally felt confident using computers at a basic level. These results were nearly identical to those found for the original English versions of the BITS and BITS-SF (Weigold A. & Weigold I. K., 2021).

Taken together, the findings indicate that the BITS-TC and BITS-SF-TC are viable measures of CSE in Taiwan. Taiwan has a strong focus on artificial intelligence, ICT, and ICT education (Haldorai et al., 2020), and tailoring related instructional interventions based on self-efficacy can be useful (see Tsai, 2019). The BITS-TC and BITS-SF-TC can be used to assist in adapting technology trainings, understanding their outcomes, and refining them for different sectors, such as education and business.

General Discussion

We translated the BITS and BITS-SF into simplified and traditional Chinese, as well as conducted two studies examining their psychometric properties. The results replicate and extend previous findings related to the BITS and BITS-SF. Both translated measures evidenced a three-factor correlated structure corresponding to the novice, advanced, and expert subscales; the subscale scores also related as expected to other measures of CSE, a measure of self-esteem, self-identified computer skill level, and occupation. Regarding the BITS-SF, the simplified Chinese version yielded four underlying classes of CSE (negligible, novice, advanced, and expert), whereas the traditional Chinese version showed three classes (novice, advanced, and expert). Both are consistent with underlying theory, the original development of the BITS-SF, and BITS-SF scoring (Dreyfus, 2004; Weigold, 2021; Weigold A. & Weigold I. K., 2021); additionally, in both studies, the underlying classes showed evidence of convergent validity.

It is possible that the different number of classes found for the BITS-SF-SC and the BITS-SF-TC are due to how data were collected. Since participants accessed the simplified Chinese study using a QR code, it is likely that most participants completing it used a smartphone and did not need basic computer skills. Conversely, the traditional Chinese study was accessed using a Web link. There are over one billion Internet users in Mainland China; 99.6% of users access the Internet through smartphones, with fewer accessing it through desktop computers (33.3%), laptops (32.6%), TVs (32.6%), and tablets (27.6%; China Internet Network Information Center, 2022). However, only six participants across both studies had a score of 0. It could be that people who have negligible self-efficacy are not likely to complete studies using technology. Alternatively, there may be few people with negligible self-efficacy in populations with easy access to computers and the Internet. If so, the results of the current studies indicate the BITS-SF can capture this in its scoring.

Implications

Given the focus on e-learning and digital technology in Mainland China and Taiwan (Bajpai et al., 2019; Hsin-fang & Tzu-hsuan, 2022), the translated BITS and BITS-SF can be used to assess which individuals are most likely to benefit from different types of technology training. For example, the three subscales of the BITS allow administrators and instructors to determine the degree of self-efficacy for different computer skill levels that employees and students might have. Additionally, the BITS-SF can be used as a CSE screening tool with large samples, such as an entire school (Weigold A. & Weigold I. K., 2021). Given that smartphones are used more frequently than computers in Mainland China to access the Internet (China Internet Network Information Center, 2022), the BITS and BITS-SF may be especially relevant in settings where computers are used, such as business and education, in this country.

Relatedly, both Mainland China and Taiwan are invested in decreasing the urban and rural educational divide through technology (Bajpai et al., 2019; Hsin-fang & Tzu-hsuan, 2022). Rural students have reported lower scores for digital self-efficacy in Mainland China and online learning self-efficacy in Taiwan compared to their urban counterparts (Li et al., 2022; Liao et al., 2016). As a result, CSE may be an important construct to consider when implementing and assessing technological initiatives. For example, the translated BITS may provide valuable information not only about major changes in CSE (e.g., moving from novice to advanced), but also about minor increases within the different subscales as individuals learn new techniques (e.g., lower scores within the advanced subscale to higher scores within the same subscale), which can then be compared across urban and rural areas. Additionally, the translated BITS-SF might be used with large samples of trainees or educators to provide population estimates of CSE across time.

The findings of the current study also have implications for future research on the BITS and BITS-SF. For example, researchers should examine potential within-country differences, such as across Mainland Chinese provinces (Rodon & Chevalier, 2017) and racial and ethnic groups living in both countries. Additionally, studies examining the BITS and BITS-SF in samples using traditional data collection methods (e.g., paper-and-pencil) would allow for a stronger estimation of the people who have CSE for different computer skill levels. The findings of the current and original studies indicate that the expert scale does not have a ceiling effect, even with IT engineers; however, it is unclear if there would be floor effects for the novice subscale in people who seldom use computers.

Relatedly, the findings of the LCAs have implications for the BITS-SF’s scoring. When combining the patterns of scores of the LCAs in the current and original studies, there is support for a score of 1 indicating negligible-to-novice CSE, rather than novice CSE. This would be consistent with the scores of 3 and 5 indicating CSE between two skill levels (Weigold, 2021). Further studies, particularly with samples expected to be low in CSE, are needed to determine the most accurate interpretation of BITS-SF low scores.

Finally, since both Mainland China and Taiwan invest in technology for education (Hsin-fang & Tzu-hsuan, 2022; Zeng, 2022), it would be helpful to examine the psychometric properties of the translated BITS and BITS-SF for use with students at different ages and levels of school so that they can potentially be used with younger samples. Relatedly, although the samples in the current study had large numbers of teachers (Study 1) and students (Study 2), the translated measures should be assessed in multiple samples of teachers and students to provide normative data for these populations.

Limitations

The results of the current studies should be understood in the context of their limitations. First, all data were collected cross-sectionally, so conclusions cannot be drawn about the longitudinal stability of the translated BITS and BITS-SF. Second, all measures were self-report surveys, which could have led to mono-method bias. Third, participants were recruited online and required to use electronic devices to complete the study. This might have resulted in biased sampling, as those without access to technology or the recruiting methods were not included. Relatedly, we only recruited one sample per study, which did not allow for an internal replication of the findings.

Conclusion

We conducted two studies assessing the psychometric properties of the simplified and traditional Chinese BITS and BITS-SF. Both measures’ scores showed evidence of a three-factor structure (BITS); underlying classes corresponding to relevant levels of CSE (BITS-SF); internal consistency; and convergent, discriminant, and concurrent validity. These replicate and extend the findings of the original English BITS and BITS-SF, indicating that they are strong measures for assessing CSE in both simplified and traditional Chinese. As the BITS and BITS-SF are modern questionnaires that correct for issues in previous CSE measures (Weigold A. & Weigold I. K., 2021), the current studies provide a strong foundation for their use in Mainland China and Taiwan. Both the BITS and BITS-SF measures are available for download at www.bitssurvey.com and free for non-commercial use.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Data Availability

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

Author Biographies

Arne Weigold, PhD, is a full professor in the Department of Psychological Sciences at Kent State University. His research interests are in scale development and translation, human-computer interaction, and research methodology in psychology.

Ingrid K. Weigold, PhD, is a full professor in the Counseling Psychology Program in the Department of Psychology at The University of Akron. Her research emphases include international student well-being, scale development, human agency, and data collection methodology.

Xiangling Zhang, PhD, is a lecturer in the School of Information Technology and Labor Education at Beijing Institute of Education. Her research areas include Educational Technology, AI Education, and Open Educational Resources.

Ning Tang, MA, is a doctoral candidate in the Counseling Psychology Program in the Department of Psychology at The University of Akron. Her research interests are psychology trainees' self-care, international students' well-being, sexual violence, gendered racial microaggressions, and post-traumatic change.

Yun Kai Chong, MA, is a doctoral student in the Counseling Psychology Program in the Department of Psychology at The University of Akron. His main research interests involve minority and international student mental health, meaningfulness, and personal growth.

References

Bajpai

Biberman

Y. Y.

(2019). ICT for education: Lessons from China. CDS Working Paper Series: Towards a new Indian model of information and communications technology-led growth and development. Columbia University.

Bandura

(1982). Self-efficacy mechanism in human agency. American Psychologist, 37(2), 122–147. https://doi.org/10.1037/0003-066X.37.2.122

Bandura

(2006). Guide for constructing self-efficacy scales. In Pajares

Urdan

T. S.

(Eds.), Self-efficacy beliefs of adolescents (pp. 307–337). Age Information Publishing.

Behr

Braun

(2022). How does back translation fare against team translation? An experimental case study in the language combination English–German. Journal of Survey Statistics and Methodology, 11(2), 285–315. smac005. https://doi.org/10.1093/jssam/smac005

Burnham

K. P.

Anderson

D. R.

(2004). Multimodel inference: Understanding AIC and BIC in model selection. Sociological Methods & Research, 33(2), 261–304. https://doi.org/10.1177/0049124104268644

Byrne

B. M.

(2012). Structural equation modeling with Mplus. Routledge.

Cassidy

Eachus

(2002). Developing the Computer User Self-Efficacy (CUSE) Scale: Investigating the relationship between computer self-efficacy, gender and experience with computers. Journal of Educational Computing Research, 26(2), 133–153. https://doi.org/10.2190/JGJR-0KVL-HRF7-GCNV

Chang

Wang

C.-Y.

Chen

G.-D.

(2009). National program for e-learning in Taiwan. Educational Technology & Society, 12(1), 5–17.

Chen

Han

(2015). The reliability and validity of the Chinese version of the revised-positive version of Rosenberg self-esteem scale. Advances in Psychology, 05(09), 531–535. https://doi.org/10.12677/ap.2015.59068

10.

Chibisa

Tshabalala

M. G.

Maphalala

M. C.

(2021). Pre-service teachers’ computer self-efficacy and the use of computers. International Journal of Learning, Teaching and Educational Research, 20(11), 325–345. https://doi.org/10.26803/ijlter.20.11.18

11.

China Internet Network Information Center . (2022, August 31). The 50th “Statistical Report on Internet Development in China”. China Internet Network Information Center. https://www.cnnic.net.cn/n4/2022/0914/c88-10226.html

12.

Compeau

D. R.

Higgins

C. A.

(1995). Computer self-efficacy: Development of a measure and initial test. MIS Quarterly, 19(2), 189–211. https://doi.org/10.2307/249688

13.

Cyberspace Administration of China . (2021). 14th Five-Year Plan for National Informatization. Cyberspace Administration of China. http://www.cac.gov.cn/2021-12/27/c_1642205314518676.htm

14.

Dong

Chai

C. S.

Zhai

(2020). Exploring the structural relationship among teachers’ technostress, technological pedagogical content knowledge (TPACK), computer self-efficacy and school support. The Asia-Pacific Education Researcher, 29(2), 147–157. https://doi.org/10.1007/s40299-019-00461-5

15.

Dreyfus

S. E.

(2004). The five-stage model of adult skill acquisition. Bulletin of Science, Technology & Society, 24(3), 177–181. https://doi.org/10.1177/0270467604264992

16.

Ferdousi

(2019). The effect of computer self-efficacy and attitude on undergraduate students’ intention to use emerging technology in classroom learning. Journal of Computer Sciences and Applications, 7(1), 50–55. https://doi.org/10.12691/jcsa-7-1-8

17.

Greenberger

Chen

Dmitrieva

Farruggia

S. P.

(2003). Item-wording and the dimensionality of the Rosenberg self-esteem scale: Do they matter? Personality and Individual Differences, 35(6), 1241–1254. https://doi.org/10.1016/S0191-8869(02)00331-8

18.

Gupta

Sasidhar

Y. U.

(2019). Conformational dynamics of Ribonuclease Sa and its S48P mutant: Implications for the stability of the mutant protein. Journal of molecular graphics & modelling, 88(2), 71–80. https://doi.org/10.1016/j.jmgm.2019.01.001

19.

Haldorai

Murugan

Ramu

(2020). Evolution, challenges, and application of intelligent ICT education: An overview. Computer Applications in Engineering Education, 29(3), 562–571. https://doi.org/10.1002/cae.22217

20.

Harkness

J. A.

Villar

Edwards

Harkness

J. A.

Braun

Edwards

Johnson

T. P.

Lyberg

Mohler

P. P.

Pennell

B.-E.

(2010). Translation, adaptation, and design. In Smith

T. W.

(Ed.), Survey methods in multinational, multiregional, and multicultural contexts (pp. 117–140). Wiley Online Library.

21.

Howard

M. C.

(2014). Creation of a computer self-efficacy measure: Analysis of internal consistency, psychometric properties, and validity. Cyberpsychology, Behavior, and Social Networking, 17(10), 677–681. https://doi.org/10.1089/cyber.2014.0255

22.

Hsia

J.-W.

Chang

C.-C.

Tseng

A.-H.

(2014). Effects of individuals’ locus of control and computer self-efficacy on their e-learning acceptance in high-tech companies. Behaviour & Information Technology, 33(1), 51–64. https://doi.org/10.1080/0144929X.2012.702284

23.

Hsin-fang

Tzu-hsuan

(2022, July 29). Tablet plan part of digital education drive: Ministry. Taipei Times. https://www.taipeitimes.com/News/taiwan/archives/2022/07/29/2003782619

24.

Hsu

Y.-L.

Lin

T.-C.

Hsu

S.-C.

(2014). A study of ICT users’ technostress – focus on technostress comping mechanisms. National Sun Yat-sen University. [Unpublished master’s thesis].

25.

International Test Commission (2017). The ITC guidelines for translating and adapting tests (2nd ed.). International Test Commission. www.InTestCom.org

26.

Kline

R. B.

(2016). Principles and practice of structural equation modeling (4th ed.). Guilford.

27.

Yang

Zhang

(2022). Cognitive bias modification for adult’s depression: A systematic review and meta-analysis. Frontiers in Psychology, 13, 968638. https://doi.org/10.3389/fpsyg.2022.968638

28.

Liao

P.-A.

Chang

H.-H.

Wang

J.-H.

Sun

L.-C.

(2016). What are the determinants of rural-urban digital inequality among schoolchildren in Taiwan? Insights from blinder-oaxaca decomposition. Computers & Education, 95, 123–133. https://doi.org/10.1016/j.compedu.2016.01.002

29.

Ling

X.-Q.

Huang

Z.-J.

(2010). Rediscussion of the Treatment of Item 8 of Rosenberg Self-Esteem Scale Chinese Version. Chinese Journal of School Doctor, 24(4), 263–265. https://www.zgxyzz.org.cn/CN/abstract/abstract15686.shtml

30.

Marakas

G. M.

Johnson

R. D.

Clay

P. F.

(2007). The evolving nature of the computer self-efficacy construct: An empirical investigation of measurement construction, validity, reliability and stability over time. Journal of the Association for Information Systems, 8(1), 16–46. https://doi.org/10.17705/1jais.00112

31.

Mei

Brown

G. T. L.

(2018). Conducting online surveys in China. Social Science Computer Review, 36(6), 721–734. https://doi.org/10.1177/0894439317729340

32.

Murphy

C. A.

Coover

Owen

S. V.

(1989). Development and validation of the computer self-efficacy scale. Educational and Psychological Measurement, 49(4), 893–899. https://doi.org/10.1177/001316448904900412

33.

Muthén

L. K.

Muthén

B. O.

(1998–2017). Mplus user’s guide (8th ed.). Author.

34.

Nylund

K. L.

Asparouhov

Muthén

B. O.

(2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling: A Multidisciplinary Journal, 14(4), 535–569. https://doi.org/10.1080/10705510701575396

35.

Pai

Y.-J.

(2017). The Relationships of Self-esteem, Well-being and Internet Addiction among Junior High School Students. [Master’s thesis, National University of Tainan]. https://hdl.handle.net/11296/924tbn

36.

Peng

(2017). Do computer skills affect worker employment? An empirical study from CPS surveys. Computers in Human Behavior, 74, 26–34. https://doi.org/10.1016/j.chb.2017.04.013

37.

Poo

M.-m.

J.-l.

N. Y.

Xiong

Z.-Q.

Tan

(2016). China Brain Project: Basic neuroscience, brain diseases, and brain-inspired computing. Neuron, 92(3), 591–596. https://doi.org/10.1016/j.neuron.2016.10.050

38.

H.-H.

Hung

M.-L.

(2011, December 10). The Computer Self-Efficacy and Computer Attitude for Female on Consumer Behavior and Willingness of Acceptance for E-Book [Paper presentation]. The 17th conference on information management & practice. International Conference Hall at Chia Nan University of Pharmacy and Science. Tainan, Taiwan. https://oplab.im.ntu.edu.tw/csimweb/system/application/views/files/IMP/20110013

39.

Rodon

Chevalier

(2017). Toward more comprehensive Chinese Internet users’ studies: Translation and validation of the Chinese-Mandarin version of the 8-item information retrieval on the web self-efficacy scale (Ch-IROWSE). International Journal of Human-Computer Interaction, 33(10), 846–855. https://doi.org/10.1080/10447318.2017.1294335

40.

Rodon

Meyer

(2018). Self-efficacy about information retrieval on the web across all domains: A short measure in French and English. Behaviour & Information Technology, 37(5), 430–444. https://doi.org/10.1080/0144929X.2018.1449252

41.

Rosenberg

(1965). Society and the adolescent self-image. Princeton University Press.

42.

Stošić

(2015). The importance of educational technology in teaching. International Journal of Cognitive Research in Science, Engineering and Education, 3(1), 111–114. https://doi.org/10.23947/2334-8496-2015-3-1-111-114

43.

Thongsri

Shen

Bao

(2020). Investigating academic major differences in perception of computer self-efficacy and intention toward e-learning adoption in China. Innovations in Education and Teaching International, 57(5), 577–589. https://doi.org/10.1080/14703297.2019.1585904

44.

Tsai

C.-Y.

(2019). Improving students’ understanding of basic programming concepts through visual programming language: The role of self-efficacy. Computers in Human Behavior, 95, 224–232. https://doi.org/10.1016/j.chb.2018.11.038

45.

Weigold

(2021). Brief inventory of technology self-efficacy (BITS) and brief inventory of technology self-efficacy – Short form (BITS-SF) manual. https://www.bitssurvey.com/

46.

Weigold

I. K.

(2021). Measuring confidence engaging in computer activities at different skill levels: Development and validation of the Brief Inventory of Technology Self-Efficacy (BITS). Computers & Education, 169(1), 104210. https://doi.org/10.1016/j.compedu.2021.104210

47.

Zeng

(2022). An empirical research on China’s policy for ICT integration in Basic Education from 1988 to 2021. Educational technology research and development: ETR & D, 70(3), 1059–1082. https://doi.org/10.1007/s11423-022-10079-y

Translation and Validation of the Brief Inventory of Technology Self-Efficacy (BITS): Simplified and Traditional Chinese Versions

Abstract

Keywords

Computer Self-Efficacy

Brief Inventory of Technology Self-Efficacy

Relevance of CSE to Mainland China and Taiwan

The Current Studies

Study 1: Simplified Chinese

Method

Participants

Measures

Procedure

Results

Discussion

Study 2: Traditional Chinese

Method

Participants

Measures

Procedures

Results

Discussion

General Discussion

Implications

Limitations

Conclusion

Footnotes

Declaration of Conflicting Interests

Funding

Data Availability

Author Biographies

References