Vygotsky’s sociocultural theory

According to Vygotsky [30], learning develops through interaction with mediating tools and more capable others. In the present study, ChatGPT functions as a mediating scaffold that supports learners while they analyze reading passages and construct mind maps. ChatGPT does not replace the learner but supports them in locating the main idea, supporting details, and additional information, and in organizing these into a structured representation. Such scaffolding matters most when students are not yet able to carry out these tasks on their own. In this respect, ChatGPT can be understood as operating within the learner’s Zone of Proximal Development, where guided assistance enables the completion of comprehension tasks that would otherwise lie just beyond reach.

This theoretical lens is particularly relevant because the intervention was not based on passive acceptance of AI output. Instead, students were trained to co-construct mind maps with ChatGPT by reading the passage first, entering structured prompts, reviewing the generated output, and then validating it against the original text. This process positions ChatGPT as a supportive mediator in learning rather than a source of ready-made answers. The instructional value of the tool, therefore, lies in its ability to guide learners’ thinking and help them engage more actively with reading passages.

Cognitive load theory

The second theoretical foundation of the study is Cognitive Load Theory [31]. RC in a foreign language often places a high burden on working memory because learners must simultaneously process vocabulary, sentence structure, and the relationships among ideas. When information is presented in a long linear form, students may struggle to retain and organize it efficiently. Mind mapping can help reduce this burden by externalizing the structure of the text into a visual format. By showing the relationships among ideas clearly, mind maps may reduce extraneous cognitive load and allow learners to focus more effectively on comprehension.

In the context of the present study, GPT-mediated mind mapping may reduce cognitive load in two ways. First, it helps students reorganize textual information into manageable units, making it easier to distinguish central ideas from secondary details. Second, the validation stage asks students to compare the AI-generated map with the original passage, drawing their attention back to both meaning and organization. GPT, on this view, does more than produce an output: it helps learners engage with the text in a more systematic and efficient way.

Taken together, these two theories provide a focused conceptual model for the intervention. Sociocultural Theory explains how ChatGPT functions as a scaffold and mediating tool during mind-map construction, while Cognitive Load Theory explains why organizing information visually may support RC by reducing unnecessary mental burden. Within this model, students’ validation of ChatGPT output is an essential step because it requires them to review, refine, and confirm the meaning of the passage rather than accept the output automatically. Thus, the theoretical framework of the study is not based on multiple separate theories, but on a coherent explanation of how guided mediation and reduced cognitive load work together to support EFL learners’ RC through GPT-mediated mind mapping.

Previous studies

A number of studies have examined how different mind mapping techniques affect RC across a variety of educational settings. Working with ChatGPT-mediated mind mapping for collaborative brainstorming, Duong [6] reported a significant correlation between students’ grasp of the method and their ability to generate and organize ideas. In the Thai context, Chaichompoo [13] found that electronic mind mapping improved both RC and summarization among English majors. Comparable gains have been documented by Mohaidat [17] with Jordanian students and by Almelhi [9] with EFL college learners. In Indonesian university classrooms, Andoko et al. [12] and Pinandito et al. [20] point to kit-build concept mapping as an effective support for English RC when it is used in collaborative learning. Work by Hamid [15], Alomari [10], and Alomari and Alhorani [11] likewise documents the benefits of digital mind mapping for RC across different educational contexts, while Adel Baghagho et al. [7] report gains on several dimensions of RC among Egyptian preparatory school students. Other studies broaden the picture: Hazaymeh and Alomery [16] show that visual mind mapping strategies enhance critical thinking and reading proficiency among English language learners; Ghorbani Shemshadsara et al. [14] find that computer-based instruction with mind mapping tools strengthens both text-structure awareness and RC at the undergraduate level; and Rittichai and Torat [21] demonstrate the value of mind mapping delivered through platforms such as Edmodo for English reading proficiency. Similar improvements in RC have been observed with electronic mind mapping applications in Novitasari’s [19] work with college students and Monliang’s [18] with high school students. Finally, Aljaser [8] and Yan and Kim [22] report positive effects of electronic mind mapping on academic achievement and on schema strategy instruction, reinforcing the broader role of mind mapping in supporting RC across educational settings.

Recent scholarship on AI in EFL education further indicates that AI-supported instruction can strengthen learners’ higher-order thinking, classroom engagement, and openness to adopting new learning tools. For example, Liu and Wang [32] found that an AI-based instructional intervention in English literature classes was associated with significant gains in EFL learners’ critical thinking. In addition, Wu et al. [33] showed that EFL learners’ intention to use AI is strongly shaped by perceived ease of use and positive attitudes toward AI. From an affective perspective, Derakhshan and Park [34] reported that AI-mediated instruction may increase positive achievement emotions while reducing negative ones. Read together, these studies indicate that AI-supported learning can shape more than performance alone; it also touches on learners’ cognitive engagement, their emotional experience of learning, and their readiness to adopt educational technology.

Using mind mapping as a teaching tool for RC also comes with difficulties for students, particularly around selecting which ideas to foreground and bringing enough creativity to the task. While recognized for its benefits, concerns persist regarding its optimal use—whether individually or in collaborative settings [23]. Kepirianto et al. [24] argue that not all types of texts are suitable for mind mapping, especially when combined with collaborative methods like Think-Pair-Share, as effective comprehension often requires sufficient background knowledge. Yan and Kim [22] highlight additional obstacles such as uneven time management, difficulties in collaboration, and usability issues with digital tools, which detract from the effectiveness of mind mapping instruction in reading education.

Participants

The study was carried out at a university in the southwestern region of Saudi Arabia, with participants drawn from the Preparatory Year Program. This program plays a pivotal role for science-stream students in the Saudi system, bridging high school and specialized university pathways such as medicine, engineering, computer science, pharmacy, and nursing. Its curriculum is designed to meet university entry requirements and delivers intensive coursework in computer science, mathematics, communication skills, and English. Additionally, the program provides psychological preparation through counseling sessions to help students adjust to university life. In the Preparatory Year, classes are gender-segregated, with male teachers instructing male students and female teachers instructing female students. Due to communication difficulties with the female section, the study specifically targeted the male section.

The study examined the extent to which specific teaching methods helped male students develop as readers. Two intact level-one reading classes, totaling 60 students, were selected through a straightforward sampling procedure and assigned to serve as the experimental and control groups. The quasi-experimental design allowed for a systematic comparison between the two groups.

As part of the intervention, the experimental group was taught through interactive reading activities, while the control group received conventional instruction. Participants were homogeneous across several characteristics: gender (all male), high-school background (science stream), age (18–19), first language (Arabic), prior English instruction (seven years), academic level (level 1), placement-test band (pre-intermediate), and course (Reading Skills).

This study received full ethical approval from the Najran University Research Ethics Committee (Approval No.: 202505-076-032278-073379) on 27/05/2025. The approval confirmed that the project, “Investigating ChatGPT-mediated mind mapping to facilitate EFL learners reading comprehension,” met the requirements of the National Committee for Bioethics at King Abdulaziz City for Science and Technology. In accordance with this approval, written informed consent was obtained from all participants before their involvement in the study. All data were anonymized to protect participant confidentiality, and the research was conducted strictly in accordance with the approved proposal, adhering to all standard conditions set by the ethics committee throughout the investigation. The results are expected to have implications for curriculum development and teaching strategies by offering insights into successful reading instruction strategies for students in the preparatory year. The program and data collection started right after obtaining ethical approval. The study started on 28/5/2025 and ended on 14/8/2025.

Moreover, this research utilized generative AI (Copilot) to assist with the search for authentic academic sources, editing, and proofreading. It was also used to rephrase and organize ideas and draft and format sections such as results and references.

Tools

Data were collected through two instruments: an RC achievement test and semi-structured interviews. The achievement test measured participants’ RC performance before and after instruction in the ChatGPT-mediated mind mapping strategy. The interviews were used to explore participants’ learning experiences, with particular attention to the difficulties they encountered when working through reading passages using ChatGPT-assisted mind maps.

Intervention

The 10-week training program using ChatGPT (free version) is designed to enhance RC skills among EFL learners by integrating mind-mapping techniques with GPT support. The program is structured around explicit instruction in RC strategies—summarizing, predicting, and questioning—so that students can actively monitor and strengthen their understanding of texts. It opens with a two-hour training session, led by an experienced EFL instructor, that is designed to prepare students to use ChatGPT’s features effectively for mind mapping.

The free version of ChatGPT was chosen because its role in this intervention was not to generate visually polished mind maps, but to help learners analyze reading passages—identifying the main idea, supporting details, and additional information—and to arrange these elements into a simple mind-map layout. The free plan suits this purpose well: it allows users to work with uploaded files, summarize texts, and receive explanations in clear and accessible language, all of which align closely with the reading-comprehension aims of the intervention.

Specialized mind-mapping platforms such as MindMeister, Coggle, Xmind, and Lucidchart, by contrast, are built primarily for the visual creation, editing, and sharing of diagrams. Their main strengths lie in layout design, templates, collaboration, and export features, rather than in guiding learners through the cognitive work of interpreting and simplifying reading passages. Their free tiers are also typically limited, often capping the number of maps or diagrams a user can create, or restricting AI-supported functions to trial access.

The 10-week duration adopted for this ChatGPT-mediated RC program is consistent with earlier research on effective training periods for RC and related cognitive skills in educational settings. Adel Baghagho et al. [7], among others, argue that sustained training of 6–12 weeks gives students enough time to internalize and apply new strategies—digital mind mapping included—without overloading their schedules, and report that mind mapping delivered over such a structured period improved EFL students’ comprehension by organizing textual information visually and fostering deeper engagement with the text. Aljaser [8] and Almelhi [9] reach a similar conclusion: gradual, week-by-week work with electronic mind maps was associated not only with better comprehension but also with more positive attitudes toward learning English, suggesting that consistent, extended practice yields the strongest cognitive gains. Alomari and Alhorani [11] demonstrate that intervals of at least eight weeks are particularly effective for RC improvements in EFL settings, providing students with enough time to familiarize themselves with mind mapping techniques and integrate them into their learning processes. Integrating ChatGPT over a structured period offers additional advantages by combining technology with traditional methods. The growing interest in AI in education, as examined by Xu et al. [36] and Limna et al. [37], points to AI tools’ ability to promote personalized learning, which can be a novel addition to face-to-face instruction. This integration into a 10-week program thus capitalizes on the optimal training period for critical thinking skills and represents an innovative approach to RC in educational technology.

Throughout the program, students engage in hands-on training sessions, where they practice applying mind mapping to identify main ideas, supporting details, and relevant information in reading passages, guided by the prescribed reading textbook: Academic Progress Reading & Writing Student Book (P2). After the initial two-hour orientation session in Week 1, the experimental group completed one GPT-mediated mind-mapping task each week during the regular reading class. Each task was carried out within one class session of approximately [30 minutes], and the activity followed the same instructional sequence across the 10-week intervention.

The program builds in follow-up activities and regular reinforcement across the study period, giving students repeated opportunities to consolidate each RC strategy. A pre-test established baseline performance for both groups, after which a post-test and semi-structured interviews were used to evaluate the program’s effect on students’ comprehension and on their experiences of using GPT for mind mapping. Taken together, the design offers a manageable yet intensive framework for integrating technology into EFL reading instruction, in line with current research on strategy training in language education.

In the first week of the semester, an experienced EFL instructor with a background in technology integration led a two-hour classroom training session. Its aim was to equip students with the skills needed to use ChatGPT’s features to apply the mind-mapping technique successfully. This orientation included: (a) a teacher demonstration of how to read a passage first and identify the main idea, supporting details, and additional information; (b) guided practice in entering prompts into ChatGPT; (c) explanation of how to review and validate AI-generated output; and (d) a short practice task completed individually or in pairs. A full sample training lesson, worksheet, and sample prompts are provided in S1 Appendix.

The independent variable examined here is the instruction on strategy use (GPT-mediated mind mapping technique), based on the hypothesis that explicit use of mind mapping mediated by GPT will impact students’ RC. Following Oxford’s [38] model of strategy training, the program involved steps such as assessing learners’ needs and resources, assigning strategies, considering anticipated benefits and motivation, preparing materials and activities, implementing training, and evaluating its effectiveness.

To determine baseline RC performance, a pre-test was taken by the experimental and control groups before the intervention. To guarantee that the two groups performed equally, the pre-test was given using the identical RC accomplishment test as the post-test. The pre-test covered the same material as the post-test and had two open-ended and three multiple-choice questions. The experimental and control groups’ RC skills were equivalent before the intervention, as evidenced by the results, which revealed no discernible differences.

After the pre-test, the training program for the experimental group commenced. The experimental group received training on mind mapping techniques using tasks and activities from the textbook, integrating identified strategies. The teacher-trainer guided students using mind mapping with ChatGPT to understand reading passages by identifying main ideas, supporting details, and additional information.

The experimental group received the necessary training on how to use GPT-mediated techniques to support RC. After this training, learners used ChatGPT to help them identify the main idea, supporting details, and additional information in reading passages and to organize these elements into simple mind maps. In contrast, the control group was taught the same reading-comprehension objectives through conventional mind-mapping instruction, using traditional classroom methods without the support of any generative AI tools. Thus, both groups were exposed to mind mapping as a comprehension strategy, but only the experimental group used ChatGPT as a mediating tool in the process.

During training, the experimental group used GPT to construct mind maps of the prescribed reading passages. The training focused on how to build mind maps that, by drawing on GPT’s features, would deepen students’ comprehension of the text. Through group activities and hands-on practice, students engaged directly with the technology during each session. Using ChatGPT, they learned to rely on a mind map to identify the key points and illustrative features of a sample reading passage drawn from the required textbook: Academic Progress Reading & Writing Student Book (P2). The procedure was built around co-construction rather than passive acceptance of the AI’s output. In each task, students first read the assigned passage on their own and jotted down what they took to be the main idea and the key details. They then submitted a guided prompt to ChatGPT to obtain a preliminary text-based mind map. Once the response appeared, they compared it against the original passage, verified whether the main idea, supporting details, and additional information had been captured accurately, and either refined the prompt or edited the mind map by hand. In this sense, ChatGPT functioned as a mediating support tool, while the final version of the map remained subject to student review and teacher confirmation.

In this intervention, ChatGPT is used only to support RC. After reading a passage, learners use ChatGPT to create a simple mind map that identifies the main idea, the supporting details, and any additional information in the text. This helps learners organize what they read in a clear visual form and improves their understanding of the passage. The reading texts used for this purpose can include topics such as hobbies, favorite pastimes, start-up companies, world records, social media, deep-sea exploration, and deep-sea creatures.

The prompt pattern taught to students was intentionally simple and consistent. Students were trained to use prompts such as: “Help me understand this reading passage. Identify the main idea, three supporting details, and two pieces of additional information. Then make a simple mind map in easy English,” and “Turn this reading passage into a simple mind map. Put the main idea in the center, add branches for supporting details, and add one branch for additional information. Use short phrases only.” These prompt models were repeated throughout the intervention so that students learned to use ChatGPT as a structured reading-comprehension aid rather than as an automatic answer generator. Table 1 presents sample mind maps generated by students using these prompts across a range of reading passages from the course textbook, with additional examples provided in S1 Appendix.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Sample RC mind maps generated with ChatGPT.

https://doi.org/10.1371/journal.pone.0336185.t001

To make validation explicit, students were instructed to check each AI-generated mind map against the original passage using questions (see S1 Appendix). If any part was incomplete, inaccurate, or oversimplified, students revised the output either by re-prompting ChatGPT or by editing the map themselves. The teacher monitored this process and confirmed the final version during classroom discussion.

Regular follow-up activities were planned to reinforce the methods gained over the 10-week study period. Thus, each instructional cycle involved reading the passage, prompting ChatGPT, reviewing the generated output, validating it against the source text, and revising the final mind map. A post-test (standardized mid-term test) was given at the end of the training session to assess RC’s progress. This evaluation functioned as a quantifiable gauge of program efficacy and individual development. Students then participated in semi-structured interviews to discuss their experiences utilizing mind maps mediated by GPT to help them understand reading passages.

Data analysis

To investigate the first research question, both descriptive and inferential statistical analyses were conducted using SPSS version 28. The researcher began by calculating descriptive statistics (means, standard deviations, medians, ranges, and interquartile ranges) for both groups’ pre-test and post-test scores. The normality of distribution was assessed using the Shapiro-Wilk test and Kolmogorov-Smirnov test with Lilliefors correction, as recommended for sample sizes under 50. Homogeneity of variances was evaluated using Levene’s test for equality of variances. Given that the experimental group’s post-test scores violated the assumption of normality (Shapiro-Wilk p = .001), the non-parametric Mann-Whitney U test was employed to compare the post-test scores between groups, rather than the independent samples t-test. For the pre-test comparison, where both groups showed normal distribution, an independent samples t-test was appropriately used to establish baseline equivalence. The reliability of the RC test was assessed using the Kuder-Richardson Formula 20 (KR-20), which yielded a coefficient of 0.79, indicating acceptable internal consistency for the measurement instrument. The practical significance of the intervention was evaluated by calculating the effect size (r) using the formula r = Z/√N, where Z is the standardized test statistic from the Mann-Whitney U test and N is the total sample size.

For the second research question, which explored participants’ learning experiences with ChatGPT for RC, the researcher conducted semi-structured interviews. The researcher developed interview protocols to guide participants in reflecting on their experiences using the ChatGPT-mediated mind map technique to comprehend reading passages in an EFL context. Participants discussed challenges they encountered when using GPT for learning RC through mind maps. Qualitative data from these interviews underwent thematic analysis, with researcher categorizing responses based on recurring themes [39]. Fig 2 illustrates Braun and Clarke’s [39] six-phase model for thematic analysis. Each phase includes a brief description to guide the systematic process of identifying and defining themes within qualitative data. This model emphasizes a structured approach to ensure in-depth and reliable insights from qualitative studies.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Braun and Clarke’s [39] six-phase model for thematic analysis.

https://doi.org/10.1371/journal.pone.0336185.g002

Role of ChatGPT-mediated mind mapping technique in enhancing EFL readers’ RC

This report presents a complete statistical analysis of an experimental study investigating the effects of a ChatGPT-mediated mind mapping technique on the RC of EFL learners. The analysis includes pre-test equivalence testing, normality assessments, and both descriptive and inferential analyses of post-test results to provide a comprehensive understanding of the intervention’s effectiveness.

Pre-Test analysis: Establishing group equivalence.

The pre-test descriptive statistics in Table 2 reveal important initial characteristics of both groups. While the measures of central tendency (mean and median) are remarkably similar between groups, indicating comparable average performance levels, the control group demonstrates substantially greater variability in scores. This is evidenced by the larger standard deviation (4.05 vs 2.58), variance (16.43 vs 6.64), and interquartile range (5.62 vs 2.50). The Experimental group’s scores were more tightly clustered around the mean, while the control group showed more dispersed performance levels.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Descriptive statistics for experimental and control groups (pre-test).

https://doi.org/10.1371/journal.pone.0336185.t002

As shown in Table 3, both normality tests (Kolmogorov-Smirnov and Shapiro-Wilk) yielded non-significant results (p > .05) for both groups, confirming that the pre-test RC scores followed a normal distribution. This satisfies the normality assumption for parametric testing of pre-test data.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Tests of normality (pre-test).

https://doi.org/10.1371/journal.pone.0336185.t003

In Table 4, the independent samples t-test confirms no statistically significant difference between groups at baseline (t(58) = 0.78, p = 0.437). This establishes that any post-intervention differences cannot be attributed to pre-existing differences in RC ability.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Independent samples t-test for pre-test RC scores.

https://doi.org/10.1371/journal.pone.0336185.t004

Post-test analysis: Evaluating intervention effects

As depicted in Table 5, the post-test descriptive statistics reveal substantial differences between groups. The experimental group shows markedly higher central tendency across all measures (means, medians, & modes). Notably, while the Experimental group maintained relatively consistent performance (SD = 2.77), the control group continued to show greater score dispersion (SD = 4.22). The experimental group’s higher minimum score (16.5 vs 10) suggests more consistently strong performance across all participants.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Descriptive statistics for experimental and control groups (post-test).

https://doi.org/10.1371/journal.pone.0336185.t005

Table 6 displays that the normality tests indicate violation of the normality assumption for the Experimental group (Shapiro-Wilk p = .001) while the control group maintained normal distribution (p = .364). This violation necessitates the use of non-parametric tests for post-test comparisons.

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Tests of normality (post-test).

https://doi.org/10.1371/journal.pone.0336185.t006

As shown in Table 7, the Mann-Whitney U test identified a statistically significant difference in post-test RC scores between the two groups (U = 180.00, p < .001). The experimental group (Mean Rank = 39.50) clearly outperformed the control group (Mean Rank = 21.50). With an effect size of r = 0.52, the result carries medium-to-large practical significance, indicating that the intervention was educationally as well as statistically meaningful.

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. Mann-Whitney U test for post-test RC scores.

https://doi.org/10.1371/journal.pone.0336185.t007

Taken together, the statistical analyses offer strong evidence that ChatGPT-mediated mind mapping supported EFL RC in this study. The pre-test analysis confirmed that the two groups entered the study with comparable RC ability, ruling out pre-existing differences as an alternative explanation for what followed. At post-test, the experimental group scored substantially higher than the control group and, as reflected in the smaller variability of their scores, showed more uniform gains across participants. Combined with a significant Mann-Whitney U result (p < .001) and a substantial effect size (r = 0.52), these patterns provide solid evidence that was linked to improved RC performance. The uniformity of the effect observed within the experimental group further suggests that the intervention may have benefited a broad range of learners rather than only the higher-achieving ones—a point worth underlining given the control group’s persistent variability throughout the study. In practical terms, the 4.29-point mean difference represents a sizeable shift, moving the experimental group from moderate to high performance levels while the control group showed only minimal gain over time.

Overall, the post-test findings lend support to ChatGPT-mediated mind mapping as an instructional approach for EFL RC. Students in the experimental group performed significantly better than their counterparts in the control group, and the size of this difference was educationally meaningful. However, given the quasi-experimental nature of the study, these findings should be interpreted as evidence supporting the usefulness of the approach, rather than as definitive proof of causation.

EFL students’ learning experiences of utilizing the GPT-mediated mind mapping technique in RC passages

Qualitative data from the semi-structured interviews were analyzed thematically following Braun and Clarke’s [39] six-phase model, which involves familiarization with the data, generating initial codes, searching for themes, reviewing themes, defining and naming themes, and producing the report. This systematic procedure enabled the researcher to identify recurring patterns in participants’ responses and to organize them into coherent themes and subthemes. The thematic analysis of the interview data revealed three major themes related to students’ experiences with GPT-mediated mind mapping in RC. These themes, along with their associated subthemes, are presented in Table 8.

Download:

• PNG
  larger image
• TIFF
  original image

Table 8. Emerging themes and subthemes from the qualitative interview data.

https://doi.org/10.1371/journal.pone.0336185.t008