Study on the Dynamic Evolution and Influencing Factors of Teachers’ Cognitive Roles in Precision Teaching and Research

Abstract

As AI integrates into education, precision teaching and research has become core to teachers’ professional development. However, superficial discussions in such activities hinder quality improvement. Focusing on teachers’ cognitive engagement, this study explores their cognitive roles, transformation laws, and influencing factors. The study selected 92 primary school teachers. Relying on precision teaching and research activities, and using LCA, LTA, and coded data from teachers’ discussion discourses, the study identified three types of emergent cognitive roles: “Shallow Observer,” “In-depth Explorer,” and “Intermediate Responder,” along with five typical role transformation paths: “Sustained Observation,” “Late Decline to Passivity,” “Sustained Responder,” “Immediate Decline to Passivity,” and “Sustained Explorer.” Further analysis showed role evolution links to teaching seniority and data literacy: Teachers with over 10 years of teaching experience and a professional title of no less than First-Level Teacher are mostly “Sustained Observer”; meanwhile, “Sustained Explorer” have the longest seniority and highest titles; in contrast, all “Sustained Responder” have less than 10 years of teaching experience. Finally, a “dual-track empowerment” strategy is proposed: deepen teaching expertise via expert lectures and classic lesson studies; improve technology usability, strengthen data literacy training, and build a tech-empowered evidence-based system—aiming to boost precision teaching and research quality.

Keywords

precision teaching and research cognitive role transformation teachers’ professional development

Introduction

In recent years, AI-empowered precision teaching and research has become a key model to support teachers’ professional development. In such activities, AI technology is used to analyze classroom teaching data and enable collaborative activities with visualized classroom behaviors. This data support helps teachers understand their own teaching performance and students’ learning status (Chen et al., 2024; Li & Zeng, 2019; Lin & Hu, 2019). Policies at home and abroad both indicate that developing precision teaching and research activities is an effective way to promote teachers’ professional development and competence improvement. UNESCO pointed out in the Education 2030 Framework for Action that “teachers should be supported to use information technology to empower their professional development and improve education quality” (UNESCO, 2015). Additionally, the CPC Central Committee and the State Council emphasized in the Opinions on Comprehensively Deepening the Reform of Teacher Team Construction in the New Era that teaching quality should be improved through precise analysis, diagnosis and support, so as to promote teachers’ professional development (State Council of the People’s Republic of China, 2018). It is thus clear that high-quality precision teaching and research activities are of great importance to teachers’ professional development in the new era.

AI-empowered precision teaching and research enables teachers to objectively analyze classroom behaviors and generate systematic and critical discussion results. However, teacher discussions in precision teaching and research still have problems of superficiality (Wang et al., 2025). To improve the quality of teachers’ collaborative activities, we can start with the depth of teachers’ cognitive engagement (Zhou & Han, 2025). This involves clarifying the laws governing changes in teachers’ cognitive engagement during teaching and research, exploring the causes of these changes, and putting forward targeted optimization strategies that adapt to different characteristics of cognitive engagement changes. These efforts provide a basis for enhancing teachers’ cognitive engagement and achieving high-quality precision teaching and research. Role analysis is a classic method for analyzing participants’ dialogue and interaction in collaborative activities. It can accurately describe and highly condense the characteristics of teachers’ cognitive engagement in precision teaching and research. To better align with the emergent and dynamic nature of cognitive processes, this study introduces “emergent cognitive roles” to extract the key cognitive characteristics of individuals (Wu et al., 2022). Emergent roles arise spontaneously during collaborative learning, and their characteristics highly represent the discussion features of individuals. Existing studies have identified emergent roles based on fine-grained cognitive coding (Liu et al., 2025; Mao et al., 2024; Volet et al., 2017). However, most existing studies focus on static roles and ignore the complexity of longitudinal evolution. The characteristics of teachers’ cognitive engagement change as the discussion progresses, leading to changes in emergent roles. This dynamism runs through the entire process of teaching and research activities (Liu, Chang, Zhang, & et al, 2025). For example, Saqr emphasized that individuals’ cognitive behaviors in collaborative activities develop and change with different discussion stages, resulting in changes in their cognitive roles over time (Saqr & López-Pernas, 2022).

This study focuses on the issue of superficial cognitive engagement among teachers in precision teaching and research. Taking emergent cognitive roles as the unit of analysis, it integrates two dimensions—individual cognition and peer cognitive interaction—and combines theories and methods such as role analysis, Latent Class Analysis (LCA), and Latent Transition Analysis (LTA). It constructs a research framework of “cognitive role identification - dynamic role transformation - influencing factor exploration,” aiming to reveal the complete change trajectory of teachers’ cognitive engagement in precision teaching and research. Based on this, the study proposes more targeted intervention strategies to promote the deepening of cognitive engagement and the improvement of teaching and research quality, thereby providing cases for building high-level and sustainable precision teaching and research activities.

Literature Review

Teachers’ Cognitive Engagement in Teaching and Research Activities and Its Influencing Factors

Teachers’ cognitive engagement is one of the key factors driving the high-quality development of teaching and research activities. Studies have shown that the teaching and research process is an intertwined and interdependent system of “knowledge-cognition-action”: based on their own teaching experience, teachers interact with other teachers to form concrete discursive cognition, transform inherent thinking patterns, and adopt flexible action strategies (Cui & Wei, 2025; Li, 2022). The migration of this concrete cognitive development to higher levels can facilitate meaning construction in teaching and research activities (Chen et al., 2024). Existing studies consistently point out that the quality of collaborative learning is closely related to participants’ cognitive engagement patterns. Zhou et al. found that learners with different performance levels exhibit distinct cognitive engagement patterns at various stages of collaborative tasks (Zhou, 2025). In collaborative learning, there are significant differences in cognitive engagement and its phased transition sequences between high-performance groups and low-performance groups, and the evolution over time shows specific trajectories (Guo, 2024). Therefore, clarifying the laws governing changes in teachers’ cognitive engagement during collaborative teaching and research activities can give full play to teachers’ initiative and critical thinking—enabling them to combine their own experience to conduct in-depth and reflective analysis of teaching problems—thereby improving the overall effectiveness of precision teaching and research activities.

In current precision teaching and research activities, there exist issues such as insufficient teacher participation motivation and superficial, shallow discussion (Sun et al., 2025; D. Wang et al., 2025). Teachers’ insufficient data literacy and teaching seniority experiences may both lead to low cognitive engagement in teaching and research activities (Huang et al., 2025; Li & Gu, 2022; Sun & Chen, 2010; D. Wang et al., 2025), while a higher level of TPACK contributes to better cognitive engagement of teachers in these activities (M. Wang et al., 2025). Therefore, to enhance teachers’ cognitive engagement in precision teaching and research, it is necessary to take the cultivation of data literacy for all teachers as the core pathway, emphasizing that data must be deeply integrated with teachers’ educational experience and professional wisdom (D. Wang et al., 2025). Teachers’ educational experience and professional wisdom are highly associated with their teaching seniority, which also determines that teachers with different teaching seniority have distinct needs for data literacy cultivation. Teachers’ teaching experience may lead to job burnout, as well as fear of intelligent technology and cognitive gaps (Huang et al., 2025; Li & Gu, 2022; Sun & Chen, 2010). Thus, by integrating teachers’ personal characteristics such as teaching seniority and data literacy to conduct an in-depth exploration of the laws governing changes in cognitive engagement, it helps to understand how to inspire teachers to combine their own experience and conduct in-depth, reflective analysis of teaching issues.

Emergent Roles in Collaborative Discussions

In high-quality synchronous teaching and research activities, teachers exhibit distinct cognitive characteristics in discussions, including proactively providing professional content, accurately guiding the direction of discussions, injecting diverse knowledge resources, and dynamically identifying cognitive misunderstandings. Additionally, teachers engage in frequent peer interactions, analyze the adaptability of experiences from a critical perspective, and achieve the reconstruction of practical knowledge (Kourkouli, 2024). Synchronous teaching and research activities are characterized by time pressure on discussions, leading to weak logical coherence in teachers’ dialogues, yet strong interactivity (Mohammadi, 2024). Discussions revolve around “practical issues,” involving in-depth experiential exchanges; collaborative dialogue accounts for the highest proportion, and the rate of cognitive engagement is relatively high (Mulaimović et al., 2024). It can thus be seen that cognitive characteristics are explicit manifestations of teachers’ performance in teaching and research activities. It can thus be seen that cognitive characteristics are explicit manifestations of teachers’ performance in teaching and research activities. To accurately depict and highly condense the characteristics of teachers’ cognitive engagement in precision teaching and research activities, and to better align with the emergent and dynamic nature of cognitive processes, this study introduces role analysis. It leverages emergent cognitive roles to extract the key cognitive characteristics of individuals (Wu et al., 2022).

From the perspective of role generation mechanisms, roles can be divided into scripted roles and emergent roles (Liu & Yu, 2012). Different from scripted roles, which may hinder individuals from solving problems spontaneously, emergent roles arise naturally during collaborative learning, and their characteristics are highly representative of individuals’ discussion features. For example, Ouyang identified six social emergent roles based on social participation behaviors, and there are significant differences in social participation characteristics among different roles (Ouyang & Chang, 2019). Dowell distinguished emergent roles using two dimensions: participation level and interaction mode, clustering roles into six types—Chatterers, Drivers, Followers, Lurkers, Socially Detached, and Theoretically Meaningful Participants—each with distinct and unique characteristics in terms of participation level and interaction mode (Dowell et al., 2019). Liu used content analysis and cluster analysis based on multi-dimensional and process-oriented discourse data in collaborative argumentation, and identified three emergent roles in collaborative argumentation: activity guides, active constructors, and task followers (Liu, Chang, Zhang, & et al, 2025).

According to the behavioral and discourse characteristics of roles across different dimensions, roles can be categorized into cognitive roles, social roles, and other types. Cognitive roles emphasize teachers’ thinking and analysis in teaching and research activities, focusing on the depth of teachers’ understanding of teaching problems; social roles emphasize teachers’ interaction and collaboration in teaching and research activities, focusing on communication and support among teachers. These two types of roles complement each other, forming a rich set of teacher roles. Shen argued that cognitive roles explain the behavioral characteristics of individuals in cognitive dialogue, and such behavioral characteristics can effectively promote knowledge exchange and interaction (Shen, 2011). Wu identified four emergent cognitive roles—knowledge constructors, task followers, independent explorers, and lurkers—based on cognitive engagement behaviors (Wu & Ouyang, 2025).

Analysis of Cognitive Role Transformation

Cognitive emergent roles in teaching and research activities are dynamic. Cognitive engagement in collaborative learning activities changes over time, which may lead to the transformation of cognitive emergent roles. The study identified different role types, including innovative connection learners, social learners, and reflective learners, and pointed out that these roles are relatively stable in most cases, but there is also the possibility of transformation under specific conditions (Xu & Du, 2023). In response to such role transformation, Saqr proposed the concept of longitudinal roles and verified the role transformation patterns of individuals in different discussion stages (Saqr & López-Pernas, 2022). Such transformation of emergent roles may have impacts on collaborative learning activities, and the impacts vary. Zhou et al. found that there are differences in the evolution of cognitive engagement patterns among groups with different performance levels (Zhou & Ye, 2024). At the same time, Zhang explored in depth the evolution laws of five typical automatically generated roles in knowledge-building communities and explained the impacts of different types of roles on knowledge-building teaching effects (Zhang et al., 2020). Wang et al.’s study, conducted in the context of synchronous teaching and research environments, reveals a progressive transformation of teachers’ roles in teaching and research activities: starting from “Technology Learner,” moving to Preliminary Explorer, then to Problem Solver, and finally to “Teaching Practitioner” (Wang et al., 2010).

To capture the characteristics of emergent roles and their dynamic transformation, researchers have developed diversified identification methods, such as content analysis, social network analysis, and latent transition analysis. Hao used content analysis and lag sequence analysis to code and analyze discussion discourses, and explored the transformation paths between discussion roles (Hao et al., 2019). Marcos-García used content analysis and social network analysis indicators to define teachers’ fine-grained emergent roles (Marcos-García et al., 2015). Swiecki combined social network analysis and cognitive network analysis indicators to determine participants’ roles (Swiecki & Shaffer, 2020). Some scholars also use probabilistic clustering methods to explore role transformation; for example, Wu used latent transition analysis combined with latent profile analysis to explore students’ roles, role transformations, and fine-grained attributes of transformations, and further explained the characteristics of their network structures (Wu & Ouyang, 2025). To conduct multi-dimensional and fine-grained analysis of the characteristics of emergent roles, this study uses latent transition analysis, with teachers’ multi-dimensional cognitive engagement as indicators, to analyze the laws of their cognitive role transformation in precision teaching and research activities.

Based on the connotation and dynamics of cognitive emergent roles, it can be concluded that in collaborative learning activities such as precision teaching and research activities, teachers’ cognitive roles may change with the development of discussion stages. Therefore, using latent transition analysis to explore teachers’ cognitive role transformation in precision teaching and research activities helps to gain an in-depth understanding of the dynamics of teachers’ cognitive engagement in teaching and research activities, and provides theoretical basis and practical guidance for optimizing the design of teaching and research activities and promoting teachers’ professional development. In summary, this study focuses on cognitive role transformation, takes individual cognition and peer cognitive interaction as cognitive characteristics, and uses Latent Transition Analysis (LTA) to identify the emergence of cognitive roles and derive the laws of typical cognitive role transformation.

The cognitive roles in this study focus on teachers’ cognitive engagement behaviors. Combining the individual dimension and the peer dimension, emergent cognitive roles are identified based on teachers’ cognitive engagement behaviors in two dimensions—individual engagement and peer interaction—in precision teaching and research activities.

The study proposes the following research questions:

RQ1: What types of cognitive emergent roles do teachers exhibit in precision teaching and research activities?

RQ2: What transformation paths do the cognitive emergent roles of teachers present in precision teaching and research activities?

RQ3: What influencing factors lead to the different cognitive role transformations of teachers in precision teaching and research activities ?

Methodology

Research Background

To explore the different emergent cognitive roles of teachers and their transformations in precision teaching and research activities, this study organized a one-month AI-empowered precision teaching and research program. Conducted in line with the existing teaching and research rhythm of the participating schools, the activity process drew on the CDIO educational model, integrating theoretical learning with practical application and advocating that participants address practical challenges through teamwork (Li, 2024). The activities focused on the same teaching content and were carried out regularly to ensure continuity and cumulative effects in the time sequence. Each teaching and research group accumulated approximately 5 hr of study activity duration, with a total of about 80 hr of discussion time across all thematic activities. Before the activities, all participating teachers received unified special training on interpreting intelligent analysis reports. Throughout the process, the same online teaching and research platform, intelligent teaching recording system, and the AI analysis reports generated by these tools were used to enhance contextual consistency and internal validity.

Participants

In this study, 92 primary school teachers from 4 subjects were recruited in Sichuan Province. These teachers were divided into 16 teaching research groups by subject, with 5-6 teachers in each group. The characteristics of the teachers in terms of age, teaching seniority, and professional titles are presented in Table 1. Specifically, the teachers’ subject areas included primary school Chinese, mathematics, English, and science. By age group: teachers aged 18–30 accounted for 67.39%, those aged 31–40 for 29.35%, and those aged 41–50 for 3.26%. By teaching seniority: teachers with 0–5 years of teaching experience accounted for 58.70%, those with 6–10 years for 27.17%, those with 11–20 years for 11.96%, and those with over 21 years for 2.17%. By professional title: teachers without a professional title accounted for 41.30%, junior title holders for 5.43%, second-level teachers for 36.96%, first-level teachers for 15.22%, and senior teachers for 1.09%. In summary, the largest proportion of participants were unrated teachers aged 18–30 with 0–5 years of teaching experience.

Table 1.

Statistical Characteristics of Teachers in Precision Teaching and Research Activities

	Category	Number	Proportion (%)
Age	18–30 years old	62	67.39
	31–40 years old	27	29.35
	41–50 years old	3	3.26
Teaching seniority	0–5 years	54	58.70
	6–10 years	25	27.17
	11–20 years	11	11.96
	Over 21 years	2	2.17
Professional title	Unrated	38	41.30
	Junior	5	5.43
	Second-level	34	36.96
	First-level teacher	14	15.22
	Senior teacher	1	1.09

The Process of Precision Teaching and Research

The process of the teaching and research activities that each teacher participated in is shown in the figure, covering three phases: Conception, Design, and Practice. The activity design draws on the CDIO educational model. This phased, same-theme design not only ensures the integrity of teaching and research sessions but also achieves the cumulative effect of teachers’ cognition and capabilities through continuity in the time sequence. In the first phase (Conception Phase), teachers, based on the intelligent analysis reports, conducted collaborative reflection driven by both data and experience, and put forward initial ideas for teaching optimization plans. This phase focused on emphasizing a comprehensive understanding of the theme of the teaching and research activities. In the second phase (Design Phase), teachers, combined with the intelligent analysis reports, conducted in-depth discussions on the preliminary plans from the Conception Phase and transformed them into specific design plans; during this phase, teachers focused on the systematisms and feasibility of instructional designs. Before the third phase, teachers conducted on-site lesson recording, lesson observation, and evaluation in accordance with the teaching optimization plans. In the third phase (Practice Phase), teachers proposed teaching optimization plans again based on previous experience, discussed the practical implementation of the design plans with reference to the intelligent analysis reports, and further conducted research and discussion on the optimization plans. After each discussion activity, researchers conducted semi-structured interviews on the precision teaching and research activities with teachers from the discussion groups randomly (Figure 1).

Figure 1.

Process of Precision Teaching and Research Activities

Data Collection Procedures

This study recorded teachers’ discussion activities across three phases—conception, design, and practice—through video and audio. The transcribed discussion texts were used as the data source, and semantic segments were divided into meaning units, resulting in 1,5327 pieces of discourse data. First, the text content was coded in accordance with the cognitive engagement coding framework. Based on the coding results, Latent Class Analysis (LCA) was applied to determine the optimal number of classes, which served as the parameter for the number of classes in the Latent Transition Analysis model. Then, LTA was conducted on the characteristics of teachers’ cognitive engagement to obtain the latent class transitions of their cognitive roles. Through the classification and screening of these latent class transitions, the typical transformation laws of teachers’ cognitive roles were derived.

Referring to the cognitive engagement coding framework proposed by Ouyang (Ouyang & Chang, 2019), this study adjusted the framework to focus on the individual dimension and peer dimension, making it more suitable for the discussion context of this research. The revised coding framework is shown in Table 2. After finalizing the coding framework, qualitative coding of teachers’ discussion discourses was conducted based on the cognitive engagement coding framework. A random 15% of the total data was selected and independently coded by two coders familiar with the framework. The Cohen’s Kappa coefficient for cognitive engagement coding between the two coders was 0.873. Subsequently, the coders discussed areas of coding discrepancy, and once they reached a consensus, they proceeded to code the remaining data. The resulting cognitive engagement coding outcomes served as the data source for cognitive emergent roles, and the cognitive engagement characteristics in the discussion discourses represented the characteristics of teachers’ cognitive emergent roles.

Table 2.

Coding Framework for Teachers’ Cognitive Engagement in Precision Teaching and Research Activities

Dimension	Cognitive engagement	Cognitive depth	Description
Individual cognition	Individual idea	II-S	Teachers simply share information
		II-M	Teachers explore their new ideas without detailed elaboration
		II-D	Teachers elaborate on their ideas through detailed explanations or with the support of resources, statistical data or experiences
Peer cognitive interaction	Agreement/Disagreement	AG-S/DA-S	Teachers merely express agreement/disagreement with others’ viewpoints
		AG-M/DA-M	Teachers express agreement/disagreement with others’ viewpoints and further explore these viewpoints without detailed elaboration
		AG-D/DA-D	Teachers agree/disagree with others’ viewpoints through detailed elaboration and/or with the support of resources and statistical data
	Asking questions/Answering questions	AsQ-S/AnQ-S	When answering peers’ questions, teachers merely ask simple questions/simply share information
		AsQ-M/AnQ-M	When asking questions or answering peers’ questions, teachers can explore new information or new ideas without the need for detailed elaboration
		AsQ-D/AnQ-D	When asking questions/answering peers’ questions, teachers elaborate on their own ideas through detailed explanations and/or with the support of resources, statistical data, or experiences

The Analytical Methods

Latent Class Analysis

Latent Class Analysis (LCA) is a statistical technique used to analyze relationships among discrete latent variables. Its core function is to identify latent classes that can explain the relationships between observed variables, reveal the inherent structure of data, and maintain local independence. In this study, Latent Class Analysis will be adopted to classify teachers into several mutually exclusive latent classes based on the explicit variables of cognitive engagement coding in teaching research discourse. Through this method, the study aims to explore whether there are heterogeneous subgroups within the teacher group and further understand the distribution proportion of each subgroup in the total population.

The data were analyzed using the LCA method to construct a Latent Class Model (LCM). The main evaluation indicators of LCM are presented in the table below. Information evaluation indicators include three information criteria: AIC (Akaike Information Criterion), BIC (Bayesian Information Criterion), and aBIC (adjusted Bayesian Information Criterion). These three indicators adhere to a uniform evaluation criterion: the lower the indicator value, the better the model fit, indicating that dividing teacher discourse into K categories is “both well-fitted and not overly complex”. BIC is generally considered the more persuasive measure. Entropy is a key indicator for assessing classification accuracy, with a value range of 0 to 1. A larger Entropy value indicates higher classification accuracy; an Entropy value greater than 0.8 means the model’s classification accuracy is over 90%.

The Bootstrap Likelihood Ratio Test (BLRT) is often used to compare the fitting differences between a k-1 class model and a k class model. The Lo-Mendell-Rubin (LMR) likelihood ratio test indicator is used to compare the fitting conditions of models with different numbers of classes and determine whether adding one class can significantly improve model fitting. For example, if the p-value of BLRT in a 3-class LCM is significant, it indicates that the 3-class model has a better fitting effect than the 2-class model; conversely, it indicates that there is no significant difference between the 3-class LCM and the 2-class LCM. In this case, the 2-class LCM should be preferred to meet the simplicity of the model (Wang & Bi, 2018a). The evaluation criteria for LMR are the same.

In addition to testing model evaluation indicators, determining the optimal LCM also requires comprehensively considering factors such as the theoretical basis of the research question, the interpretability of the classification model, and whether the sample size of each class is reasonable (Wang & Bi, 2018b). Generally, we will not choose classes where the distribution of classes is significantly unbalanced or the sample size of a certain group is less than 5% of the total sample (van Eijk et al., 2023).

Latent Transition Analysis

LTA is an extension of LCA in longitudinal studies, aiming to estimate the dynamic transition probabilities of individuals moving between different latent classes over time. By analyzing transition probabilities, this model depicts the developmental trajectories of teachers’ cognitive roles. Specifically, the probability that a teacher remains in a specific cognitive role reflects the stability of that cognitive role in precision teaching and research activities, while the probability that a teacher transitions from one cognitive role to another reveals the developmental trends of cognitive roles.

Based on the constructed LCM, LTA estimates changes in individuals’ latent states across different time phases through a transition matrix, thereby constructing a Latent Transition Model (LTM). Therefore, the LTA method can explain the phased developmental laws of individuals from the perspective of transition probabilities. Similar to the evaluation indicators of LCM, the main evaluation indicators of LTM include AIC and BIC, and Loglikelihood (Abarda et al., 2020).

Results

RQ1: Latent Class Characteristics of Cognitive Roles

To explore the different types of emergent cognitive roles of teachers in precision teaching research activities, this study defines three time points—T1, T2, and T3—which correspond to the three phases of precision teaching and research activities: the Conception Phase, the Design Phase, and the Practice Phase. Using the 15 cognitive behaviors from the cognitive engagement coding framework as explicit response indicator variables, the frequency of each behavior exhibited by an individual teacher is compared with the average frequency of the corresponding behavior across all teachers. A frequency higher than the average indicates high cognitive engagement, while a frequency lower than the average indicates low cognitive engagement.

LCA was conducted at time points T1, T2, and T3 respectively to construct LCMs with 2 to 4 latent classes, and model fit analysis was performed. The results are shown in Table 3.

Table 3.

LCM Model Fit Indices for LCA at Time Points T1, T2, and T3

Time	Number of classes	AIC	BIC	aBIC	Entrony	LMR(p)	BLRT(p)	Probability
T1	2	1380.72	1457.868	1360.037	0.851	.016*	***	0.54/0.46
	3	1374.48	1491.44	1343.12	0.892	.019*	.05	0.45/0.18/0.37
	4	1370.887	1527.671	1328.854	0.924	.660	.18	0.24/0.35/0.23/0.18
T2	2	1166.822	1240.665	1142.911	0.938	.000*	***	0.45/0.55
	3	1150.4	1262.36	1114.15	0.955	.105	***	0.59/0.29/0.12
	4	1144.373	1294.441	1095.779	0.929	.140	.02*	0.125/0.29/0.32/0.26
T3	2	1126.764	1201.748	1103.967	0.894	.017*	***	0.21/0.79
	3	1115.73	1229.42	1081.17	0.956	.134	***	0.58/0.18/0.24
	4	1110.469	1262.856	1064.139	0.982	.283	.02*	0.57/0.13/0.23/0.07

Note. *p < .05, **p < .01, ***p < .001; p-values between 0.05 and 0.07 are considered marginally significant.

According to the evaluation of model fit indices, smaller values of AIC, BIC, and aBIC indicate better model fit, while an Entropy value closer to 1 means more accurate classification. At time point T1, compared with the 2-class model, the 3-class model showed decreased AIC and BIC values and an increased Entropy value. However, the p-values of LMR and BLRT for the 4-class model were not significant, indicating that the 3-class model was more optimal. Considering practical significance, theoretical basis, and the principle of model parsimony, the 3-class model was determined as the optimal model at time point T1. Similar patterns were observed at time points T2 and T3, where the class probabilities were evenly distributed and no class had an excessively small sample proportion. Synthesizing the above indices, the 3-class model was identified as the optimal LCM for time points T1, T2, and T3—meaning teachers’ cognitive roles in precision teaching and research activities were classified into 3 latent classes.

Based on the optimal number of classes from LCA, this study used LTA to construct a 3-class latent transition model, resulting in three cognitive roles: “Shallow Observer,” “In-depth Explorer,” and “Intermediate Responder.” Table 4 summarizes the proportion of teachers in each cognitive role and the frequency of cognitive engagement behaviors.

Table 4.

Average Frequencies and Proportions of Cognitive Characteristics Across Different Cognitive Roles

Category		Role 1: Shallow observer		Role 2: In-depth explorer		Role 3: Intermediate responder
Number		N = 166	60.14%	N = 34	12.32%	N = 76	27.54%
Individual cognition	II-S	7.27	30.89%	14.29	16.12%	19.34	19.14%
	II-M	7.49	31.81%	33.82	38.16%	37.84	37.44%
	II-D	3.72	15.82%	24.12	27.21%	10.5	10.39%
Peer cognitive interaction	AsQ-S	0.87	3.71%	5.03	5.67%	3.17	3.14%
	AsQ-M	1.38	5.86%	3.32	3.75%	9.7	9.60%
	AsQ-D	0.3	1.28%	1.82	2.06%	1.32	1.30%
	AnQ-S	0.4	1.71%	0.35	0.40%	2.03	2.00%
	AnQ-M	0.6	2.53%	0.29	0.33%	6.87	6.80%
	AnQ-D	0.47	2.00%	0.32	0.36%	1.42	1.41%
	AG-S	0.55	2.33%	1.76	1.99%	3.64	3.61%
	AG-M	0.36	1.51%	2.65	2.99%	3.7	3.66%
	AG-D	0.06	0.26%	0.65	0.73%	1.11	1.09%
	DA-S	0.02	0.10%	0.06	0.07%	0.18	0.18%
	DA-M	0.03	0.13%	0.15	0.17%	0.18	0.18%
	DA-D	0.01	0.05%	0	0.00%	0.07	0.07%

Role 1: Shallow Observer

There are 166 “Shallow Observer,” accounting for 60.14% of the total participants. During the development of reflective thinking in teaching and research activities, the frequencies of Shallow Observer’s cognitive engagement behaviors at the individual level (II-S: mean frequency = 7.27; II-M: mean frequency = 7.49; II-D: mean frequency = 3.72) and their questioning and responding behaviors at the peer level (AsQ-S: mean frequency = 0.87; AsQ-M: mean frequency = 0.3; AsQ-D: mean frequency = 0.4; AnQ-S: mean frequency = 1.38; AnQ-M: mean frequency = 0.6; AnQ-D: mean frequency = 0.47) are significantly lower than those of other role types, showing a distinct characteristic of silence. Based on their low-level performance in both individual and peer cognitive engagement, this study defines this role as “Shallow Observer.”

Role 2: In-Depth Explorer

There are 34 “In-depth Explorer,” accounting for 12.32% of the total participants. In-depth Explorer have significantly higher frequencies of deep cognitive behaviors at the individual cognitive engagement level (II-D: mean frequency = 24.12, accounting for 27.21%), and shallow questioning behaviors (AsQ-S: mean frequency = 5.03, accounting for 5.67%) and deep questioning behaviors (AsQ-D: mean frequency = 1.82, accounting for 2.06%) at the peer cognitive interaction level compared to other behavior types. In addition, this role also performs prominently in intermediate engagement behaviors, with frequencies second only to “Intermediate Responder.” Specifically, these include intermediate cognitive engagement behaviors at the individual level (II-M: mean frequency = 33.82, accounting for 38.16%), intermediate questioning behaviors at the peer level (AsQ-M: mean frequency = 3.32, accounting for 3.75%), and intermediate agreement behaviors at the peer level (AG-M: mean frequency = 2.65, accounting for 2.99%). In precision teaching and research activities, In-depth Explorer show a tendency for in-depth exploration: they often cite classics, share their own experiences and insights in depth, and demonstrate a high level of commitment and exploratory spirit towards teaching and research activities. Based on their prominent performance in deep individual cognitive behaviors and deep questioning behaviors, this study defines this role as “In-depth Explorer.”

Role 3: Intermediate Responder

There are 76 “Intermediate Responder,” accounting for 27.54% of the total participants. Intermediate Responder have the highest mean frequencies of shallow cognitive engagement (mean frequency = 19.34, accounting for 19.14%) and intermediate cognitive engagement (II-M: mean frequency = 37.84, accounting for 37.44%) at the individual level, while their deep cognitive engagement behaviors (II-D: mean frequency = 10.50, accounting for 10.39%) are at a moderate level. Meanwhile, the characteristic of actively sharing information but tending to make superficial rather than in-depth knowledge contributions is also reflected in their peer cognitive interaction behaviors. At the peer level, this role has the highest mean frequencies of response-type cognitive interaction behaviors, including shallow responses (AnQ-S: mean frequency = 2.03, accounting for 2.00%), intermediate responses (AnQ-M: mean frequency = 6.87, accounting for 6.80%), deep responses (AnQ-D: mean frequency = 1.42, accounting for 1.41%), shallow agreements (AG-S: mean frequency = 3.64, accounting for 3.61%), intermediate agreements (AG-M: mean frequency = 3.70, accounting for 3.66%), deep agreements (AG-D: mean frequency = 1.11, accounting for 1.09%), shallow disagreements (DA-S: mean frequency = 0.18, accounting for 0.18%), intermediate disagreements (DA-M: mean frequency = 0.18, accounting for 0.18%), and deep disagreements (DA-D: mean frequency = 0.07, accounting for 0.07%). Additionally, teachers in this role exhibit the most frequent intermediate questioning behaviors involving simple inquiries (AsQ-M: mean frequency = 9.70, accounting for 9.60%), but relatively fewer deep questioning behaviors that contribute the most to cognition (AsQ-D: mean frequency = 1.32, accounting for 1.60%). In precision teaching and research activities, Intermediate Responder actively share information related to teaching and research tasks: based on content shared by others, they actively ask questions, and proactively respond to, agree with, or disagree with others’ viewpoints. It can be seen that Intermediate Responder are more inclined to respond to others’ viewpoints, but perform weakly in proactively exploring task solutions or conducting in-depth analysis combined with their own experiences; their behaviors mainly focus on promoting the progress of teaching and research activities by connecting to the current discussion content. Based on their prominent characteristics in information sharing and response behaviors, this study defines this role as “Intermediate Responder.”

RQ2: Latent Class Transformation of Teachers’ Cognitive Roles

Before analyzing the transformation trends of different cognitive roles, the proportions of the three role classes at the three time points are first presented, followed by the changes among the three cognitive roles as the three discussion stages progress, derived from LTA. The specific changes are shown in Table 5 and Figure 2. From the proportions of role classes at the three time points, it can be observed that there are differences in the distribution proportions of the three cognitive roles, and teachers’ cognitive roles change as the discussion proceeds. The diagonal of the transition probability matrix represents the probability that a teacher remains in their original cognitive role, reflecting the stability of their cognitive role. Among them, “Shallow Observer” has the highest stability, with probabilities of remaining in their original role being 91% and 100% respectively; followed by “In-depth Explorer,” with probabilities of remaining in their original role being 58% and 100% respectively; the least stable are “Intermediate Responder,” with probabilities of remaining in their original role being 68% and 46% respectively. The off-diagonal elements represent the probability that a teacher with one cognitive role transitions to another cognitive role, reflecting the developmental trend of teachers’ cognitive roles. There are slight differences in the transition situations between the three time points (from T1 to T2 and from T2 to T3). From T1 to T2, “Shallow Observer” mainly remains stable, but there is a possibility of transitioning to “In-depth Explorer” (7%) and “Intermediate Responder” (3%); most “In-depth Explorer” remain stable, but a considerable proportion of teachers transition from “In-depth Explorer” to “Shallow Observer” (33%) and “Intermediate Responder” (9%); 68% of “Intermediate Responder” also maintain their own cognitive roles, but there is also a possibility of transitioning to “Shallow Observer” (25%) and “In-depth Explorer” (7%). This transition situation changes from T2 to T3: The roles of “Shallow Observer” and “In-depth Explorer” are very stable with no transitions occurring. However, teachers who were “Intermediate Responder” at T2 have poor stability—only 46% of them maintain the role of “Intermediate Responder,” 49% transition to “Shallow Observer,” and 5% transition to “In-depth Explorer.”

Table 5.

Latent State Probabilities and Latent Transition Probabilities of Cognitive Roles From T1 to T3

Time	Shallow observer	In-depth explorer	Intermediate responder
T1	0.51	0.16	0.33
T2	0.55	0.10	0.35
T3	0.74	0.11	0.15
Transition probabilities from T1 to T2
Shallow observer	0.91	0.07	0.03
In-depth explorer	0.33	0.58	0.09
Intermediate responder	0.25	0.07	0.68
Transition probabilities from T2 to T3
Shallow observer	1	0	0
In-depth explorer	0	1	0
Intermediate responder	0.49	0.05	0.46

Note. T1 refers to the 1st precision teaching and research activity, T2 refers to the 2nd precision teaching and research activity, and T3 refers to the 3rd precision teaching and research activity. In the “Transition Probabilities from T1 to T2”, the rows represent the latent states of teachers’ cognitive roles at the earlier time point (T1 for the former, T2 for the latter).

Figure 2.

Transformation Categories of Different Cognitive Roles

By combining the transition probabilities of teachers’ cognitive roles in precision teaching and research activities and the actual cognitive emergent roles of each teacher across the three discussion sessions, this study identified 9 potential cognitive role transformation categories in precision teaching and research activities, as shown in Table 6.

Table 6.

Typical Transformation Categories of Different Cognitive Roles

Role transformation	Quantity and proportion
Shallow observer-shallow observer-shallow observer	n = 41
Shallow observer-shallow observer-shallow observer	44.57%
Intermediate responder-intermediate responder-shallow observer	n = 14
	15.22%
Intermediate responder-intermediate responder-intermediate responder	n = 11
	11.96%
In-depth explorer-in-depth explorer-in-depth explorer	n = 9
In-depth explorer-in-depth explorer-in-depth explorer	9.78%
In-depth explorer-shallow observer-shallow observer	n = 6
In-depth explorer-shallow observer-shallow observer	6.52%
Intermediate responder-shallow observer-shallow observer	n = 4
Intermediate responder-shallow observer-shallow observer	4.35%
Shallow observer-intermediate responder-shallow observer	n = 3
Shallow observer-intermediate responder-shallow observer	3.26%
Shallow observer-intermediate responder-intermediate responder	n = 3
	3.26%
Intermediate responder-intermediate responder-in-depth explorer	n = 1
	1.09%

RQ3: Typical Transformation Categories of Cognitive Roles and Their Associated Characteristics

To ensure the generalizability of the study, similar cognitive role transformation categories were merged, and atypical role transformation categories with a sample size less than 5% of the total sample were excluded, resulting in five typical cognitive role transformation categories. As shown in Table 7, teachers categorized as “Sustained Observer” account for 44.57%; these teachers maintain the role of “Shallow Observer,” showing inactive participation and insufficient in-depth cognitive engagement. Teachers categorized as “Late Decline to Passivity” account for 15.22%; they were “Intermediate Responder” at T1 and T2 but became “Shallow Observer” at T3. These teachers actively responded to peer teachers and promoted teaching and research tasks in the first two discussions, but spoke less and only engaged in superficial cognitive participation in the final discussion. Teachers categorized as “Sustained Responder” account for 11.96%; they maintain the role of “Intermediate Responder,” actively responding to peers’ questions and briefly sharing their own views. Teachers categorized as “Immediate Decline to Passivity” account for 10.87%; they were either “In-depth Explorer” or “Intermediate Responder” in the first discussion but became “Shallow Observer” in both T2 and T3, showing active performance in the first discussion but inactive performance in the latter two. Teachers categorized as “Sustained Explorer” account for 9.78%; they maintain the role of “In-depth Explorer” and actively use their own experience to participate in discussions in depth.

Table 7.

Typical Transformation Categories of Different Cognitive Roles

Transformation category	Quantity & proportion	Role transformation
Category 1: Sustained observer	n = 41	Shallow observer - shallow observer - shallow observer
Category 1: Sustained observer	44.57%	Shallow observer - shallow observer - shallow observer
Category 2: Late decline to passivity	n = 14	Intermediate responder - intermediate responder - shallow observer
Category 2: Late decline to passivity	15.22%
Category 3: Sustained responder	n = 11	Intermediate responder - intermediate responder - intermediate responder
Category 3: Sustained responder	11.96%
Category 4: Immediate decline to passivity	n = 10	In-depth explorer - shallow observer - shallow observer
Category 4: Immediate decline to passivity	10.87%	Intermediate responder - shallow observer - shallow observer
Category 5: Sustained explorer	n = 9	In-depth explorer-in-depth explorer-in-depth explorer
Category 5: Sustained explorer	9.78%	In-depth explorer-in-depth explorer-in-depth explorer

To further explore the associated characteristics of teachers with different cognitive role transformations, this study conducted statistical analyses on teaching seniority and professional titles among teacher groups with different cognitive role transformations. The results are shown in Figures 3 and 4 Teachers in the “Sustained Exploration” category obviously have longer teaching seniority and higher professional titles; among teachers in the “Sustained Response” category, there are no teachers with more than 10 years of teaching seniority; and there are no significant differences in professional title characteristics among teachers in the other four categories.

Figure 3.

Distribution of Teachers’ Teaching Seniority Across Different Cognitive Role Transformations

Figure 4.

Distribution of Teachers’ Professional Titles Across Different Cognitive Role Transformations

Discussion

This study explored the characteristics of cognitive emergent roles of teachers and the categories of their cognitive role transformations by analyzing the discourse texts generated during teachers’ precision teaching and research activities. Through LCA, the number of latent classes was set to 3, and the cognitive characteristics of the texts from teachers’ teaching and research activities were clustered, resulting in 3 types of cognitive emergent roles and the probabilities of cognitive role transformations during the three discussions. Synthesizing the above analysis results, it can be concluded that three distinct cognitive roles emerge among teachers in precision teaching and research activities: “Shallow Observer,” “In-depth Explorer,” and “Intermediate Responder,” as well as five classic cognitive role transformation patterns. Furthermore, this study compared the differences in characteristics such as teaching seniority and professional titles among teachers with the five cognitive role transformations, and deeply explored the reasons for their different cognitive transformations by analyzing interview texts.

The three emergent cognitive roles reflect the typicality of cognitive engagement characteristics and enrich the features of cognitive emergent roles. The cognitive engagement characteristics of the three cognitive emergent roles are typical: “Shallow Observer” has the lowest frequency of cognitive engagement behaviors in all dimensions compared to the other two roles; “Intermediate Responder” exhibits more shallow and intermediate behaviors at the individual level, as well as more response and agreement behaviors at the peer level. These three types of roles align with the social cognitive discussion process and partially resemble the characteristics of cognitive roles identified through clustering in existing role emergence studies. In previous studies, Dowell et al. clustered roles such as Followers and Lurkers based on participants’ sociocognitive processes (Dowell et al., 2019), and Wu et al. identified four roles—Knowledge Constructors, Task Followers, Independent Explorer, and Lurkers—based on three dimensions: cognition, regulation, and social interaction (Wu & Ouyang, 2025). The characteristics of “Shallow Observer” in this study correspond to the Lurkers in existing studies: teachers in this category express less but still participate to a certain extent, rather than being completely detached from teaching and research activities. The “Intermediate Responder” share some characteristics with the Followers mentioned above: their participation depth is not the highest, but they actively respond to peer teachers. Meanwhile, this study also enriches the characteristics of cognitive emergent roles. The “In-depth Explorer” in this study had the highest frequency of deep individual cognitive engagement behaviors and deep questioning behaviors, fewer cognitive response behaviors, and a moderate frequency of shallow cognitive engagement behaviors. The characteristics of “In-depth Explorer” are inconsistent with existing studies. Previous studies have shown that learners who actively participate in social interaction have the deepest level of cognitive engagement (Ouyang & Chang, 2019). There are also studies proposing similar roles such as Independent Explorer (Wu & Ouyang, 2025); both roles tend to express their own experiences and viewpoints rather than responding to others. However, different from Independent Explorer, “In-depth Explorer” also exhibit a high frequency of deep questioning behaviors—they do not ignore interaction with peers, but are more inclined to enrich the background of problems, supplement arguments for their viewpoints, and guide peers to engage in individual cognitive participation and response behaviors through proactive interaction behaviors such as questioning. This study identifies three types of emergent cognitive roles through clustering—Shallow Observer, Intermediate Responder, and In-depth Explorer. These roles not only validate the role classification logic of the social cognitive orientation but also expand existing models through the key indicator of In-depth Explorer. Meanwhile, the categorization of these three roles provides a foundation for analyzing cognitive role transformations.

Based on the three emergent cognitive roles, this study identifies five typical patterns of cognitive role transformation. It further compares differences in characteristics—such as teaching seniority and professional title—among teachers with these five transformation patterns. By integrating existing research and conducting in-depth exploration of interview texts to investigate the causes of these different cognitive transformations, the study finds that whether teachers can overcome job burnout is the key to their evolutionary shift toward the two extremes of Sustained Explorer or Sustained Observer. Additionally, the lack of data literacy leads teachers’ cognitive roles to evolve toward Shallow Observer.

Break Through the “Teaching Seniority Trap,” Overcome Barriers of Technology Apprehension and Job Burnout, and Promote the Positive Evolution of Cognitive Roles. This study found that nearly 50% of teachers with over 10 years of teaching experience and a professional title of no less than First-Level Teacher fall into the category of “Sustained Observer”; in contrast, “Sustained Explorer”—who lead teaching and research activities—have the longest average teaching seniority, with over 80% holding formal professional titles. This correlation between teaching experience and cognitive roles creates an interesting dialogue with previous research. Chen (2017) pointed out that teachers’ teaching seniority has a negative impact on their teaching reflection and participation in professional development, which is attributed to job burnout caused by long-term teaching experience, as well as apprehension toward intelligent technology and cognitive gaps (Huang et al., 2025; Li & Gu, 2022; Sun & Chen, 2010). This study partially confirms this conclusion: most teachers with rich teaching experience are “Sustained Observer” who show weak willingness to engage cognitively, making it difficult to mobilize their enthusiasm for teaching and research activities. However, the existence of “Sustained Explorer” suggests that teaching seniority and negative attitudes are not necessarily positively correlated. “Sustained Explorer” have the longest average teaching seniority and the highest average professional title level; they focus on core teaching issues in precision teaching and research activities, demonstrating in-depth and efficient cognitive engagement with few irrelevant behaviors. It can thus be seen that overcoming technology apprehension and reversing the trajectory of burnout are key to high-quality inquiry-based teaching and research activities. It is recommended to invite teachers proficient in technology to share application scenarios in actual teaching and research, allowing “Sustained Observer” to intuitively perceive the practical value of technology. Additionally, “one-on-one technical support” can be provided: pairing technical specialists or key teachers with “Sustained Observer” to answer their questions about technology use in real time, gradually eliminating their technology anxiety and transforming them into “Sustained Explorer” who fully leverage their own experience. Further comparison reveals that all “Sustained Responder” have less than 10 years of teaching experience. They exhibit the highest frequency of cognitive engagement behaviors but fewer deep individual cognitive engagement behaviors, tending instead to respond to or agree with others’ viewpoints. These young teachers are highly motivated but lack teaching experience, making it difficult for them to internalize their experiences to participate in teaching and research activities. Therefore, for “Sustained Responder,” two measures are proposed: on the one hand, push benchmark excellent lesson cases with in-depth interpretations to help them transform practical experience; on the other hand, organize expert seminars focusing on teaching and research pain points, with supporting interactive Q&A sessions to help them update their concepts and consolidate their theoretical foundation. A comparison of cognitive characteristics shows that overcoming technology apprehension, reversing job burnout, and possessing rich teaching experience are all indispensable prerequisites for promoting teachers to maintain the role of “In-depth Explorer.”

Differences in data literacy shape micro-level disparities in the evolution of cognitive roles, leading to the phenomenon of “one report, two applications.” The Sustained Explorer fully leverages their ability to interpret data, combining intelligent analysis reports to quickly identify problems in teaching and conduct in-depth analysis—from identifying temporal discrepancies in data to pinpointing specific causes in the teaching process. For instance, Teacher K2-H noted in a dialogue during the practice phase: “By comparing the AI-generated analysis report with the teaching design, we can quickly spot inconsistencies between the teaching process and the scheduled time nodes in the design. The AI report shows that the ‘reading aloud’ segment was planned for 5 min, but it actually took 8 minutes in practice. What caused this extra 3 min?” In contrast, teachers lacking data literacy tend to resist or question intelligent analysis reports, confining their cognitive engagement to a superficial level. The Immediate Decline to Passivity faces difficulties in interpreting data during the conception phase, developing emotional resistance to data analysis reports; some even fully transfer the responsibility of teaching evaluation to intelligent tools, revealing their lack of ability to analyze specific contexts based on data indicators. For example, Teacher K4-L stated in an interview during the conception phase: “If the data in this analysis report can indicate the quality of a lesson, it should be presented to teachers directly and intuitively, rather than requiring teachers to analyze it themselves.” Teachers in the Late Decline to Passivity category express doubts about intelligent analysis reports during the practice phase, creating a conflict between objective data analysis results and subjective directions for teaching optimization. This arises because these teachers lack the ability to properly apply data—they cannot reasonably attribute indicator discrepancies, and their reasoning ends in “confusion” without forming an inquiry framework of “data → attribution → optimization.” For example, Teacher W3-H mentioned during the practice phase: “The frequency of teacher-student interaction, interaction formats, and blackboard writing duration have all decreased, but the blackboard writing content in the teaching design remains unchanged. This result is confusing.” However, blackboard writing duration and blackboard writing content are not the same concept. This comparative finding aligns with the macro conclusion: the lack of teachers’ data literacy restricts the depth of their cognitive engagement in precision teaching and research activities (Sun, 2023). When teachers have insufficient data literacy, intelligent analysis reports are expected to be “final answers” rather than “starting points for inquiry,” compressing the depth of cognitive engagement to emotional responses or superficial questions. This makes it difficult to fully serve as a link between teachers’ teaching expertise and intelligent analysis, and even harder to achieve the reflective nature of high-quality teaching and research activities. Therefore, efforts should be made in three aspects: optimizing the usability of precision teaching and research technologies to lower the threshold for teachers’ use, improving teachers’ data literacy, and establishing a technology-empowered evidence-based teaching and research system. These three aspects work together to provide support and guarantee for the implementation of high-quality teaching and research activities.

By integrating two influencing factors of cognitive engagement—teaching seniority and data literacy—this study systematically analyzes the internal mechanisms underlying the transformation of different cognitive roles. It enriches research on the influencing mechanisms of teachers’ cognitive engagement in precision teaching and research, and deepens the micro-level mechanism of how data literacy affects the evolution of cognitive roles. Additionally, it provides differentiated improvement strategies for teachers with different cognitive roles and offers a three-in-one practical pathway for constructing a high-quality precision teaching and research system.

Conclusion

Precision teaching and research activities empowered by artificial intelligence technology are an important model for supporting teachers’ professional development, but there are still problems of superficialization and shallowness. This study focuses on teachers’ cognitive roles in precision teaching and research activities, explores the cognitive emergent roles of teachers and their dynamic transformations, decomposes the characteristics of teachers’ cognitive engagement into individual cognition and peer cognitive interaction, and through role transformation analysis, identifies three cognitive emergent roles:” Shallow Observer,” “In-depth Explorer,” and “Intermediate Responder” and five cognitive role transformation patterns:” Sustained Observation,” “Late Decline to Passivity,” “Sustained Responder,” “Immediate Decline to Passivity,” and “Sustained Exploration.” Furthermore, it conducts an in-depth analysis of how teaching seniority and attitudes towards technology are associated with teachers’ cognitive role transformations. The research results show the following: First, the three cognitive emergent roles highly summarize the different cognitive characteristics of teachers in precision teaching and research activities, reflecting the typicality of cognitive emergent roles and enriching their characteristics; second, the transformation of teachers’ cognitive roles is related to their teaching seniority characteristics; third, the transformation of teachers’ cognitive roles is related to their attitudes towards technology. The study reveals the positive impact of teachers’ wisdom and technical means on promoting teachers’ cognitive engagement, and puts forward targeted suggestions: using expert lectures and in-depth study of classic lesson cases to deepen teachers’ wisdom, while attaching importance to technology usability, improving teachers’ data literacy, and forming a technology-empowered teaching and research system to change teachers’ attitudes towards technology. These two approaches work in parallel to form an effective path for high-quality precision teaching and research activities, thereby promoting teachers’ professional development.

This study has limitations in terms of sample structure and the scope of influencing factors. In the research sample, teachers with over 10 years of teaching experience account for 14%, among whom those with more than 20 years of experience make up less than 3%. As a result, the conclusions have limited generalizability to senior teachers with over 20 years of experience, restricting the external validity of the study. Secondly, the research on influencing factors has not integrated school-level teaching and research culture, meso-level collaborative atmosphere, micro-level focus of technology use, and individual-level AI-TPACK literacy into a unified analytical framework.

Future work should be expanded in three aspects: First, adopt stratified sampling to increase the proportion of teachers with long teaching seniority and senior professional titles, and incorporate dimensions of region and school stage to test the cross-context robustness of the study’s conclusions. Second, based on social network analysis and hierarchical linear modeling, integrate teaching and research culture, collaborative atmosphere, focus of technology use, and AI-TPACK into a predictive model to analyze the causal chain of role evolution. Third, introduce learning communities to conduct evidence-based intervention experiments, promoting teachers to move from marginal response to in-depth exploration, developing replicable models of precision teaching and research activities, and ultimately achieving the synergistic improvement of individual professional growth and community knowledge innovation.

Footnotes

ORCID iD

Shihan Yang

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by 融合多模态数据的信息化课堂教学交互行为识别及模式挖掘研究 (62277021).

Declaration of Conflicting Interests

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

References

Abarda

Dakkon

Azhari

Zaaloul

Khabouze

(2020). Latent transition analysis (LTA): A method for identifying differences in longitudinal change among unobserved groups. Procedia Computer Science, 170, 1116–1121. https://doi.org/10.1016/j.procs.2020.03.059

Chen

C. J.

(2017). Teachers’ professional development and its influencing factors from an international perspective: An empirical study based on TALIS data. Comparative Education Review, 39(6), 84–92.

Chen

F. J.

Zhang

G. Q.

Zhang

(2024a). Connotative characteristics, value orientation, and development paths of precise teaching and research. China Distance Education, 44(3), 68–78.

Chen

Liang

Y. M.

Cao

L. N.

(2024b). Research and practice on the cognitive development of teaching and research groups from the perspective of group learning theory. China Educational Technology, 127(11), 95–101.

Cui

H. B.

Wei

F. Y.

(2025). Conceptual model, practical dilemmas, and optimization paths of organized teaching and research. Journal of Teaching and Management(15), 52–56.

Dowell

N. M. M.

Nixon

T. M.

Graesser

A. C.

(2019). Group communication analysis: A computational linguistics approach for detecting sociocognitive roles in multiparty interactions. Behavior Research Methods, 51(3), 1007–1041. https://doi.org/10.48550/arXiv.1801.03563

Guo

(2024). Research on the characteristics of group cognitive engagement in collaborative learning. In: Proceedings of the 15th international conference on education technology and computers (ICETC '23), Spain, 26–28 September 2023 (pp. 348–354). Association for Computing Machinery. https://doi.org/10.1145/3629296.3629353

Hao

X. J.

Wang

Peng

Z. J.

(2019). Analysis of group online learning process: Dynamic transformation of learners’ roles. Modern Distance Education, 3, 38–48.

Huang

W. J.

Zhang

Z. M.

(2025). How can artificial intelligence empower teachers’ professional development: Theoretical model and practical orientation. Modern Distance Education Research, 37(1), 35–44.

10.

Kourkouli

(2024). Unlocking in-depth forum discussion and perceived effectiveness: Teaching and social presence categories in online teacher communities. Teaching and Teacher Education, 146(3), Article 104630. https://doi.org/10.1016/j.tate.2024.104630

11.

(2022). Structural dimensions and dynamic model of the transformation mechanism of educational theory into practice—A grounded theory analysis based on teacher groups. Research in Educational Development, 42(20), 8–20. https://doi.org/10.14121/j.cnki.1008-3855.2022.20.008

12.

(2024). Effect and improvement strategy of applying CDIO education model to the training of broadcasting and hosting talents in the AI era. Media, (21), 76–78.

13.

S. J.

X. Q.

(2022). What affects the teaching effect of artificial intelligence education? An analysis based on teachers’ technical cognition and teaching practice. Modern Educational Technology, 32(8), 92–99.

14.

Zeng

X. Y.

(2022). Artificial intelligence empowering high-quality development of teaching and research: Theoretical framework, practical blueprint, and development context of intelligent and precise teaching and research. China Educational Technology, 122(11), 99–107.

15.

Lin

Z. R.

X. Y.

(2019). Precise teaching and research: Data-driven improvement of teachers’ teaching and research efficiency. Digital Education, 5(6), 42–46.

16.

Liu

(2012). Role analysis and design strategies in collaborative learning. Journal of South China Normal University (Social Science Edition), 37(3), 20–25.

17.

Liu

Q. T.

Chang

Y. B.

Zhang

Miao

E. H.

X. J.

(2025). Study on the emergence of learners’ roles and knowledge construction discourse patterns in collaborative argumentative learning. E-Education Research, (2), 72–79.

18.

Liu

Q. T.

Chang

Y. B.

Zhang

G. Q.

X. J.

L. J.

(2025). Exploring the emotional interaction of emergent roles in collaborative argumentation with multidimensional and fine-grained analysis. Thinking Skills and Creativity, 58. https://doi.org/10.1016/j.tsc.2025.101902

19.

Mao

Z. Q.

X. R.

Y. Y.

(2024). Identifying emergent roles and their relationship with learning outcomes and collaborative problem-solving skills. Thinking Skills and Creativity, 54(4), Article 101642. https://doi.org/10.1016/j.tsc.2024.101642

20.

Marcos-García

J.-A.

Martínez-Monés

Dimitriadis

(2015). DESPRO: A method based on roles to provide collaboration analysis support adapted to the participants in CSCL situations. Computers & Education, 82, 335–353. https://doi.org/10.1016/j.compedu.2014.10.027

21.

Mohammadi

(2024). Teachers’ CALL professional development in synchronous, asynchronous, and bichronous online learning through project-oriented tasks: Developing CALL pedagogical knowledge. Journal of Computers in Education, 11(2), 401–422. https://doi.org/10.1007/s40692-023-00260-4

22.

Mulaimović

Lazarides

Richter

(2024). Comparing quality and engagement in face-to-face and online teacher professional development. British Journal of Educational Technology, 56(1), 61–79. https://doi.org/10.1111/bjet.13480

23.

Ouyang

Chang

Y. H.

(2019). The relationships between social participatory roles and cognitive engagement levels in online discussions. British Journal of Educational Technology, 50(3), 1396–1414. https://doi.org/10.1111/bjet.12647

24.

Saqr

López-Pernas

(2022). How CSCL roles emerge, persist, transition, and evolve over time: A four-year longitudinal study. Computers & Education, 189(3), Article 104581. https://doi.org/10.1016/j.compedu.2022.104581

25.

Shen

Y. S.

(2011). Analysis of influencing factors of CSCL collaborative roles. China Educational Technology, 27(8), 19–23.

26.

State Council of the People’s Republic of China . (2018). Opinions of the CPC central committee and the state council on comprehensively deepening the reform of teacher team construction in the new era. Chinese Government Network. https://www.gov.cn/xinwen/2018-01/31/content_5262659.htm

27.

Sun

Z. D.

Chen

(2010). Analysis of factors influencing teachers’ teaching reflection. Journal of the Chinese Society of Education, (9), 71–73.

28.

Sun

F. Q.

Zeng

X. Y.

Feng

(2025). IPTR: A new paradigm of teaching research. China Educational Technology, 48(7), 25–33.

29.

Sun

S. T.

(2023). Exploration of data-based precise teaching and research: Taking S middle school in Zhejiang Province as an example (Master’s thesis). East China Normal University.

30.

Swiecki

Shaffer

D. W.

(2020). iSENS: An integrated approach to combining epistemic and social network analyses. In Proceedings of the tenth international conference on learning analytics & knowledge, Frankfurt, Germany, 23–27 March 2020 (pp. 305–313). Association for Computing Machinery. https://doi.org/10.1145/3375462.3375505

31.

UNESCO . (2015). Incheon declaration: Education 2030: Towards inclusive and equitable quality education and lifelong learning for all. Author.

32.

Van Eijk

N. L.

Wetherall

Ferguson

O'Connor

D. B.

O'Connor

R. C.

(2023). A latent class analysis using the integrated motivational-volitional model of suicidal behaviour: Understanding suicide risk over 36 months. Journal of Affective Disorders, 336(1), 9–14. https://doi.org/10.1016/j.jad.2023.05.028

33.

Volet

Vauras

Salo

A.-E.

Khosa

(2017). Individual contributions in student-led collaborative learning: Insights from two analytical approaches to explain the quality of group outcome. Learning and Individual Differences, 53(3), 79–92. https://doi.org/10.1016/j.lindif.2016.11.006

34.

Wang

D. Q.

Chen

Z. L.

Shao

W. H.

(2025a). From “negative energy” to “empowerment”: Design of thinking chain prompts based on LLMs and construction of teaching and research AI agents—A case study of intelligent classroom teaching analysis. China Educational Technology, 125(3), 111–117.

35.

Wang

Chen

Maheshi

Gašević

(2025b). Intelligent teaching analytics for collaborative reflection: Investigating pre-service teachers’ perceptions, experiences and shared regulation processes. International Journal of Educational Technology in Higher Education, 22(1), Article 45. https://doi.org/10.1186/s41239-025-00538-w

36.

Wang

M. C.

X. Y.

(2018a). Latent variable modeling and mplus application: Advanced chapter (1st ed.). Chongqing University Press.

37.

Wang

M. C.

X. Y.

(2018b). Regression mixed model: Method progress and software implementation. Advances in Psychological Science, 26(12), 2272–2280.

38.

Wang

Y. P.

Chen

N. S.

Levy

(2010). Teacher training in a synchronous cyber face-to-face classroom: Characterizing and supporting the online teachers’ learning process. Computer Assisted Language Learning, 23(4), 277–293. https://doi.org/10.1080/09588221.2010.493523

39.

Ouyang

(2025). Using an integrated probabilistic clustering approach to detect student engagement across asynchronous and synchronous online discussions. Journal of Computing in Higher Education, 37(1), 299–326. https://doi.org/10.1007/s12528-023-09394-x

40.

Han

Huang

(2022). Identifying learners’ interaction patterns in an online learning community. International Journal of Environmental Research and Public Health, 19(4), 2245. https://doi.org/10.3390/ijerph19042245

41.

(2023). What participation types of learners are there in connectivist learning: An analysis of a cMOOC from the dual perspectives of social network and concept network characteristics. Interactive Learning Environments.

42.

Zhang

Yuan

S. L.

Zhang

Y. B.

(2020). Analysis of the evolution process of generative roles in knowledge construction communities. E-Education Research, 41(2), 53–59.

43.

Zhou

Han

(2025). Using epistemic network analysis to examine the cognitive engagement process in online collaborative learning. Acta Psychologica, 253, Article 104737. https://doi.org/10.1016/j.actpsy.2025.104737

44.

Zhou

J. M.

(2024). Investigating cognitive engagement patterns in online collaborative learning: A temporal learning analytic study. Interactive Learning Environments, 32(10), 6997–7013. https://doi.org/10.1080/10494820.2023.2299976