SRL refers to self-formed ideas, feelings, and actions that help individuals achieve their objectives (Zimmerman and Schunk, 2001; Seifert and Har-Paz, 2020). Technology-mediated SRL facilitates learners with flexible learning models and improves their language learning outcomes and motivation (An et al., 2020). Prior studies primarily focused on the effectiveness of technology-enhanced language learning within classroom instruction (An et al., 2020). There is a lack of empirical investigation of the effect of technology-mediated SRL on improving language skills. Of the limited number of previous studies that addressed technology-enhanced SRL, most reported positive relationships to language learning outcomes. For instance, Öztürk and Çakiroglu (2021) compared two groups of university students with and without SRL strategies in flipped learning. The findings indicated that SRL facilitated learning English speaking, reading, writing, and grammar. Similarly, students with SRL capabilities exhibited enhanced learning outcomes in blended learning settings (Zhu et al., 2016). In contrast, Sun and Wang (2020) found low-frequency use of SRL strategies among 319 sophomores Chinese EFL students in processes of learning writing, although SRL strategies significantly predicted writing proficiency. The students were reported to lack practice in writing during classroom sessions due to large classroom size and limited class time (Sun and Wang, 2020).

In terms of the instrument to measure SRL, the Motivated Strategies Learning Questionnaire (MSLQ) is frequently used (Pintrich et al., 1993). The MSLQ has been shown to be valid and reliable for use among undergraduate students. The original MSLQ assesses cognitive, meta-cognitive, and resource management strategies (Broadbent, 2017). Cognitive strategies involve the preparation, elaboration, and management of studies. Meta-cognitive strategies primarily refer to self-control. Resource management includes time management, effort regulation, and peer learning (Broadbent, 2017). According to Broadbent's (2017) review of 12 SRL regulated online studies, meta-cognition and resource management strategies positively influence learning outcomes, while cognitive strategies have the least amount of empirical evidence to suggest their utility. As SRL theory developed from a focus on meta-cognition to recognizing its multifaceted nature, it included motivation factors that influence learning (An et al., 2020). Pintrich (2004) noted an issue of the MSLQ is that it does not include motivational and affective factors that determine essential emotional strategies (Pintrich, 2004; An et al., 2020). Therefore, this study adopted a revised version of the MSLQ that includes four SRL aspects: cognitive, metacognitive, resource management, and emotional strategies (Supplementary Appendix 1).

The ultimate goal for the technology-supported SRL, in this context, is to improve students' writing performance and writing quality. More specfically, lexical complexity is an essential indicator of EFL writing (Lemmouh, 2008; Zhu and Wang, 2013), but few studies have addressed the effect of SRL on linguistic complexity in EFL programs. O'Dell et al. (2000) suggested that lexical complexity primarily involves lexical diversity, lexical sophistication, and lexical density (the ratio of content words to tokens). The lexical diversity aims to measure lexical variability, while lexical sophistication compares the ratio of advanced words to the total tokens. Treffers-Daller et al. (2018) highlighted the importance of integrating lexical diversity, the range of words used to measure lexical complexity, and lexical sophistication, with reference to less frequently used words as defined by various standards. Previous literature on lexical complexity development is inconclusive; some studies discovered growth after training, while others did not (Knoch et al., 2014; Kalantari and Gholami, 2017). Bulté and Housen (2014) affirmed the possibility of capturing changes in linguistic complexity in L2 writing over a short period. Inquiry into the effects of various technologies on lexical complexity is necessary so that language teachers can support students with desirable technology-mediated SRL strategies, thus enabling students to achieve enhanced learning outcomes, such as better writing quality.

The technology tools adopted in the technlogy-mediated SRL in this research are the Icourse and Pigai. Based on previous studies (Golonka et al., 2014; Yang and Dai, 2015; Zhai, 2017), the definitions of Icourse and Pigai are presented in Table 1. Massive open online courses (MOOCs) are often criticized for high dropout rates and low student engagement (Gilliland et al., 2018; de Moura et al., 2021). Icourse, as a SPOC platform, is claimed to be a valid alternative as course designers, usually course lecturers, permit the course syllabus to be flexible in difficulty and more adaptable to different student characteristics (Ruiz-Palmero et al., 2020). Guo et al. (2021) quantitatively assessed the impact of the SPOC-based blended learning model embedded in the undergraduate course of International English Language Testing System (IELTS) writing at a Chinese university in Beijing. IELTS is an international standardized proficiency English test for non-English speakers. Assessments were made of writing performance through classroom observation, questionnaires, and achievement tests in pre, mid, and final terms. The experimental group outperformed the control group in the final term test results, but there were no significant differences in pre and mid-term results. However, the study did not include linguistic parameters for evaluating SPOC platforms' effects on EFL learning of writing. Of the few studies that did include linguistic measurement, Cheng et al. (2017) addressed the impact of SPOC learning management systems on 35 Chinese undergraduate EFL learners' writing performance in terms of essay length, accuracy, and lexical complexity. The findings revealed that the SPOC learning platform helped the learners to write with increased accuracy and fluency and with an increased ratio of advanced academic vocabulary in the post-test compared to the pre-test. However, the study did not include comparison with a control group that did not use the SPOC platform. Overall, prior studies highlighted the positive role SPOC platforms play in assisting the EFL learning of writing, in terms of improving writing test scores, accuracy, and fluency. Figures 1–3 illustrate how Icourse functions as a SPOC learning management system to support browsing course materials, answering online quizzes, and interacting via discussion boards.

Besides the learning management system, Pigai is used as an AWE tool in this research. AWE aims to provide prompt writing revision feedback to learners (Liao, 2016). The major difference between AWE and teacher feedback is that AWE calculates the language gap between the EFL learner's language use and that of the native speaker (Li et al., 2019). While teacher feedback mainly relies on teachers' knowledge and teaching experience. Figures 4, 5 illustrate how Pigai works as an AWE tool to support learning how to write proficiently in English. Figure 4 shows how Pigai gives an overall mark to students' essays based on lexical, syntactic, semantic, and content parameters. A general remark on the vocabulary and sentence use is also displayed at the lower right corner of the screen. Figure 5 illustrates how Pigai provides detailed feedback regarding confusing words, synonyms, and convertible sentence patterns to expand students' vocabulary and sentence use.

Overall, positive findings support the applicability and efficiency of Pigai (Lin et al., 2020). For instance, Li and Zhang (2020) reported a positive role of Pigai in improving Chinese EFL learners' writing performance and writing self-efficacy by indicating errors in students' writing in real time and thereby enabling them to acquire vocabulary and sentence-construction knowledge. In contrast, some researchers have argued Pigai has deficits (Wu, 2017). For example, Pigai is less effective at providing feedback that helps logical thinking and content structure organization, which are also crucial factors for successful compositions, in addition to vocabulary and grammar (Wu, 2017). While the technology of Pigai constantly updates and adapts to emerging pedagogical needs, Hou (2020) called for additional studies of Pigai to keep pace with its technological advances.

Determining effective ways to support EFL learning is complex. Careful consideration should be made of the combination of various technologies, rather than favoring one specific technology over another (Lam et al., 2018).

According to the research aim, the research questions of the current study were as follows:

Purposive sampling was used to recruit the participants. The initial plan was to recruit 300 sophomore students from water conservancy engineering, mechanical engineering, electronic engineering, and allied subjects. However, although the intention was to have 100 students in each group, only 280 students agreed to participate in this study. Of these students, 99 were assigned to the control group, 90 to the Icourse group, and 91 to the Icourse+Pigai group. The participants were from the same Henan province in the People's Republic of China to ensure that they shared a similar EFL learning background. Their average age was 19 years (SD = 1.169), and each had at least 10 years of EFL learning experience since their primary education.

The research complied with all ethical stipulations of the ethics committee at the University of Malaya. Before conducting the research, the relevant university administrators were fully informed, and all the students voluntarily participated in the study, and each signed an online consent form before participating in the study. All participants remained anonymous during the entire research process.

An explanatory sequential mixed-methods approach was used to address the research questions (Figure 6). First, a quasi-experimental study was conducted to obtain quantitative comparative data, with follow-up qualitative data derived from student interviews. This study assessed three groups: a control group that received no technology-mediated SRL, an Icourse-assisted SRL group, and an Icourse and Pigai supported SRL group. Icourse and Pigai supported self-regulated learning of academic writing both in and outside the classroom. The academic writing course outline is presented in Table 2. These systems include pre-class online learning of vocabulary, watching online instruction videos about writing skills, lead-in quizzes, sentence paraphrase practice, online forum discussion, after-class review, essay writing online submission, and receiving feedback and revisions. The same academic writing course was delivered to all three groups, using the same textbook in classroom instruction. The same five units of content were covered in one academic term and the class frequency was the same, namely three times per week for 90 mins per class. Each group was recruited from a general polytechnic-focused university. All three universities were ranked at the same level. The three senior lecturers in the three groups shared a similar educational background, namely, holding a master's degree and having taught 10–12 courses. Lecturers similarly monitored students' SRL processes by setting up tasks, quizzes, and activities with deadlines, and answered students' questions online when needed.

A one-way between-groups ANCOVA was conducted to compare the effectiveness of Icourse (group 2) and Icourse+Pigai (group 3) supported self-regulated learning on the participants' writing performance as compared to the control group (group 1) after one academic term. The independent variable was the technology tools used, and the dependent variable was the post-test score. Participants' pre-test scores were used as covariates in this analysis. Preliminary checks were conducted to ensure that the assumptions of normality, linearity, homogeneity of variances, homogeneity of regression slopes, and reliable measurement of the covariate were met. After adjusting for pre-test scores, there were significant differences in mean scores among the three groups (Figure 7) [F_{(1, 211)} = 4.03, df = 2, p = 0.019, partial η² = 0.04; Table 3]. According to the pairwise comparisons shown in Table 4, the Icourse+Pigai group (M = 78.44, SD = 9.71) significantly outperformed the control group (M = 73.35, SD = 13.76; p = 0.01). The difference between the Icourse group (M = 75.16, SD = 11.46) and the control group (M = 73.35, SD = 13.76; p = 0.23) was not statistically significant.

Table 5 illustrates that the pre-test comparison indicated no statistically significant differences in lexical diversity, lexical density, or lexical sophistication among the three groups. After the intervention, significant differences were observed in all three lexical complexity indicators in the Icourse group, but only in lexical diversity and density in the Icourse+Pigai group after the intervention. The latter group showed no significant differences in lexical sophistication after the intervention. In contrast, the control group exhibited no statistically significant difference from pre-test to post-test for any of the three lexical complexity indicators.

The Kruskal-Wallis post-test revealed no statistically significant differences in lexical diversity and sophistication across the three groups (control group, n = 73; Icourse group, n = 70; Icourse+Pigai group, n = 72), χ(2,215)2 = 5.53, p = 0.063 (diversity), χ(2,215)2 = 6.02, p = 0.049 (density), χ(2,215)2 = 0.06, p = 0.970 (sophistication). The result leads the null hypothesis to be rejected that the distribution of lexical density is the same across the three groups, as there was a significant difference between the Icourse group and the Icourse+Pigai group in lexical density after the intervention.

A one-way between-groups ANOVA was conducted to explore the impact of technology tools on SRL strategies. Since the assumptions required to conduct ANOVA were met and homogeneity of variances was not violated (p = 0.36), the three groups (control group, Icourse group, Icourse+Pigai group) were compared (Figure 8). There was a statistically significant difference (p < 0.05) in SRL strategies among the three groups, F = 8.59, df = 2, p < 0.01 (Table 6). Despite reaching statistical significance, the actual differences in the mean scores among the three groups were minor. The effect size, calculated using η², was 0.06. Post-hoc comparisons using the LSD indicated that the mean score in the control group (M = 3.47, SD = 0.47) differed significantly from that of the Icourse group (M = 3.75, SD = 0.45) and that of the Icourse+Pigai group (M = 3.65, SD = 0.51). The Icourse+Pigai group did not differ significantly from the Icourse group (p = 0.15).

A list of 308 frequently occurring codes was found initially in the student transcripts and then reorganized into 46 categories of third-tier code families. Many of the themes identified in the initial coding concerned the qualities of Icourse and Pigai and the advantages and disadvantages of using the tools in academic writing. Likewise, other factors related to student needs and preferences also emerged, such as increased essay practice and unwillingness toward online peer review. As the codes were grouped and sorted, 15 categories of second-tier code families were identified, such as Icourse function, Icourse quality, Pigai function, Pigai quality, study needs, teacher influence, and peer influence. The 15 categories were then grouped as 7 broader themes and then categorized as 3 main themes and ultimately classified as two main categories as internal and external factors (Figure 9). For example, internal factors referred to study needs and study preferences, and external factors included teacher influence and peer influence.

Table 7 presents 22 categories from the 46 third-tier code families that distinctly show the differences and similarities among participants' perceptions between the two experimental groups. The five participants in the Icourse group expressed a stronger desire for an increased amount of writing practice (Icourse group: 11 citations, Icourse+Pigai: 5 citations). Compared to no complaints of Icourse drawbacks in the Icourse+Pigai group, students in the Icourse group complained about its drawbacks, such as lack of essay practice (5 citations) and inability to produce calligraphy (3 citations). Compared to the Icourse group, the most distinct feature in the Icourse+Pigai group was that students referred to self-regulated learning more frequently (Icourse group: 6 citations, Icourse+Pigai group: 10 citations). As seldom mentioned in the Icourse group, the Icourse+Pigai group referred more to the Pigai benefit of high efficiency of AWE (5 citations) and reduced teacher essay evaluation pressure (5 citations).

The Icourse+Pigai group significantly outperformed the control group in writing performance, while the Icourse group showed no significant statistical difference from the control group. The writing performance results indicate that Icourse+Pigai-mediated SRL is more conducive to enhanced writing performance than is Icourse-mediated SRL. This may be because Icourse-supported SRL fails to satisfy students' study needs for more opportunities for writing practice. As revealed by the interview results, students in the Icourse group expressed a stronger desire for frequent writing practice available through technological support.

This research finding is consistent with Rüth et al. (2021), who found that testing and quizzes were more effective for learning than was repeated exposure to learning materials. Writing practice provides relevant cognitive load, that is, knowledge construction processes that unavoidably lead to learning (Sweller et al., 1998; Nückles et al., 2020). Pigai, which supports online writing submission and provides AWE services, enables students to learn through self-regulated writing practice. This might partially explain the higher writing scores in the Icourse+Pigai group. However, the participants perceived Icourse, the learning management system, as essential in their SRL, since Pigai does not allow exposure to learning materials, online discussion, and MOOC learning. The participants felt that Icourse and Pigai are irreplaceable because the two technological tools play their own roles in SRL.

The participants' psychological study preferences also might have played a role in their SRL technological use.

No statistically significant differences in lexical complexity were found among the three groups in the pre-test. From the post-test lexical complexity results, the technology-tools-supported SRL did not significantly affect students' lexical diversity and sophistication compared with the control group. The result is consistent with the participants' interview results, in that they felt negative about Icourse and Pigai's ability to significantly improve their lexical complexity. They stated that Pigai focuses more on lexical accuracy than lexical complexity in error correction.

The negative perceptions are not consistent with Jia's (2016) finding that students perceived a higher level of satisfaction regarding improving their lexical complexity of writing with Pigai mediation and used a higher frequency of Basewords 2 and 3 (less frequent words) according to Range 32 software analysis. She also stated that lexical diversity and lexical density improved after a 12-week intervention. This is consistent with Zuo and Feng's (2015) result that Pigai's scoring criteria focus more on lexical accuracy than lexical complexity. Students tend to adjust their writing strategies according to the scoring criteria applied by Pigai to obtain high scores. Our results indicate that SRL supported by both the Icourse group and the Icourse+Pigai group affected students' writing performance in terms of lexical complexity but did not significantly improve on it in the current phase of technological development.

Finally, ANOVA of SRL strategies revealed significant differences between the control and technology-supported groups. The results indicated that both Icourse and Icourse+Pigai positively related to the participants' use of SRL strategies. This aligns with van Alten et al. (2020) study, which found that providing students with technological SRL prompts is an effective strategy for improving SRL. They found that providing online videos in the process of flipped learning was positively related to students' learning outcomes. Likewise, Öztürk and Çakiroglu (2021) demonstrated that technology-mediated SRL positively enhanced students' writing skills in a flipped learning environment. According to Broadbent and Poon's (2015) review, enhanced SRL strategies positively influence learning outcome because, despite cognitive skills having a relatively negligible influence on improving learning outcomes, metacognition, time management, and critical thinking skills are positively related to learning outcomes.

However, the Icourse and Icourse+Pigai groups exhibited no significant difference in the use of SRL strategies, indicating that variation in technology tools did not significantly affect the participants' SRL strategy use. Students' psychological study preferences may partially explain this finding. Psychological study preference is compared to physical study preferences, such as, visual, aural or kinesthetic influences on study preferences. In this research, it refers to the possible psychological factors that affects students' choices on some educational modes over others. In previous studies, study preference primarily referred to sensory modality preferences. This denotes those students make study choices physically, through vision or auditory reactions (Hu et al., 2018). However, the study preferences in this research primarily referred to students' psychological factors. For instance, the interview results reflected those participants tended not to use the peer evaluation function in Pigai, even if they were told that it could be helpful to their writing. They expressed feelings of “distrust” and “embarrassment” regarding showing their essays to classmates.

Our results are consistent with van Alten et al. (2020), who found that some students disliked the SRL prompts even though the SRL support encouraged students to be more conscious of their learning. Likewise, Yot-Domínguez and Marcelo (2017) found that students tended not to use mobile-related technology tools in their SRL, but rather used mobile devices for social communication purposes. Students' psychological study preferences affect their study choices regarding technology prompts, which may subsequently influence their SRL strategy use.

The research shed light on the possible factors to consider when improving students' technology-enhanced SRL experience. Findings indicate that students' perceptions toward technology-supported SRL on their academic learning of writing vary, to some degree, by both internal and external factors (Figure 9). The state-of-art technology innovations alone do not guarantee an effective learning process and outcome (Hao et al., 2021). Chew and Ng (2021) emphasized the importance of integrating the effects of students' personality and proficiency in their word contribution in online forums. Different personality traits, such as, introverts or extroverts, may lead to different word productions in their online discussion with the same technological tool (Chew and Ng, 2021). Similarly, Lai et al. (2018) reported that various external and internal factors influence students' perceived study engagement. They proposed a new perspective in viewing technological use as diversified, which means one technological use can generate multiple forms of technology supported learning experiences. Rather than viewing technology as a whole entity in itself, students' psycho-social factors are also essential in contributing to their technologically supported learning experience (Lai et al., 2018). Our research consistently supported their finding by recognizing the importance of integrating students' internal needs and psychological factors with external factors: technological quality and performance and environmental impact are equally important as technological advances in the design and implementation of technology-supported SRL for students.