The relationship between higher-order thinking and problem-solving skills development among pre-service teachers using generative AI: an analysis based on moderated mediation

Abstract

Background

Artificial intelligence (AI), as a smart and connected technology, has significantly expanded the educational landscape.As more educators and learners begin to rely on GAI to assist with tasks such as instructional design and information generation, its potential to support problem-solving has gained increasing attention. However, the extent to which GAI contributes to the development of higher-order thinking skills—such as creative thinking, critical thinking, and metacognitive awareness—and how these thinking processes interact to influence problem-solving ability remains underexplored. A more comprehensive understanding of this relationship is needed to guide the effective integration of GAI in educational practice.

Methods

The study encompassed 473 pre-service teachers from three distinct higher education institutions, with specialties in science, computer science, and mathematics, and included a four-week generative AI-supported instructional design training program. And a semistructured interview comprising open-ended questions was administered to 50 pre-service teachers within the experimental group to present their views on generative AI-assisted teaching. Assessments were conducted before and after this program using a higher-order thinking skills survey. The relationship between thinking and problem-solving ability development of pre-service teachers using generative artificial intelligence was analyzed with a moderated model.

Results

The study encompassed 473 pre-service teachers from three distinct higher education institutions, with specialties in science, computer science, and mathematics, and included a four-week generative AI-supported instructional design training program. And a semistructured interview comprising open-ended questions was administered to 50 pre-service teachers within the experimental group to present their views on generative AI-assisted teaching. Assessments were conducted before and after this program using a higher-order thinking skills survey. The relationship between thinking and problem-solving ability development of pre-service teachers using generative artificial intelligence was analyzed with a moderated model.

Conclusions

This study confirms that training pre-service teachers using generative AI to foster creative thinking can elevate their critical thinking, thereby impacting their problem-solving abilities. Moreover, metacognitive thinking amplifies the impact of creative thinking on critical thinking, resulting in a moderated mediation effect.

Introduction

Since the launch of GAI in November 2022, generative artificial intelligence (GAI) has been developing at an exponential rate, marking a new phase in the evolution of internet technology. Its impact on the education sector has become an irreversible trend [1]. As an advanced form of modern technological development, GAI has permeated the learning space of pre-service teachers, significantly influencing teaching objectives and problem-solving abilities. These abilities are typically cultivated through prolonged training in modern education. However, GAI’s efficiency in knowledge production, dissemination, and application can drastically shorten this training process [2]. In this context, it is inevitable that traditional education’s process of cultivating pre-service teachers’ problem-solving abilities will undergo profound transformation.

The article primarily focuses on pre-service teachers’ problem-solving abilities in terms of critical thinking, creative thinking, and metacognition. The purpose of education is to cultivate pre-service teachers into critical thinkers, including the ability to engage in effective thinking, employ appropriate reasoning, conduct independent evaluations, and effectively address real-life problems. Creative thinking involves generating new and better problem-solving methods, which are considered imaginative, constructive, and innovative [3]. Additionally, metacognition is defined as thinking about thinking [4].Generative AI can help teachers quickly generate syllabi, lesson plans, activity schemes and assessment tasks, significantly reducing preparation time and improving efficiency. Metacognition consists of two main components: monitoring and evaluating one’s own cognition regarding the knowledge that should be acquired as a learner (knowledge) and regulating one’s own cognition (regulation). All three types of thinking are not only influenced by GAI but also impact the development of pre-service teachers’ problem-solving skills. Therefore, the research significance of this study is important for the professional prospects of pre-service teachers.

The primary reason pre-service teachers rely on GAI is that it generates unexpected content in the form of images, text, and audio [5]. In the past, people had to collect a large amount of sensory data before making rational judgements when acquiring and generating information and knowledge. Currently, GAI is revolutionising existing educational and teaching methods by quickly providing facts and perspectives from databases, thereby assisting pre-service teachers in developing problem-solving skills [6]. However, the cultivation of problem-solving skills hinges on fostering diverse thinking patterns. For example, different groups require distinct problem-solving approaches in specific contexts, such as the creative thinking of art teachers and the critical thinking of history teachers. Therefore, the impact of GAI on the development of pre-service teachers’ thinking and problem-solving skills is a potential and valuable area of exploration.

Although GAI has been highly effective in supporting teacher professional development, research on GAI’s impact on teachers’ problem-solving abilities is limited, and studies on the factors and mechanisms influencing these abilities are also relatively scarce. On one hand, most research primarily focuses on how students’ creative and critical thinking abilities influence their problem-solving capabilities, with limited attention given to teachers’ problem-solving skills. On the other hand, while the cultivation of problem-solving abilities in educational processes is a frequently discussed topic, there is little discussion of the impact of GAI on cognitive and problem-solving abilities [7]. To address these gaps and better discuss how to provide inclusive and adaptive AI learning environments, this study explores the impact of critical and creative thinking developed in GAI environments on pre-service teachers’ problem-solving abilities.

Literature review

Generative AI and educational practice

In its latest report, United Nations Educational, Scientific and Cultural Organization (UNESCO) defines GAI as a digital technology for automatically generating content based on human cognitive symbolic representation systems [8]. The two core technologies of GAI are rapid content generation and dialogue understanding. These technologies have been effectively extended and expanded in the education sector, significantly enhancing support for self-directed learning, such as inspirational guidance, customised learning content, and learning path planning (Fischer & Scharff, 1998). Additionally, data indicates that 64.24% of students hope teachers will appropriately incorporate GAI into their teaching processes [9].

GAI is one of the earliest representatives of generative technology, demonstrating strong interactive capabilities in code programming, real-time question-answering, and scenario translation [10]. Shoufan found through a survey that pre-service teachers generally believe GAI’s functions are highly effective for learning [11]. One reason is that interactive dialogue can enhance critical and creative thinking through immediate feedback [12]. Additionally, Yu et al. particularly emphasised the importance of GAI’s multilingual translation and human-machine interaction for foreign language teaching [13]. GAI not only provides pre-service teachers with abundant language learning resources but also simulates real-life language environments.

For pre-service teachers, the popularity of GAI among students stems from the educational reforms it has sparked, which have significantly impacted teaching and learning methods. Through continuous dialogue and text generation, it provides pre-service teachers with a midwifery approach to learning that is not constrained by educational resources [14]. Additionally, by leveraging intelligent dialogue to understand pre-service teachers’ learning status, needs, and behaviours, and using deep learning algorithms to recommend learning resources, it addresses their diverse and personalised requirements [15]. A related study indicates that GAI’s initial responses may prompt further teacher-student interaction and encourage pre-service teachers to apply their knowledge and reasoning skills [16].

Furthermore, many studies have emphasised that GAI significantly influences pre-service teachers’ learning abilities, thinking processes, and teaching methods by constructing learning scenarios [17]. For example, GAI can assist teachers in developing course outlines, including writing a complete learning content outline and teaching schedule [18]. Another study indicates that GAI is beneficial for blended learning reforms, as it can stimulate students’ interest in learning, broaden their knowledge base, improve learning efficiency, and reduce educational inequality [19]. Therefore, the advantages of GAI in programming, debugging code, enhancing classroom assessment, and assisting in instructional design are increasingly being recognised by pre-service teachers [20].

Problem-solving abilities, creative thinking, critical thinking, metacognition

A problem refers to a dilemma where an individual wants to accomplish something but cannot take swift action [21]. Pre-service teachers must master numerous problem-solving skills in education before becoming qualified teachers. American educational psychologist Jonassen categorises problems into well-structured problems and ill-structured problems [22]. In well-structured problems, the problem has a clear objective, and solutions can be found by following fixed rules and principles. In ill-structured problems, the problem description is vague and incomplete, so solutions to ill-structured problems are often unpredictable. They may have multiple solutions or no solution at all. This poses a significant challenge for pre-service teachers developing problem-solving skills, but GAI can break through this limitation of having no clear objectives.

GAI can also quickly address problems involving different types of thinking. Based on the differing approaches to problem-solving, problem-solving can be categorised into conventional problem-solving and creative problem-solving [23]. Conventional problem-solving refers to the process of addressing issues using existing or readily available resources. Creative problem-solving involves seeking innovative, flexible, and original solutions to problems. As a high-level comprehensive creative activity, the process of creative problem-solving requires individuals to overcome mental biases and approach problems from a completely new perspective [24]. Undoubtedly, the emergence of GAI has reorganised the frameworks in the minds of pre-service teachers to address various creative problems.

Creative thinking refers to the cognitive process of solving problems in novel and unique ways [25], including divergent thinking and convergent thinking. Numerous studies have shown that creative thinking has a positive impact on problem-solving abilities [26]. For example, students with higher scores in innovative thinking solve problems faster than those with lower scores [27]. Additionally, for an educational group, the application of creative thinking across different disciplines can significantly influence the group’s problem-solving abilities in those areas. Especially when addressing real-world problems in specific higher education work environments, creative thinking is also necessary [28]. For instance, applying creative thinking in the entrance exams for architecture design students positively impacts their ability to solve problems in specific fields [29]. In the field of computer science, the use of innovative thinking in computer models and algorithms enhances individuals’ creativity and learning abilities in problem-solving [30].

Furthermore, the development of creative thinking is closely related to learning scenarios and teaching methods associated with GAI. For example, a mobile physics learning programme supported by intelligent Android devices has been proven to enhance the creative thinking and problem-solving abilities of teaching teams [31]. Smart robots and digital coding education significantly influence the creative abilities of preschool children [32] The use of remote-controlled cars in mechanical innovation courses significantly improved participants’ problem-solving skills. This study primarily examines the impact of creative thinking on pre-service teachers’ problem-solving abilities in the context of integrating GAI into teaching environments. Therefore, we propose the hypothesis. H1: When using GAI for teaching support, the development of creative thinking among pre-service teachers has a positive impact on their problem-solving abilities.

Critical thinking refers to a purposeful, logical cognitive process that emphasises analysing, evaluating, and reasoning to judge the authenticity, validity, and value of information, ultimately leading to reasonable decisions or reliable conclusions [33]. There is no clear consensus on its definition, but it generally encompasses aspects such as questioning, rationality, and reflection. For example, Robert Ennis defines critical thinking as the reasonable and mature thinking individuals engage in when determining what to believe or do [34]. McPeck views critical thinking as the skills and tendencies that emerge in reflective sceptical activities [35]. Paul defines critical thinking as the distinction between inadequate dialogue and self-directed thinking [36]. From this, it can be inferred that critical thinking for pre-service teachers is not only about self-criticism of teaching methods but also about avoiding erroneous decisions caused by biases and emotional reactions toward students.

Creative thinking and critical thinking are key components of higher-order cognitive processes and are crucial for the development of pre-service teachers’ skills [37]. Some studies indicate that student-centred teaching focused on cultivating critical thinking and creative thinking can foster interdisciplinary higher-order thinking [38]. Different teaching strategies can enhance students’ critical thinking. Students’ composite computational thinking and creative higher-order thinking can promote the development of critical thinking and language output in complex relational forms [39]. There is a significant correlation between students’ creative and critical thinking abilities in solving unconventional problems [40]. H2: When using generative artificial intelligence for teaching support, the development of pre-service teachers’ creative thinking and problem-solving abilities is mediated by critical thinking.

Metacognition, which refers to self-regulation and awareness of one’s own cognitive processes, is considered an important component of creative thinking. The emergence of creativity is the result of the interaction between associative, executive, and metacognitive factors [41]. For example, metacognitive thinking plays an important role in divergent thinking, particularly by stimulating creativity through factors such as language skills and emotional motivation [42]. Enhancing metacognitive strategy training in language and reading research can improve higher-order thinking skills in language and reading research, thereby influencing reading outcomes. Pre-service teachers can implement specific methods to train students’ metacognitive strategies to enhance their ability to search for problems, evaluate problems, and solve problems, which are important aspects of metacognitive thinking. Reflective courses in higher education can provide sufficient teaching scenarios, feedback, and summative assessments to cultivate students’ metacognitive and reflective thinking, enhance students’ behavioural motivation, and improve teaching effectiveness Therefore, this study proposes Hypothesis 3 to assess metacognitive thinking as a moderating variable. H3: When using generative artificial intelligence for teaching support, the development of pre-service teachers’ creativity and critical thinking is mediated by metacognitive thinking.

In summary, the widespread adoption of GAI means that simple knowledge transmission and skill training are increasingly being replaced by algorithms and code, shifting the core value of the teacher’s role from ‘knowledge transmitter’ to ‘learning facilitator’ and ‘thinking cultivator.’ This transformation makes pre-service teachers’ higher-order thinking abilities and problem-solving skills more important than ever. In this context, this study used GAI to explore the relationship between pre-service teachers’ creative thinking and problem-solving abilities and constructed a theoretical model. First, creative thinking can directly influence problem-solving abilities; critical thinking plays a fully mediating role between creative thinking and problem-solving abilities; and metacognitive thinking has a moderating effect between creative thinking and critical thinking (see Fig. 1).

Fig. 1.

A conceptual model of problem-solving abilities affecting pre-service teachers

Methods

The moderated mediation model

The Moderated Mediation Model (meMO for short) is a statistical model used to analyze the complex relationships among independent variables, dependent variables, mediating variables and moderating variables. It combines the mediating effect and the moderating effect, and can reveal how the moderating effect is indirectly achieved through mediating variables.

Sample size, participants and context

A total of 473 pre-service teachers from three different higher education institutions voluntarily participated in this study [43]. All participants were fully informed about the study and had the right to withdraw at any time during the survey process. The participants, who specialized in science, computer science, and mathematics, were between the ages of 23 and 26. Among the participants, 60% were female and 40% were male. In terms of educational background, 45% of the participants were enrolled in science-related programs, 35% in computer science programs, and 20% in mathematics programs. The study was approved by the Institutional Review Board (IRB) of [Institution Name], ensuring that all ethical guidelines were followed. Out of 473 collected scales, 19 were deemed invalid due to incomplete entries or inconsistent responses, resulting in 436 valid scales, yielding a response effectiveness rate of 96%.

Measures and procedures

The quasi-experiment is structured into five phases: pre-test, experiment, post-test, interviews, and summary. All subjects are divided into two groups, with the group using GAI-assisted teaching scenarios as the experimental group and the group not using GAI-assisted teaching as the control group.

Pre-test and experimental phase

Before the experiment began, a pre-test was conducted using the Teacher Efficacy Scale and the Advanced Thinking Scale to collect baseline data from both the experimental and control groups [44]. During the four-week experimental phase, both groups received training in teaching thinking and problem-solving skills, but only the experimental group integrated GAI-assisted instructional activities. Specifically, the experimental group engaged in tasks such as optimizing existing teaching plans with GAI, generating complete lesson plans on given topics, simulating teacher-student dialogues using GAI, and refining their instructional designs based on AI-generated feedback. These activities required them to critically evaluate AI outputs, adjust prompts, and reflect on cognitive strategies. In contrast, the control group followed traditional methods, including manual lesson planning, group discussions, textbook-based case analysis, and classroom simulations without AI tools. Both groups received equal training time and identical learning objectives, with the only variable being the use of generative AI in the experimental group.

The main tasks assigned to the experimental group over the month-long experiment include: Using GAI to optimize a given teaching design; In the second week, generating a teaching design for a specified topic using GAI and continuing teaching training with GAI.3. In the third week, simulating classroom exercises using GAI for the teacher-student interaction segment of the specified teaching design; In the fourth week, generating and optimizing teaching designs using GAI and conducting simulated classroom training. In contrast, the control group did not use GAI to assist the teaching scene, but adopted traditional classroom teaching methods and thinking methods.

Post-test and summary phase

During the post-test phase, both groups are re-assessed using the Self-Efficacy Scale and the Advanced Thinking Scale [45]. The data collected before and after the experiment are compared to analyze the causal relationships between specific variables. Additionally, interviews with approximately 50 participants are conducted to understand their experiences, feelings, and challenges encountered during the GAI-assisted teaching scenario training. This helps in gaining deeper insights into their views on using GAI for teaching and the impact of generative artificial intelligence on their thinking styles and problem-solving abilities. In the summary phase, the collected data and interview content are analyzed and coded. This analysis focuses on the impact of GAI on pre-service teachers’ thinking styles and problem-solving abilities during the teaching process. The results are then discussed in detail.

Data collection

Higher order thinking

This study constructed and revised the Higher-Order Thinking Scale based on Bloom’s taxonomy of educational objectives, synthesizing relevant theories of higher-order thinking [46]. The scale comprises four dimensions: creative thinking, critical thinking, metacognitive thinking, and problem-solving ability. Its design was informed by several classical and contemporary studies: the critical thinking dimension was primarily based on the frameworks of Ennis and Facione [47]. The creative thinking dimension drew upon the Torrance Tests of Creative Thinking and subsequent developments; the metacognitive thinking dimension referred to Flavell’s metacognition theory and the Metacognitive Awareness Inventory developed by Schraw and Dennison [48]. While the problem-solving ability dimension was grounded in the problem-solving theories proposed by Newell and Simon as well as related research in educational psychology [49]. During the item development process, multiple rounds of discussion and revision were conducted by experts in education and psychology to ensure that the items scientifically captured the essence of the four dimensions.

The first part of the scale collects basic demographic information, which is closely related to the development of higher-order thinking. The second part consists of 44 items distributed as follows: items 1–16 assess critical thinking, examining questioning, deconstruction, and dialectical thinking in the learning process; items 17–25 measure creative thinking, focusing on construction, divergence, and expansion; items 26–34 evaluate metacognitive thinking, including planning, regulation, and monitoring; and items 35–44 assess problem-solving ability. The overall reliability of the scale was high, with a Cronbach’s α of 0.931 and a KMO value of 0.871, indicating good construct validity and reliability.

Interviews and data analysis

Over the course of one month, approximately 50 pre-service teachers from the experimental group participated in individual semi-structured interviews to explore their perceptions, evaluations, and experiences related to using AI-assisted teaching, particularly with GAI. Each interview lasted between 20 and 30 min and was conducted face-to-face in a private setting to ensure a safe and open atmosphere for sharing. An interview protocol was developed in advance and included open-ended questions covering areas such as perceived effectiveness of GAI, challenges encountered, changes in thinking processes, and integration of AI tools into teaching design. All participants provided written informed consent, and interviewers followed ethical guidelines to ensure participant confidentiality and voluntary participation.

The interviews were audio-recorded and transcribed verbatim, after which the transcripts were imported into NVivo 12 for qualitative analysis. A content analysis approach was used, whereby similar codes were grouped into sub-themes (e.g., instructional efficiency, critical thinking, problem-solving). The coding process followed several stages: (1) initial open coding to capture meaningful text units, (2) categorization of codes into broader conceptual categories, and (3) iterative refinement into final themes. Two researchers independently coded the transcripts, and intercoder reliability was calculated (Cohen’s κ = 0.82), indicating substantial agreement. Discrepancies were resolved through discussion to ensure analytical rigor. A final thematic map and visual summary were generated to illustrate the qualitative findings.

In addition to qualitative analysis, quantitative data were collected separately through validated scales measuring creative thinking, critical thinking, metacognitive thinking, and problem-solving ability. These quantitative data were managed and analyzed using SPSS 24.0. Descriptive statistics, normality testing, and correlation analysis were conducted, and the PROCESS 3.0 macro (Model 4 and Model 7) developed by Hayes was applied to test mediation and moderated mediation models. The use of SPSS was thus limited to the quantitative strand of the study, not the interview data. This mixed-methods design allowed for triangulation of results: qualitative insights helped to explain and contextualize the statistical findings, thus providing a more comprehensive understanding of how generative AI influenced participants’ cognitive and instructional development.

Results

Test for common method bias

Initially, we employed Harman’s single-factor test to check for the presence of common method bias. The results indicated that the variance explained by the first factor was 24.058%, which is below the critical threshold of 40%. Thus, common method bias was not a significant issue in this study.

Descriptive statistics and correlation analysis

Next, Table 1 presents the mean values, standard deviations, and correlations among the study variables. A random coefficient regression model was used to test the hypothesis. For mediation effect testing, it is required that the mediator significantly correlates with both the independent and dependent variables. Pearson’s correlation analysis (Table 1) showed that the mediator, critical thinking, was significantly positively correlated with both the independent variable, creative thinking, and the dependent variable, problem-solving ability (p < .01), allowing for further mediation analysis. The study also explored the moderating effect of metacognitive thinking between creative thinking and problem-solving ability. Pearson’s correlation results indicated that metacognitive thinking was significantly positively correlated with creative thinking (p < .01), enabling analysis of the moderating effects.

Table 1.

Correlations & descriptive statistics

	sex	Place of Residence	Creative Thinking	Critical Thinking	Problem Solving Ability	Metacognitive Thinking
sex	1
Place of Residence	− 0.115*	1
Creative Thinking	0.159**	-0.071	1
Critical Thinking	0.083	-0.052	0.604**	1
Problem Solving Ability	0.187**	-0.041	0.428**	0.406**	1
Metacognitive Thinking	0.063	-0.078	0.406**	0.278**	0.261**	1
Mean	0.510	0.580	3.500	3.573	3.479	3.804
Std. Deviation	0.500	0.494	0.703	0.604	0.447	0.749

Testing for moderated mediation effects

Table 2 presents the results of the mediation effect tests for the hypotheses. Hypothesis 1 predicted a positive correlation between creative thinking and problem-solving abilities in pre-service teachers using generative AI (GAI). We conducted the mediation effect test using Model 7 of the PROCESS 4.1 macro, developed by Hayes, with gender and place of residence as control variables, critical thinking as the mediating variable, creative thinking as the independent variable, and problem-solving ability as the dependent variable. The results indicated a significant positive correlation between the creative thinking and problem-solving abilities of pre-service teachers using GAI (β = 0.408, t = 9.338, p < .001). This finding aligns with recent studies that demonstrate how GAI can support problem-solving by offering diverse perspectives, generating novel solutions, and enabling learners to iteratively refine their reasoning [50]. Unlike prior works that only examined general thinking patterns or focused on emotional and psychological strategies, emerging research directly links GAI use to enhanced problem-solving performance, especially when learners are guided to integrate AI-generated input into their own cognitive processes. Moreover, effective problem-solving is strongly influenced by “specific learning environments” and the cultivation of “creative thinking skills.” In the context of GAI, these environments include AI-enhanced classrooms where learners are encouraged to critically engage with AI outputs rather than passively accept them, and to leverage creative thinking skills—such as divergent idea generation and flexible perspective-taking—to adapt AI suggestions to complex, real-world problems. By situating GAI within structured yet creativity-supportive learning contexts, pre-service teachers can more effectively translate creative thinking into robust problem-solving ability. Therefore, Hypothesis 1 was supported, and the present findings extend prior research by clarifying the mechanisms through which GAI contributes to both creativity and problem-solving in teacher education.

Table 2.

Mediation model testing

variable	model 1		model 2		model 3
	Problem Solving Ability		Critical Thinking		Problem Solving Ability
	β	t	β	t	β	t
sex	0.122	2.784**	-0.014	-0.368	0.126	2.922**
Place of Residence	0.002	0.045	-0.011	-0.288	0.005	0.107
Creative Thinking	0.408	9.338***	0.605	15.552***	0.266	4.977***
Critical Thinking					0.235	4.441***
R	0.444		0.604		0.482
R2	0.197		0.365		0.232
F	35.392***		82.598		32.625

Hypothesis 2

posits that the impact of pre-service teachers’ creative thinking on their problem-solving abilities is mediated by critical thinking. This was tested using mediation effect Model 4. The results showed that the independent variable, creative thinking, had a significant positive impact on the mediator, critical thinking (β = 0.605, t = 15.552, p < .001), indicating a significant mediation effect in the first half of the path. Creative thinking (β = 0.266, t = 4.977, p < .001) and critical thinking (β = 0.235, t = 4.441, p < .001) both exerted significant positive effects on the dependent variable, problem-solving ability, thereby confirming both direct and mediating effects. The mediation effect was validated through bootstrap methods, with confidence intervals for the mediation effect, direct effect, and total effect excluding zero, indicating partial mediation by critical thinking between creative thinking and problem-solving ability. These findings highlight the pivotal mediating role of critical thinking. Notably, recent research has shown that generative AI tools such as ChatGPT can simultaneously enhance and undermine critical thinking, depending on how learners engage with them. On the one hand, ChatGPT can stimulate inquiry, argumentation, and reflection, thereby strengthening critical thinking; on the other hand, it may foster cognitive shortcuts or overreliance, leading to reduced independent judgment. Against this backdrop, the mediating role identified in the present study underscores the need to cultivate robust critical thinking skills in pre-service teachers, enabling them to harness the benefits of generative AI without succumbing to its potential drawbacks. From a practical standpoint, this suggests that teacher education programs should design targeted training methods for “new technologies and contexts.” For example, embedding structured critical reflection tasks, scenario-based evaluation of AI outputs, and guided practice in questioning and counter-argumentation can help pre-service teachers develop resilient critical thinking habits. Such strategies are essential to ensure that creative thinking translates effectively into problem-solving ability in AI-supported learning environments. Therefore, Hypothesis 2 was supported.

Table 3 presents the results of testing for moderating effects. Hypothesis 3 posited that metacognitive thinking moderates the influence of creative thinking on critical thinking among pre-service teachers using generative AI (GAI). The moderating mediation effect was assessed using Model 7. The results indicated that creative thinking (β = 0.650, t = 14.231, p < .001) and metacognitive thinking (β = 0.101, t = 2.211, p < .05) both exerted significant positive effects on critical thinking. Moreover, the interaction term between creative thinking and metacognitive thinking (β = 0.086, t = 3.354, p < .001) also had a significant positive impact on critical thinking. These findings suggest that metacognitive thinking plays a crucial moderating role in enhancing the translation of creative thinking into critical thinking.

Table 3.

Mediating effect with regulation

Variables	Model 1		Model 2
	Critical Thinking		Problem Solving Ability
	β	t	β	t
sex	-0.030	-0.386	0.251	2.922**
Place of Residence	-0.015	-0.193	0.009	0.107
Creative Thinking	0.650	14.231***	0.266	4.977 ***
Critical Thinking			0.235	4.441***
Metacognitive Thinking	0.101	2.211*
Creative Thinking* Metacognitive Thinking	0.086	3.354***
R	0.618		0.482
R2	0.382		0.232
F	53.133***		32.625***

Importantly, metacognitive thinking is not a monolithic construct but encompasses several essential processes, including self-regulation, goal-setting, and reflective monitoring. The moderating role identified in this study implies that pre-service teachers who actively engage in setting clear learning goals, monitoring their reasoning processes, and reflecting on their thought patterns are better able to transform creative ideas into well-grounded critical judgments. This result is particularly meaningful in the context of GAI usage, where overreliance on automated outputs may foster “metacognitive laziness”—a reluctance to monitor and evaluate one’s own thinking [51].Thus, the present findings highlight that cultivating strong metacognitive habits is essential for maximizing the benefits of GAI in teacher education. For educators, this means designing instructional interventions that prompt learners to articulate goals, engage in self-questioning, and reflect critically on both AI-generated and self-generated ideas. For learners, it underscores the importance of consciously regulating their engagement with GAI, ensuring that they remain active evaluators rather than passive recipients of information. Therefore, Hypothesis 3 was supported, and the study provides practical insights for mitigating risks of metacognitive complacency while enhancing higher-order thinking in AI-assisted learning environments.

Similarly, a simple effects analysis was conducted, categorizing metacognitive thinking and innovative thinking into groups based on one standard deviation above and below the mean. Further simple slope analysis indicated (see Fig. 2) that at a lower level of metacognitive thinking (M-1SD), innovative thinking had a significant positive predictive effect on critical thinking (simple slope = 0.565, t = 13.256, p < .001). At a higher level of metacognitive thinking (M + 1SD), innovative thinking also had a significant positive predictive effect on critical thinking (simple slope = 0.736, t = 12.157, p < .001). This demonstrates that as metacognitive thinking increases, the positive predictive effect of innovative thinking on critical thinking also significantly rises.

Fig. 2.

Metacognitive thinking moderates the impact of creative thinking on critical thinking

Qualitative analysis

Following the interviews, qualitative analysis was conducted to assess the application of GAI in testing the cognitive development and problem-solving abilities of pre-service teachers (Lu et al., 2024). Our findings revealed that employing GAI in higher education teaching presents a dual-edged impact. For instance, when assessing the respondents’ willingness to continue using tools such as GAI and other AI-generated Content (AIGC) for instructional training, two-thirds of the interviewees expressed their intent to persist with the use of GAI tools, while one-third indicated their reluctance to utilize GAI and similar AIGC for instructional support. This prompted an analysis of the pros and cons of employing GAI in the realm of higher education teaching, focusing on its influence on teachers’ pedagogical thinking and competencies. On one hand, the study explored the advantages of AIGC in enhancing the pedagogical thinking and problem-solving abilities of pre-service teachers, including features such as pre-set student generation, training in cognitive content, analysis of instructional materials and methodologies, and facilitation of instructional interaction. On the other hand, the study scrutinized the drawbacks associated with using GAI and other AIGC in pre-service teachers’ pedagogical thinking and problem-solving capacities, such as unexpected responses, over-reliance, and perceived inferiority compared to conventional search engines. Please refer to Fig. 3 for a comprehensive depiction.

Fig. 3.

Perceptions of pre-service teachers regarding the application of GAI

Specifically, when applying GAI and other AIGC to testing the development of pre-service teachers’ instructional thinking and problem-solving abilities, AIGC exhibits various advantages in pre-service teachers’ instructional thinking and problem-solving.

Advantages in creative thinking, as mentioned by PT-1: “We recently used GAI for instructional design, and I think it can only play a supplementary role, such as in the design of the teaching process, it may not have such innovative ideas for us, and we may still need to redesign it when introducing the scenario. I can converse smoothly with GAI, but I find that the information it provides is extracted from some articles, which may not completely match my questions.”

Advantages in critical thinking, as stated by PT-6: “I am similar to my classmates before. so I will continue to use it. On the one hand, if GAI’s version can continue to be updated and iterated, it will definitely become more intelligent. Secondly, the content of material analysis and the comparison of problem thinking provided by it can be truly referenced. In terms of teaching process, I think the simulated classroom exercises we did before are very good, which can provide great help for teachers to think about problems from the perspective of students. And I think the current use of GAI or other AI is not very clear. I think we can explore more ways to train critical thinking and creative thinking, and further develop the application of AI in teaching. For example, I did not think that AI could help us simulate classroom teaching before, but now I feel that it can do a good job in this aspect. So I think GAI has great potential and is worth continuing to use and develop.”

Advantages in problem-solving, as mentioned by PT-4: “I think where GAI helps me is in writing material analysis and student situation analysis in instructional design. It can help me integrate this part of the content, which is very convenient and saves me the trouble of searching for it myself. But other than that, it is not helpful, especially in the teaching process. For example, when I did the instructional design of ‘possibility of events’ last time, it could not recognize that the balls of the same color were the same in the problem of catching balls. I had to ask many times before it could give the correct answer, but if I wrote it myself, I could quickly write it out.”

When applying GAI and other AIGC tools to test the instructional thinking development and problem-solving abilities of pre-service teachers, there are numerous drawbacks in the use of AIGC in pre-service teachers’ instructional thinking and problem-solving. For example, in terms of the accuracy of problem-solving, PT-15 expressed, “I feel that GAI’s responses are not very satisfactory. You have to ask it questions very meticulously, and it doesn’t feel very intelligent. Moreover, the answers it gives are not necessarily correct. Especially for our computer programming code, some of the code it provides simply doesn’t work, and it’s not as good as what I find on some forums.”

Regarding the repetition rate in usage, PT-16 stated, “I probably won’t use it much. I feel that GAI’s assistance in teaching training doesn’t have any particularly strong effect. But if I sometimes need to write some ad hoc lesson plans or other summary materials, I might still use it for assistance. But I feel that GAI’s assistance in instructional design is not as convenient as finding existing, more mature lesson plans through Baidu Wenku or other channels.”

Discussion

The primary objective of this study is to examine the influence of creative thinking, critical thinking, and metacognition on pre-service teachers’ problem-solving abilities in the context of GAI. The results indicate that pre-service teachers’ creative thinking is associated with GAI, and the use of GAI positively influences their problem-solving abilities; critical thinking plays a fully mediating role in the process by which GAI influences pre-service teachers’ problem-solving abilities; and metacognitive thinking moderates the process by which creative thinking generated through GAI influences critical thinking in pre-service teachers. These three conclusions align with previous research, which has demonstrated that creative thinking effectively promotes technological development and academic achievement; new teaching models and methods can enhance critical thinking and problem-solving abilities [52]. Additionally, different thinking styles influence problem-solving abilities in distinct ways [53]. Changes in emotional and psychological thinking patterns can facilitate problem-solving and the achievement of core values and goals [54].

Secondly, in certain educational environments, higher-order thinking based on critical thinking can significantly improve students’ language abilities [55]. Advanced thinking strategies applied in educational environments can profoundly influence students’ development at higher levels [56]. Educators and learners can use new technologies and scenarios to effectively train their critical thinking skills, thereby positively influencing their problem-solving abilities. Furthermore, implementing specific learning environments to train creative thinking skills can improve participants’ academic performance [57]. The reason critical thinking enhances students’ abilities is that critical thinking involves approaching problem situations from an opposing perspective. During the critical process, participants can question problem situations from different angles, which helps individuals expand their problem-solving search space, break free from existing mental frameworks, and activate their thinking.

Furthermore, in environments where pre-service teachers use GAI to cultivate high-level creative thinking, those with stronger critical thinking skills are better equipped to solve related problems, thereby enhancing their capabilities. Specifically, metacognitive thinking combined with learning environments incorporating scenario-based and technological elements can enhance creative thinking and scientific reasoning skills [58]. Furthermore, training in various learning environments, such as problem-based learning and scenario demonstrations, can enhance advanced cognitive processes such as critical, reflective, and metacognitive thinking. For example, in multilingual learning environments, students can use GAI to simulate conversations, practising language skills by improving pronunciation and expanding vocabulary.

Finally, the results of the interview analysis indicate that most pre-service teachers expressed optimism about the application prospects of GAI in teaching during the interviews. They believe that with the continuous innovation of GAI, GAI will bring more convenience and possibilities to practical teaching. For example, through big data, teaching resources can be optimally allocated, or through the Internet of Things, an intelligent teaching environment can be constructed to enhance students’ learning experience; through GAI, personalised teaching and intelligent assessment can be achieved.

Limitations and future research

This study inevitably has limitations. Firstly, our sample was relatively small and focused specifically on pre-service teachers, primarily within computer-related fields. This concentration may mean that their cognitive levels could be higher than those in other disciplines, potentially limiting the generalizability of the results to other groups. If the study were conducted with a larger, more diverse sample from different disciplines and groups, the results might have broader applicability.

Secondly, this study employed a moderated mediation model to examine the impact of pre-service teachers’ creative thinking on problem-solving abilities under the application of GAI teaching scenarios and technology. Theoretically, this model provided evidence of a mediating effect between educators and learners shaped by technology, pedagogical methods, and contexts. Practically, it offers insights for enhancing and improving teachers’ problem-solving abilities, highlighting the significance of discussing causal relationships between different types of thinking in the training and development of cognitive strategies.

This research was conducted in accordance with the ethical principles for human participant research as outlined by the American Psychological Association, ensuring the protection of personal data and providing participants with informed consent and the option to withdraw.

Conclusion

The results indicate that pre-service teachers, under the influence of GAI technologies and environments, engage more deeply in creative thinking, thereby developing higher levels of critical thinking and enhanced problem-solving abilities. This study confirms that training pre-service teachers using GAI to foster creative thinking can elevate their critical thinking, thereby impacting their problem-solving abilities. Moreover, metacognitive thinking amplifies the impact of creative thinking on critical thinking, resulting in a moderated mediation effect.In addition, qualitative analysis was conducted to assess the application of GAI in testing the cognitive development and problem-solving abilities of pre-service teachers ( pre-service teachers).We should seek advantages and avoid disadvantages, and better play the positive role of artificial intelligence in learning space in the realm of higher education.

Abbreviations

DE

Direct Effect

ME

Mediation Effect

ME

Moderation Effect

MM

Moderated Mediation

Author contributions

‘Z.T. and L. wrote the main manuscript text, Z,T prepared Figs. 1, 2 and 3. All authors reviewed the manuscript.’

Funding

This study was funded by the second batch of the Ministry of Education of China’s Artificial Intelligence initiative(Grant No. 4045C5282300438).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Hangzhou Normal University (2024022). All procedures performed in studies involving human participants were in accordance with ethical standards of the Declaration of Helsinki. Participants have the right to voluntarily choose to participate in this study. Informed consent was obtained electronically prior to the collection of data from participants.

Consent for publication

Not applicable.