A Knowledge Graph-Based Approach to Enhancing Grounded Theory Research

A Knowledge Graph-Based Approach to Enhancing Grounded Theory Research | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A Knowledge Graph-Based Approach to Enhancing Grounded Theory Research Haitao Song, Qianqian Wu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6433331/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Traditional grounded theory (GT) heavily depends on researchers' cognitive abilities, making theoretical construction highly subjective and inconsistent. This reliance on individual expertise results in significant variations in research quality, limiting the interpretability, reproducibility, and scalability of GT. Recent advancements in knowledge extraction technologies have demonstrated the potential to bridge this gap by enabling non-expert researchers to achieve expert-level analytical capabilities. To address these challenges, this study proposes a computational grounded theory method based on knowledge graphs (CGT-KG), integrating knowledge graph techniques to enhance theory construction. By systematically representing multi-dimensional concept-theory relationships, CGT-KG reduces subjectivity, improves transparency, and strengthens theoretical validation. Taking the construction of a psychological structure model for efficient mathematics learning as a case, the paper verifies that the knowledge hypergraph strengthens grounded theory in three aspects: ① multi-dimensional concept-theory relationship, the knowledge hypergraph overcomes the limitations of binary relationships, making theory construction richer and more structured, and the visualized results promote knowledge sharing among researchers; ② enhanced theory verification, the color coding of nodes and edges provides a new perspective for theoretical saturation test and theoretical sampling; ③ automatic hypothesis discovery, a new theory generation framework based on analogical reasoning is proposed, making systematic exploration of new theories possible. Business and commerce/Information systems and information technology Social science/Complex networks Social science/Education Computational Grounded Theory Theory Development Knowledge Graph Knowledge Visualization Theory Construction Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1. Introduction Grounded theory has been hailed as "the most influential research paradigm in contemporary social sciences" and "the cutting edge of the qualitative research revolution"[ 1 ]. It is a methodology that combines the depth and validity advantages of qualitative research with the procedural rigor and reliability advantages of quantitative research[ 2 ][ 3 ][ 4 ]. Despite its widespread acclaim for its in-depth exploration of the real world, standardized operational procedures, and emphasis on practice, grounded theory is not without its challenges. As Suddaby (2006)[ 5 ]points out, grounded theory is not perfect or easy; it is a controversial, inclusive, and constantly evolving emerging research method. In 1968, Glaser and Strauss[ 6 ] formally proposed classical grounded theory, which was subsequently extended to disciplines such as education, psychology, and management. This method grants researchers a high degree of freedom[ 7 ], requiring a series of “subjectivities”, “predispositions”, and “personalities” throughout the research process[ 8 ]. However, when dealing with complex problems or when researchers' academic levels vary, the limitations of traditional methods that heavily rely on researchers' personal qualities become apparent, significantly hindering the promotion and application of grounded theory[ 9 ][ 10 ]. To make grounded theory accessible to a wider audience and fields, scholars began to explore ways to improve it and lower the threshold for use, enabling even non-expert researchers to construct high-quality theoretical results. To make grounded methods easier to understand and practice, Strauss and Corbin (1998) [ 11 ] proposed a procedural grounded theory with clear and detailed processes. The strictly standardized coding process strengthens the systematicness and transparency of the research. Non-expert researchers often find it difficult to sort out clues for constructing theories. In response, Charmaz (2000)[ 12 ] proposed constructivist grounded theory, which provides a clear theoretical construction guide to help scholars with different academic backgrounds construct theories. Although the above two mainstream grounded theory improvement methods have made efforts to reduce the professional requirements for users, most scholars still cannot use grounded theory freely. This is because researchers have different theoretical processing abilities[ 13 ][ 14 ]. In addition, to cope with the massive amount of qualitative text data in the era of big data, grounded research is often conducted in the form of team collaboration[ 15 ][ 16 ]. However, the members' reliance on logical reasoning in their minds during data analysis is an implicit and subjective approach that naturally isolates individuals, and the analysis process lacks preservation and sharing, which greatly compromises the reliability, intersubjective validity, and replicability of research results [ 17 ]. There is an urgent need for a new improvement approach to objectively record the research and analysis process, and to compensate as much as possible for the differences in theoretical processing ability between non-expert and expert scholars, so as to achieve the vision of "everyone can do high-quality grounded theory analysis". 2. Research Framework Mitigating the adverse impact of researchers' subjective judgments on theory construction and enhancing the accuracy, objectivity, and replicability of findings remain critical challenges in qualitative research, particularly in grounded theory studies. Scholars have explored various strategies to reduce subjectivity. Establishing clear and consistent coding criteria prior to data analysis provides researchers with a uniform interpretive framework[ 18 ]. Researchers are also encouraged to engage in continuous self-reflection and reflexive analysis throughout the research process, and to disclose any information that may influence data interpretation[ 19 ]. Collaborative coding is recognized as a method to mitigate oversight or bias in theory development and to introduce diverse perspectives[ 20 ][ 21 ]. During this process, interrater reliability (IRR) or interrater agreement (IRA) are widely used to assess the reliability and consistency of the coding process[ 22 ][ 23 ]. Furthermore, mixed methods approaches, such as triangulation involving multiple data sources and coders, have been proposed to reduce subjectivity and enhance the credibility of findings[ 24 ]. The use of computer-assisted qualitative data analysis software (CAQDAS), such as NVivo, Atlas.ti,and QDA Miner, facilitates data management and analysis for research teams[ 25 ]. While these strategies collectively contribute to enhancing the rigor of qualitative research, they largely rely on post-hoc validation to refine coding outcomes, which can be time-consuming and labor-intensive. This paper argues that enhancing researchers' theoretical processing capabilities to a consistently high level is pivotal. These capabilities encompass analytical thinking, logical reasoning, and memory. Therefore, this study proposes a computational grounded theory method based on knowledge graphs (CGT-KG), the integration of knowledge graphs as a knowledge visualization and data storage tool, leveraging computational techniques to assist in enhancing researchers' theoretical processing capabilities. The research object of grounded theory is social phenomena. Any social phenomenon has some kind of association. There is no completely isolated phenomenon or theory, and the complex relationship that connects concepts or categories is the theory that researchers expect to obtain[ 26 ], and these relationships are breeding the generation of new knowledge [ 27 ]. Knowledge graph is a powerful tool that can display the complex and intertwined relationships between knowledge units or knowledge groups. Knowledge graph uses nodes and edges to show entities and the relationships between entities, which can transform the implicit and abstract analysis process of researchers into an explicit and concrete knowledge network[ 28 ]. Through the visualized knowledge network, researchers can more intuitively understand and analyze the relationships between various concepts, improve the depth and breadth of analysis and reasoning, reduce communication costs, and make research results traceable and reproducible. Compared with traditional relational databases, the graph data structure adopted by knowledge graph has stronger knowledge representation ability and scalability, and can retain researchers' analysis ideas and the process of constructing theories. In addition, by constructing the connection between concepts and theories from a macro perspective, knowledge graph assists researchers to more efficiently observe the theoretical framework and accelerate the discovery of new theories. 3. A Knowledge Graph-Based Approach to Grounded Theory Research This paper proposes a knowledge graph-based approach to grounded theory research, introducing graph visualization into grounded theory, with knowledge graph serving as a visual tool to strengthen the results of multiple steps. Compared with traditional grounded theory research methods, the improved method adds the step of constructing a conceptual knowledge graph in the original coding stage, and also improves the axial coding and selective coding, theoretical saturation test and theoretical sampling. The specific implementation ideas are shown in Fig. 1 : (1) Defining the problem and literature review, reviewing the literature in related fields, and determining the problem to be studied; (2) Initial sampling, preliminary selection of research questions and data collection; (3) Open coding, analyzing the data and extracting concepts from it; (4) Constructing a conceptual knowledge graph, extracting the relationships between concepts and theories based on the literature, and constructing a knowledge graph based on this; (5) Axial coding and selective coding, further deepening and summarizing according to the clear presentation of relationships in the conceptual knowledge graph to form multiple categories and paradigms; (6) Theoretical saturation, if there are ambiguous relationships or independent hyperedges in the knowledge graph, it indicates that the theory is not saturated and needs to continue to supplement data; (7) Theoretical sampling, supplementing literature data for ambiguous relationships and independent hyperedges, and then returning to step 3 until a complete conceptual knowledge graph is obtained; (8) Theoretical development, combing the relationship between concepts and theories and the implicit information of the graph structure based on the conceptual knowledge graph, and mining and constructing new theories; (9) Conclusions and suggestions, comparing the newly constructed theory with existing theories, analyzing the differences and reasons, and summarizing the advantages and contributions of the new theory, etc. 3.1 Constructing a Conceptual Knowledge Graph In 1966, Kerlinger (1966) [ 29 ] proposed that "a theory is a set of related constructs (concepts), definitions, and propositions, which expresses a systematic understanding of phenomena through specific relationships between variables." This indicates that theory can be regarded as a relational expression of a set of related concepts, corresponding to independent variables, dependent variables, and the relationships between variables in the literature. This paper uses hypergraph to construct a conceptual knowledge graph to describe the above relationships, and uses the graph as a data support environment for research to participate in key links to achieve a strengthening effect. The edges in ordinary graphs can only connect two nodes, but many theories involve more than two concepts, so ordinary graphs often fail to completely describe and explain the structure and connotation of each theory in the theoretical framework. The new concept derived from hypergraph - hyperedge can connect any number of nodes, which means that hypergraph can describe more complex data relationships [ 30 ][ 31 ], express richer theoretical connotations, and can be used to solve the problems of easy edge explosion and easy loss of high-order structural information in ordinary graphs [ 32 ]. It has been proven to be able to effectively solve many complex real-world problems[ 33 ]. Taking the "Three-Component Model of Working Memory" proposed by Baddeley and Hitch (1992) [ 34 ]in 1974 as an example, this model proposes that working memory includes three parts: the central executive system, the visuospatial sketchpad, and the phonological loop. Subsequently, Baddeley (2000) [ 35 ] added a fourth element - the episodic buffer - to the original model, expanding it into a "Four-Component Model of Working Memory". The analysis of the theoretical framework is shown in Fig. 2 (a): the foundation of the theoretical framework is the four atomic concepts of phonological loop, central executive system, visuospatial sketchpad, and episodic buffer. Therefore, concept nodes are created in the bottom layer plane. The three-component model of working memory and the four-component model of working memory respectively generate the composite concepts "working memory 1" and "working memory 2". A composite concept refers to a concept whose meaning is explained by atomic concepts, and should be distributed in different layers of the graph from the atomic concepts to distinguish the hierarchical relationship. Therefore, the upper layer plane is constructed and the corresponding nodes A and B of the composite concepts are created on this plane. As shown in Fig. 2 (a), the traditional graph structure can easily describe the relationships between concepts, but it lacks the depiction of the theory itself, and cannot express the fact that a theory is composed of two or more concepts, let alone present the association between concepts and theories, and between theories. Moreover, as the facts continue to enrich, the number of edges and theoretical levels will continue to increase, which is very unfavorable for observing and analyzing the implicit information in the graph. This paper uses hypergraph to model the theoretical framework, mapping atomic concepts to nodes, and mapping composite concepts and theories to connectable hyperedges. In fact, in many cases, the composite concept and the theory that generates the concept are equivalent, so the corresponding hyperedges and nodes can be transformed into each other. As shown in Fig. 2 (b), the hypergraph structure simplifies the multiple edges between concept nodes in Fig. 2 (a) into polygons to represent the connotation of the theory and its generated composite concepts, and displays the concepts and theories on the same plane. In addition, through the intuitive graphical structure of hyperedges, the hierarchical relationship between theories and their interrelationships are clearly shown. For example, the coupling part and the difference part between theory A and theory B are clear at a glance, which promotes researchers' understanding and communication of theories. The introduction of hypergraph into theory is of great significance. It enables grounded theory research to no longer be limited to the level of connections between concepts, but to rise to the theoretical level, and can intuitively observe the existing theories, the subordinate relationship between concepts and theories, and the affinity relationship between theories from the knowledge graph. The construction of the conceptual knowledge graph is carried out synchronously with open coding. Open coding refers to the process by which researchers interpret original text data, extract assertions, and refine concepts, as shown in Table 1 . When constructing the conceptual knowledge graph, researchers need to further refine the relationship types between concepts, such as correlation, causality, time relationship, context relationship, etc., and record them in the form of triples, and then represent the theory and its composition in the form of hyperedges. This step makes the complex thinking process in the researcher's mind explicit and structured. Open coding provides conceptual entities for the conceptual knowledge graph, while the conceptual knowledge graph visualizes the coding results and complements the relationships between concepts, making the information richer and more intuitive, and easy to share. This not only helps the research team record and verify research results, but also reduces the adverse effects of individual subjective thinking differences, facilitates the team to trace and reproduce the theory construction process, and makes up for the differences in theoretical processing ability among members of the research team. Table 1 Partial results of open coding on original literature Original Literature Data Assertion Triple Involved Concept General self-efficacy is correlated with and predictive of depression levels; higher self-efficacy is associated with lower depression levels (Self-efficacy, related, depression level) Self-efficacy, depression leve Among personality traits, extraversion, conscientiousness, and agreeableness have significant effects on mathematics achievement and are good predictors of mathematics achievement (Personality traits, related, mathematics achievement) Personality and personality traits, mathematics achievement Differences in ability attribution and cognitive strategies exist between students of different genders, but have not reached a significant level; there are no significant differences in other aspects (Gender, unknown, ability attribution)(Gender, unknown, cognitive strategies) Learning attribution, gender, cognitive strategies 3.2 Axial Coding and Selective Coding In the new research method, the conceptual knowledge graph, as a data support environment, replaces the traditional open coding data foundation. In the axial coding stage, researchers need to divide numerous concepts into systematic category structures based on the hierarchical relationship between concepts in the conceptual knowledge graph. In the selective coding phase, researchers build upon the axial coding structure, integrating the research theme to analyze the intrinsic connections among categories, refine the core category, and ultimately identify a comprehensive theoretical framework that can integrate all categories. Compared with using open coding data, the conceptual knowledge graph will provide richer and more intuitive information. First, the main task of axial coding is to discover and establish various connections between concept categories to express the organic connection between various parts of the data [ 36 ]. The conceptual knowledge graph records the verifiable relationships between concepts based on literature data, providing a solid theoretical basis for axial coding to classify categories; and the graphical presentation method helps researchers to clarify complex conceptual relationships, making axial coding more systematic and clear. Secondly, in the selective coding stage, the conceptual knowledge graph further helps researchers analyze the internal connections between categories, identify core categories, and verify whether these core categories can统领 all other categories, ensuring the completeness and consistency of the theoretical framework. In addition, the conceptual knowledge graph, through its visualization characteristics, enables researchers to intuitively display and explain complex conceptual relationships and theoretical frameworks, reducing communication costs and improving team collaboration efficiency. 3.3 Theoretical Saturation Test and Theoretical Sampling Theoretical saturation test and theoretical sampling is a process of continuous iteration, dynamically adjusting the direction of researchers' data collection to ensure the depth and breadth of research results. The founders of grounded theory, Strauss and Glaser (2017) [ 37 ], first proposed the concept of theoretical saturation and defined it as "if sampling continues, no new categories or related themes will appear." The "sampling" here refers to theoretical sampling, and its connotation is that the concepts, categories, or theories being formed guide researchers in the next step of what data to collect, where to collect, etc. [ 38 ][ 39 ]. Although the traditional theoretical sampling method is systematic and comprehensive, it often faces problems such as complex data management, difficult relationship identification, and cumbersome data updates when dealing with large amounts of data and complex relationships. These challenges limit the efficiency and quality of theoretical sampling to a certain extent. In addition, color coding is also applicable to hyperedges. To distinguish composite concepts, known theories, and theories to be tested, Fig. 3 uses green, blue, and orange for identification respectively. Among them, the composite concept hyperedge is a special case of the known theory hyperedge, and the key to discovering the theory to be tested lies in finding competitive theories. Competitive theories refer to the existence of multiple theories for the same phenomenon, and researchers need to choose among these theories. The relationships between these theories can be further subdivided into three situations: mutually exclusive, complementary, and the same; relatively, there is no need for selection between non-competitive theories. In Fig. 3 , theory C and theory D constitute a set of competitive theories, and they give different explanations for the relationship between node 9 and node 11. Therefore, the union of the two theories, that is, the new hyperedge E, is the theory to be tested. Another advantage of the knowledge graph is that it can dynamically manage the relationship types between concepts and adjust the color coding of relationships in real time to share research progress information, which further enhances the flexibility and scalability of grounded theory. The underlying logic of the theoretical saturation test is based on the idea proposed by Strauss and Glaser that "various concepts should be closely intertwined to form a unified and internally connected whole." That is, there are common concepts among various theories in the same research field, and various theories rely on the common concepts as intersection points to form a concept relationship network. It is impossible for a single concept or theory to be separated from the theoretical system and become its own school. Therefore, it is possible to judge whether there is a lack of theory through the graph structure of the hypergraph. A theoretically saturated hypergraph should conform to the characteristic that "mutually connected hyperedges cover all concept nodes." Combined with the definition of theoretical saturation mentioned above, it can be known that theoretical saturation needs to satisfy two conditions: there are no ambiguous relationships in the conceptual knowledge graph and there are no independent nodes and hyperedges. 4. Experiment and Analysis This study selected a highly cited paper published in 2014, "The Psychological Structures Model of Highly Effective Mathematics Learning Based on the NVivo10 Qualitative Analysis" [40], as the benchmark. It adopts the procedural grounded theory method, which can ensure the validity of the comparison results of this study. This paper screened 47 literatures from 2000 to 2013 (hereinafter referred to as "specified years") as data according to the research theme, and used NVivo10 tool for coding; this study also used the 47 literatures mentioned in this paper as the basic data, and used knowledge graph to visualize the coding process, and finally compared the theoretical results of the two research methods. 4.2 Grounded Theory Research Example Based on Conceptual Knowledge Graph This study constructed a conceptual knowledge graph based on hypergraph, and used the relationships between concepts in it to sort out and summarize to construct theories. The results are shown in Table 2. Comparing the research results of the traditional method and the new method, it is found that the new method largely reproduces the results of the traditional method, indicating that the addition of the knowledge graph is practical and feasible. It is worth noting that for the differences in conclusions between the two methods, the new method provides a visual process of theory construction, helping researchers to trace back and analyze the reasons for these differences, making the theoretical framework more reliable and reusable. Table 2 Comparison of Psychological Models Based on NVivo and Knowledge Graph NVivo-based psychological model (traditional approach) Knowledge graph-based psychological model (novel approach) Main Category- First-level nodes Subcategory- Second-level nodes Reference points Main Category- First-level nodes Subcategory- Second-level nodes Reference points Psychological Mechanism Selective Attention 2 Psychological Mechanism Selective Attention 5 Working memory 6 Cognitive Processing （ add ） 18 Implicit learning 6 Memory Capacity （ update ） 23 Learning strategies Cognitive strategies 28 Implicit Cognition （ update ） 7 Metacognitive strategies 16 Learning strategies Cognitive strategies 25 Resource management strategies 10 Metacognitive strategies 24 Mathematical learning literacy Mathematical ability 23 Resource management strategies 18 Mathematical view 16 Self-regulated learning （ add ） 12 Learning habit 17 Mathematical learning literacy Mathematical ability 18 Metacognition Metacognitive knowledge 8 Mathematical view 5 Metacognitive monitoring 25 Learning habit 13 Non-intellectual factors Willpower 14 Metacognition Metacognitive knowledge 12 Academic emotions 4 Metacognitive monitoring 13 Personality traits 3 Non-intellectual factors Willpower 7 Learning attitude 20 Academic emotions 21 Cognitive style （ remove ） 2 Personality traits 3 Learning motivation 19 Learning attitude 16 Mathematical achievement motivation （ remove ） 15 Learning motivation 47 Influence of external factors 3 Self-awareness （ add ） 23 Influence of external factors 16 Taking the secondary node implicit cognition as an example, according to the concept-relationship triples in the conceptual knowledge graph, it can be traced back to the original text of the literature: implicit cognition refers to a kind of cognitive processing activity that does not require conscious effort to complete, mainly including unconscious perception, implicit memory and implicit learning, etc. Then, find the triples corresponding to this material, namely (unconscious perception, subclass, implicit cognition), (implicit memory, subclass, implicit cognition), (implicit learning, subclass, implicit cognition). The axial coding link to summarize the three under the implicit cognition concept is a natural result. By introducing the conceptual knowledge graph into grounded theory research, researchers can make the research process transparent, so as to achieve process traceability and result reproducibility. 4.3 Theoretical Function of Hypergraphs: Explanation and Prediction of Theories Information visualization tools have transformed the way humans perceive the world [27]. Knowledge hypergraphs, by visualizing the theoretical frameworks of specific domains, employ higher-level abstractions and intuitive graphical representations to provide in-depth insights into theoretical outcomes. They democratize access to knowledge repositories, offering ordinary scholars capabilities comparable to those of experts. This not only empowers researchers to efficiently organize and interpret existing theoretical findings but also facilitates the exploration of novel theories through the analysis of theoretical network structures. As shown in the previous educational psychology case, the hypergraph model can parse complex theoretical relationships in an intuitive and efficient way. Figure 4 shows a partial structure of the memory-related theoretical model. In this figure, each node represents an atomic concept (such as long-term memory), and each hyperedge corresponds to a theory (such as the multi-store model) or a composite concept (such as working memory). The labels on the edges indicate the relationships between nodes (such as function). The advantage of the hypergraph lies in its ability to clearly present the hierarchical, inclusive, and associative relationships between concepts and theories: Hierarchical relationship of concepts (or theories): Memory is a sub-concept of a psychological mechanism, while working memory 1, working memory 2, etc. are sub-concepts of memory. Inclusive relationship of concepts (or theories): The composite concept working memory 2 includes the central executive system, the visuospatial sketchpad, the phonological loop, and the episodic buffer. The multi-store model divides the memory system into long-term memory, short-term memory, and sensory memory. General association between concepts (or theories): The working memory multi-component model connects working memory 2 and long-term memory, connecting working memory with the multi-store model. Through such a concise hypergraph, the multi-faceted relationships between 11 atomic concepts, 3 composite concepts, and 2 theories are successfully visualized, making the originally complex and independent concepts (or theories) simple and clear. This fully demonstrates the powerful ability of hypergraphs in theoretical explanation. Analogical reasoning is an important way to generate theoretical hypotheses [41][42]. Kant once said [43]: "Whenever reason lacks a reliable line of argument, the method of analogy can often guide us forward." Therefore, this paper summarizes four methods of generating new theoretical hypotheses based on the idea of analogical reasoning, as shown in Figure 5: Analogical transfer: This is the most basic and intuitive application of the analogical reasoning method, which refers to analogy with the structure of existing theories, and trying to add new elements to construct new theories. Taking the multi-component model of working memory as an example, Baddeley proposed a new model "four-component model of working memory" by adding new components to the original model. Analogizing the graph structure reasoning and development mode of this theory, there may be new components that can develop into new theories. Deductive refinement: It is assumed that lower-level concepts will inherit some attributes of higher-level concepts, so lower-level concepts can inherit the research framework applicable to higher-level concepts. Taking long-term memory as an example, as a component of memory, long-term memory should have some basic attributes of memory, so it can inherit the analysis dimensions of memory. In this model, the analysis dimensions of memory include time limit, function, encoding and retrieval mechanism, decay and forgetting mode, neural basis, etc., then the above dimensions can be directly used to explore the various attributes of long-term memory. By searching relevant literature for different attributes and checking for research directions that are still blank, the possibility of new theories can be explored. Based on the relevant research on memory alone, the analysis of long-term memory supplements content such as classical conditional reflex effect and priming effect. This process is the formation of a new theory, and the specific results are shown in Table 3, with the bold content being the knowledge not added in the original graph. The duration dimension is temporarily empty, and there are two possibilities: First, this dimension is not applicable to the research of long-term memory; second, it is still in the theoretical exploration stage and needs further in-depth exploration. Table 3 Applying a memory research framework to long-term memory Dimension Memory capacity Long-term memory Duration Multi-store model (sensory memory, short-term memory, and long-term memory) - Function Working memory theory (working memory 1, working memory 2) Explicit memory（Episodic memory、Semantic memory）、Implicit memory (procedural memory, classical conditioning effects, priming effects) Encoding and retrieval Re-encoding、Levels of processing theory、Encoding specificity principle Semantic encoding 、 Associative encoding、Levels of processing theory、 Dual-coding theory、Recoding 、 False memory Decay and forgetting Forgetting curve、Interference theory Forgetting curve、Interference theory、Retrieval-induced forgetting Neural basis Hippocampus、Prefrontal cortex、Amygdala、Cerebellum Prefrontal cortex、Amygdala、Frontal lobe、Medial temporal lobe Storage Memory trace、Neural plasticity Unlimited capacity 、 Durable Related Concepts Relearning Short-term memory、Sensory memory、Age 、 Disease In addition, you can also choose the general system analysis method 5W1H to conduct a preliminary analysis of long-term memory. The 5W1H method provides a systematic and comprehensive framework by clarifying the six dimensions of Why (why), Who (research subject), What (what), When (when), Where (where), and How (how), which helps to explore the potential mechanisms of long-term memory from multiple perspectives. The specific results are shown in Table 4. Through this general analysis framework, extensive exploration can be carried out in the early stage of research to accumulate multi-dimensional data about long-term memory. Table 4 Examining the Research Gap in Working Memory Using the 5W1H Analysis Method 5W1H Long-Term Memory WHY (Purpose) To store and retrieve information, supporting knowledge accumulation, experience summarization, and skill formation WHO(Related Concepts) Multi-store model (sensory memory, short-term memory, and long-term memory) WHAT (Content) Implicit memory、Explicit memory WHEN (Timeline) When information transfers from short-term memory to long-term memory 、 During learning and recall WHERE (Location) Hippocampus、Cerebral cortex、Amygdala、Cerebellum, etc HOW (Mechanism) Through deep processing, review and consolidation, and emotional association to consolidate and retrieve information These two methods have their own unique advantages. The method of inheriting the memory analysis framework can conduct in-depth analysis based on existing theories, ensuring that the results are closely aligned with the research theme; while 5W1H, as a general analysis framework, is more suitable for the initial stage of research, and helps to explore various potential problems and characteristics of long-term memory from multiple perspectives. Choosing an appropriate research framework according to the development stage of the research object's field, or combining the two to complement each other, helps to discover new research dimensions of the research object, fill existing research gaps, and promote further development of the field. Table 5 Generalizing the Research Framework to the Remaining Three Concepts Framework Subject Mechanism Duration Memory capacity Multi-component model of working memory Multi-store model Selective attention Filter theory、Multiple-channel model Early selection model、 Late selection model Implicit memory Incubation effect、Implicit learning theory Implicit temporal representation Cognitive processing Information processing theory、Levels of processing theory Cognitive load theory Universal framework: It is assumed that many lower-level concepts of the same higher-level concept have similar local graph structures. Therefore, the existing research framework is extended to other concepts to discover new theories. For example, selective attention, implicit cognition, and cognitive processing and memory all belong to the category of psychological mechanisms. Therefore, analogical reasoning is performed, and the research framework of memory is applied to the other three nodes. In fact, the feasibility of this method has been confirmed. As shown in Table 5, there are indeed corresponding theories for studying the other three concepts based on the research framework of operation mechanism and time limit (time). If it is found that there is no theory when generalizing the theoretical framework, there may be two reasons: First, it does discover a new direction of research in the field; second, this attribute is not suitable for the research object. Shifted objectives: It is assumed that many lower-level concepts of the same higher-level concept have similar associations with the research subject. For example, long-term memory, sensory memory, and short-term memory belong to the same theory - the multi-store theory. There are also theories describing the association between working memory and long-term memory. Therefore, the association between working memory and the above elements can be discussed. 5. Discussion and Conclusion 5.1 Discussion Logical reasoning is crucial to the development of human intelligence. With the increasing penetration of data-driven AI models in various fields, their "black box" problem has become increasingly prominent. To ensure the reliability of AI, there is an urgent need to constrain and test the calculation process of AI through rigorous logical reasoning. Therefore, the demand for logical reasoning has also increased, and theory, as the cornerstone of logical reasoning, guides the direction of reasoning [44]. Rapidly building a solid theoretical foundation has become an urgent problem to be solved. Computational grounded theory is an efficient method for constructing a theoretical system. It integrates explanatory and predictive thinking [45], providing solid theoretical support for the development of scientific research, thereby solving the "black box" problem of AI and promoting its application in more fields. To break through the limitations of inefficient theory construction, more and more researchers have begun to turn their attention to the field of computational grounded theory. Brailas et al. (2014) [46] proposed networked grounded theory, introducing community discovery technology to cluster the coding network and automatically select the coding process; Nelson (2020)[17] formally proposed the idea of computational grounded theory, and provided a methodological framework that has both the excellent interpretive ability of human beings and the rigorous, reliable and reproducible characteristics of computer technology, which has received widespread attention from scholars in the field. If grounded research is divided into three stages: coding, inductive reasoning, and theory construction, then existing research results have significantly accelerated the tedious preparatory work in the early stage. However, how to improve the efficiency of the inductive reasoning and theory construction stages is still an urgent problem to be solved. Grounded theory automation is one way to achieve this goal, and scholars have already tried to integrate methods such as large models [47][48] and machine learning [49] into grounded theory. The knowledge graph-based computational grounded method proposed in this paper aims to use the idea of knowledge visualization to break through the limitations of existing research. With the powerful functions of analyzing and mining relationships of knowledge graph, grounded research has the ability of complex reasoning and knowledge discovery, which not only improves the efficiency of inductive sorting theory, but also finds the pattern of discovering theoretical hypotheses, and enhances the explanatory and predictive nature of grounded research theoretical results. With the conceptual knowledge graph as a data support environment, subsequent research can use natural language processing technology to further realize grounded research automation. For example, using natural language processing technology to extract assertions and process them into triples, and then visually display complex concepts and relationships in the form of hypergraphs, thereby improving coding efficiency and consistency; by introducing a reasoning engine into the knowledge graph, complex causal relationships and associations hidden in the data can be identified, thereby expanding the depth and breadth of research; the knowledge graph provides unique advantages in relationship analysis and mining, and can discover patterns and implicit relationships in the data through network analysis algorithms, thereby realizing the development and innovation of the theoretical system. 5.2 Conclusion Management theory has long faced problems such as theory lagging behind practice and uneven levels of theory construction [50]. On the one hand, the rapidly changing and complex environment has placed higher demands on management theory; on the other hand, theoretical innovation highly depends on the personal experience of researchers, resulting in long research cycles and strong subjectivity of results. The knowledge graph-based computational grounded theory method proposed in this paper provides a more efficient, transparent, and explanatory new approach for theory construction and prediction. This study selected a traditional grounded research with scientific literature as qualitative data as a benchmark, and verified the performance improvement of grounded research after introducing the knowledge graph through comparative experiments. The results show that, by using the graph structure to describe the relationship between concepts and theories, the new method strengthens the research process of grounded theory in several links. In the data analysis link, the introduction of the knowledge hypergraph realizes the construction of multi-faceted relationships between theory and concepts, successfully breaking through the limitations of binary relationships. Moreover, the visualized research results facilitate the dissemination, sharing, and reuse of knowledge, enhance the consistency of researchers' interpretation of research and analysis, and improve communication efficiency in multi-person collaborative research. In the theory construction link, different types of edges and nodes in the graph are distinguished by color coding for targeted theoretical sampling, thereby improving the validity and reliability of supplementary data and information. In the conclusion review link, the complete construction process of the theory is visually displayed, which significantly improves the explanatory power of the theory. The reproducibility and traceability also confirm the improvement of the objectivity of the research conclusions. This series of improvements has enhanced the validity, completeness, and reliability of grounded theory. As Hofman et al. (2021)[51] said, research that integrates explanatory and predictive thinking will inject new vitality into the development of disciplines. The conceptual knowledge hypergraph proposed in this study can not only deeply explain existing theories and reveal the deep reasons behind phenomena, but also predict scientific hypotheses based on four new theory discovery paths, providing direction for future research. This method is general and not only applicable to the current research field, but also expected to be extended to other disciplines.With the continuous advancement of technology and the deepening of research, computational grounded theory will surely play an increasingly important role in promoting the development of multiple disciplines Declarations Author Contribution Author contributions: Song. proposes research topics based on knowledge graphs for grounded theory, design experimental plans, revise papers and finalize them.Wu. Participated in topic selection and experimental method design, managed the data, modeled the psychological structure of efficient mathematics learning using knowledge graph-based grounded theory, validated the method's effectiveness, visualized the findings, and drafted the initial manuscript. References Charmaz, K. (2000). Constructivist and objectivist grounded theory. Handbook of qualitative research, 2, 509-535. Turner, C., & Astin, F. (2021). Grounded theory: What makes a grounded theory study? European Journal of Cardiovascular Nursing, 20(3), 285–289. Coleman, G., O’Connor, R. (2007). Using grounded theory to understand software process improvement: A study of Irish software product companies. Information and Software Technology, 49(6), 654–667. Birks, M., & Mills, J. (2015). Grounded theory: A practical guide. Sage. Suddaby, R. (2006). From the editors: What grounded theory is not. Academy of management journal, 49(4), 633-642. Glaser, Barney G., Anselm L. Strauss, and Elizabeth Strutzel. "The discovery of grounded theory; strategies for qualitative research." [J],Nursing research,1968,17(4): 364. Wagner S M , Lukassen P , Mahlendorf M .Misused and missed use — Grounded Theory and Objective Hermeneutics as methods for research in industrial marketing[J].Industrial Marketing Management, 2010, 39(1):5-15. Saldana, Johnny. 2015. The Coding Manual for Qualitative Researchers. 3rd ed. Los Angeles, CA: Sage Bryant, A., & Charmaz, K. (2010). The SAGE handbook of grounded theory: Paperback edition. Sage publications. Chun Tie, Y., Birks, M., & Francis, K. (2019). Grounded theory research: A design framework for novice researchers. SAGE Open Medicine, 7, 2050312118822927. Strauss, Anselm, and Juliet Corbin." Basics of qualitative research techniques." (1998). Charmaz, Kathy." Constructivist and objectivist grounded theory." ?Handbook of qualitative research?2 (2000): 509-535. Vollstedt, M., & Rezat, S. (2019). An introduction to grounded theory with a special focus on axial coding and the coding paradigm. Compendium for early career researchers in mathematics education, 13(1), 81-100. Jonsen, K., & Jehn, K. A. (2009). Using triangulation to validate themes in qualitative studies. Qualitative research in organizations and management: an international journal, 4(2), 123-150. Guest, G., & MacQueen, K. M. (Eds.). (2008). Handbook for team-based qualitative research. Rowman Altamira. Berente, N., & Seidel, S. (2014). Big data & inductive theory development: Towards computational Grounded Theory?. Nelson L K.Computational grounded theory: A methodological framework[J].Sociological methods & research, 2020, 49(1):3-42 Cofie, N., Braund, H., & Dalgarno, N. (2022). Eight ways to get a grip on intercoder reliability using qualitative-based measures. Canadian Medical Education Journal, 13(2), 73-76. Jones, M., & Alony, I. (2011). Guiding the use of Grounded Theory in Doctoral studies–an example from the Australian film industry. MacQueen, K. M., McLellan-Lemal, E., Bartholow, K., & Milstein, B. (2008). Team-based codebook development: Structure, process, and agreement. Handbook for team-based qualitative research, 119, 119-135. Schreier, M. (2012). Qualitative content analysis in practice. Jacobs University Bremen. McDonald, N., Schoenebeck, S., & Forte, A. (2019). Reliability and inter-rater reliability in qualitative research: Norms and guidelines for CSCW and HCI practice. Proceedings of the ACM on human-computer interaction, 3(CSCW), 1-23. Díaz, J., Pérez, J., Gallardo, C., & González-Prieto, Á. (2023). Applying inter-rater reliability and agreement in collaborative grounded theory studies in software engineering. Journal of Systems and Software, 195, 111520. Jonsen, K., & Jehn, K. A. (2009). Using triangulation to validate themes in qualitative studies. Qualitative research in organizations and management: an international journal, 4(2), 123-150. Houghton, C., Murphy, K., Meehan, B., Thomas, J., Brooker, D., & Casey, D. (2017). From screening to synthesis: using nvivo to enhance transparency in qualitative evidence synthesis. Journal of clinical nursing, 26(5-6), 873-881. Li, Z. G. (2007). The study of grounded theory in business research. In Oriental Forum (Vol. 4, pp. 90-94). Chen, Y., Chen, C. M., Liu, Z. Y., Hu, Z. G., & Wang, X. W. (2015). The methodology function of CiteSpace map** knowledge domains. Studies in Science of Science, 33(2), 242-253. Ji, S., Pan, S., Cambria, E., Marttinen, P., & Philip, S. Y. (2021). A survey on knowledge graphs: Representation, acquisition, and applications. IEEE transactions on neural networks and learning systems, 33(2), 494-514. Kerlinger, Fred Nichols." Foundations of behavioral research." [J].1966 Antelmi, A., Cordasco, G., Polato, M., Scarano, V., Spagnuolo, C., & Yang, D. (2023). A survey on hypergraph representation learning. ACM Computing Surveys, 56(1), 1-38. Zhou, D., Huang, J., & Schölkopf, B. (2006). Learning with hypergraphs: Clustering, classification, and embedding. Advances in neural information processing systems, 19. Tian, L., Zhou, X., Wu, Y. P., Zhou, W. T., Zhang, J. H., & Zhang, T. S. (2022). Knowledge graph and knowledge reasoning: A systematic review. Journal of Electronic Science and Technology, 20(2), 100159. Estrada, E., & Rodriguez-Velazquez, J. A. (2005). Complex networks as hypergraphs. arxiv preprint physics/0505137. Baddeley A.Working memory[J].Science, 1992, 255(5044):556-559 Baddeley, Alan." The episodic buffer: a new component of working memory?."[J]. Trends in cognitive sciences ,2000,4 (11): 417-423. Chen, X. (2008). The New Development of Qualitative Research and Its Significance to Social Science Research. Educational Research and Experiment, No. 2, 14-18. Glaser, Barney, and Anselm Strauss.Discovery of grounded theory: Strategies for qualitative research.[M] Routledge, 2017. Glaser, B. G., & Strauss, A. (2012). The discovery of grounded theory. Aldine Transaction. (Original work published 1967) Conlon, C., Timonen, V., Elliott-O’Dare, C., O’Keeffe, S., & Foley, G. (2020). Confused about theoretical sampling? Engaging theoretical sampling in diverse grounded theory studies. Qualitative health research, 30(6), 947-959. Guangming, W., Wenjuan, S., & ****, S. (2014). The psychological structures model of highly effective mathematics learning based on the NVivo10 qualitative analysis. Studies of Psychology and Behavior, 1(1), 1-13. Gentner, D., & Maravilla, F. (2017). Analogical reasoning. In International handbook of thinking and reasoning (pp. 186-203). Routledge. Chen, P., & Garcia, W. (2010, July). Hypothesis generation and data quality assessment through association mining. In 9th IEEE International Conference on Cognitive Informatics (ICCI'10) (pp. 659-666). IEEE. Kant, I. (1797). Allgemeine Naturgeschichte und Theorie des Himmels, oder Versuch von der Verfassung und dem mechanischen Ursprunge des ganzen Weltgebäudes: nach Newtonschen Grundsätzen abgehandelt. Liu, H., Ning, R., Teng, Z., Liu, J., Zhou, Q., & Zhang, Y. (2023). Evaluating the logical reasoning ability of chatgpt and gpt-4. arxiv preprint arxiv:2304.03439. Chen, Z., & Chen, Y. (2024). Computing grounded theory: a quantitative method to develop theories. The Journal of Chinese Sociology, 11(1), 17. Brailas, A. V. (2014). Networked grounded theory. The Qualitative Report, 19(8), 1-16. Übellacker, T. (2024). AcademiaOS: Automating Grounded Theory Development in Qualitative Research with Large Language Models. arxiv preprint arxiv:2403.08844. Zhou, Y., Yuan, Y., Huang, K., & Hu, X. (2024). Can ChatGPT perform a grounded theory approach to do risk analysis? An empirical study. Journal of Management Information Systems, 41(4), 982-1015. Muller, M., Guha, S., Baumer, E. P., Mimno, D., & Shami, N. S. (2016, November). Machine learning and grounded theory method: convergence, divergence, and combination. In Proceedings of the 2016 ACM International Conference on Supporting Group Work (pp. 3-8). Van de Ven, A. H. (2007). Engaged scholarship: A guide for organizational and social research. Oxford University Press. Hofman, J. M., Watts, D. J., Athey, S., Garip, F., Griffiths, T. L., Kleinberg, J., ... & Yarkoni, T. (2021). Integrating explanation and prediction in computational social science. Nature, 595(7866), 181-188. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6433331","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":492939130,"identity":"4536cfb4-53a1-4d24-aec8-301ea47ef7ce","order_by":0,"name":"Haitao Song","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyUlEQVRIiWNgGAWjYBACAwhlA+WyEaMFoigNiJlJ03KYBC3m8u0PHxf8Om/PPyP/AMOHssMM/LMb8GuxbOMxNp7Zdztxxo1kBsYZ5w4zSNw5QMBhx3jYpHl7bicYSCQzMPO2HWYwkEggpIX9GVDLOXuwlr/EaWEwk+b5cYBxA0gLI3FacoyNeRuSE2eceWxwsOdcOo/EDUJaDh9/+Jjnj509f3viwwc/yqzl+GcQ0AIGjG0Q+gAQ8xChHgT+EKluFIyCUTAKRiYAAK74PvaHsRg6AAAAAElFTkSuQmCC","orcid":"","institution":"South China University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Haitao","middleName":"","lastName":"Song","suffix":""},{"id":492939131,"identity":"2732f4a9-070b-47ee-8927-ef7a302f381f","order_by":1,"name":"Qianqian Wu","email":"","orcid":"","institution":"South China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Qianqian","middleName":"","lastName":"Wu","suffix":""}],"badges":[],"createdAt":"2025-04-12 08:53:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6433331/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6433331/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88340376,"identity":"052e30a0-ac38-45e1-a828-d44bed508c1b","added_by":"auto","created_at":"2025-08-05 12:38:11","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":121349,"visible":true,"origin":"","legend":"\u003cp\u003eImproved Grounded Theory Research Based on Knowledge Graph\u003c/p\u003e\n\u003cp\u003eIn the process of CGT-KG, a conceptual knowledge graph is constructed based on the open coding data, which will be used as a data foundation in the subsequent steps. The nodes and edges in the graph are also refined in the subsequent steps.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6433331/v1/c128131a4bd0fe2f0d156ef6.png"},{"id":88338050,"identity":"fa9c8444-60ea-4bfe-87f4-2080937f6cdc","added_by":"auto","created_at":"2025-08-05 12:22:11","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":195835,"visible":true,"origin":"","legend":"\u003cp\u003eHypergraphs can represent complex relationships in theories\u003c/p\u003e\n\u003cp\u003eCompared with ordinary graphs, hyperedges in hypergraphs can describe the relationship between multiple nodes. For example, working memory theory contains multiple components. And it avoids the edge explosion problem.Node A and hyperedge A are two equivalent morphisms of theory A.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6433331/v1/8d413887ef165ea428a14e93.png"},{"id":88340377,"identity":"6b66d056-60ea-4850-aa11-4f82c387082c","added_by":"auto","created_at":"2025-08-05 12:38:11","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":159719,"visible":true,"origin":"","legend":"\u003cp\u003eColor coding aids in theory saturation testing and theoretical sampling\u003c/p\u003e\n\u003cp\u003eThe relationship of different states is described by edges and hyperedges of different colors, which guides scholars to supplement the unknowable relationship and testable theory, and improves the conceptual knowledge graph.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6433331/v1/dffe9ef3315aa0f5a892064b.png"},{"id":88339847,"identity":"ae1032cb-1ba7-44fa-9c5d-a00be6c2f114","added_by":"auto","created_at":"2025-08-05 12:30:11","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":199470,"visible":true,"origin":"","legend":"\u003cp\u003eHypergraphs provide a framework for interpreting memory theories\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6433331/v1/0ae22d987e8d4a6d92d2fece.png"},{"id":88338056,"identity":"2d000d90-dbbb-47a1-a4b3-2ccc49bd5e79","added_by":"auto","created_at":"2025-08-05 12:22:11","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":218802,"visible":true,"origin":"","legend":"\u003cp\u003eFour pathways for predicting new theories using hypergraphs\u003c/p\u003e\n\u003cp\u003e①Using the structure of existing theories as an analogy to construct new theories.②Exploring new theories based on the different attributes of the research object.③Extending the applicability of the research framework to various concepts.④Discussing the impact of different factors on the research subject.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6433331/v1/313c0aa4c38ff9ea7dff2c68.png"},{"id":96913947,"identity":"28079775-17fc-415d-a2ed-4631b5d59921","added_by":"auto","created_at":"2025-11-27 14:04:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1379857,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6433331/v1/d4943283-ab40-418a-ab17-af154855fc9e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Knowledge Graph-Based Approach to Enhancing Grounded Theory Research","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eGrounded theory has been hailed as \"the most influential research paradigm in contemporary social sciences\" and \"the cutting edge of the qualitative research revolution\"[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. It is a methodology that combines the depth and validity advantages of qualitative research with the procedural rigor and reliability advantages of quantitative research[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e][\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e][\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Despite its widespread acclaim for its in-depth exploration of the real world, standardized operational procedures, and emphasis on practice, grounded theory is not without its challenges. As Suddaby (2006)[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]points out, grounded theory is not perfect or easy; it is a controversial, inclusive, and constantly evolving emerging research method. In 1968, Glaser and Strauss[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] formally proposed classical grounded theory, which was subsequently extended to disciplines such as education, psychology, and management. This method grants researchers a high degree of freedom[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], requiring a series of \u0026ldquo;subjectivities\u0026rdquo;, \u0026ldquo;predispositions\u0026rdquo;, and \u0026ldquo;personalities\u0026rdquo; throughout the research process[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. However, when dealing with complex problems or when researchers' academic levels vary, the limitations of traditional methods that heavily rely on researchers' personal qualities become apparent, significantly hindering the promotion and application of grounded theory[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e][\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eTo make grounded theory accessible to a wider audience and fields, scholars began to explore ways to improve it and lower the threshold for use, enabling even non-expert researchers to construct high-quality theoretical results. To make grounded methods easier to understand and practice, Strauss and Corbin (1998) [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] proposed a procedural grounded theory with clear and detailed processes. The strictly standardized coding process strengthens the systematicness and transparency of the research. Non-expert researchers often find it difficult to sort out clues for constructing theories. In response, Charmaz (2000)[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] proposed constructivist grounded theory, which provides a clear theoretical construction guide to help scholars with different academic backgrounds construct theories.\u003c/p\u003e\u003cp\u003eAlthough the above two mainstream grounded theory improvement methods have made efforts to reduce the professional requirements for users, most scholars still cannot use grounded theory freely. This is because researchers have different theoretical processing abilities[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e][\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. In addition, to cope with the massive amount of qualitative text data in the era of big data, grounded research is often conducted in the form of team collaboration[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e][\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. However, the members' reliance on logical reasoning in their minds during data analysis is an implicit and subjective approach that naturally isolates individuals, and the analysis process lacks preservation and sharing, which greatly compromises the reliability, intersubjective validity, and replicability of research results [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. There is an urgent need for a new improvement approach to objectively record the research and analysis process, and to compensate as much as possible for the differences in theoretical processing ability between non-expert and expert scholars, so as to achieve the vision of \"everyone can do high-quality grounded theory analysis\".\u003c/p\u003e"},{"header":"2. Research Framework","content":"\u003cp\u003eMitigating the adverse impact of researchers' subjective judgments on theory construction and enhancing the accuracy, objectivity, and replicability of findings remain critical challenges in qualitative research, particularly in grounded theory studies. Scholars have explored various strategies to reduce subjectivity. Establishing clear and consistent coding criteria prior to data analysis provides researchers with a uniform interpretive framework[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Researchers are also encouraged to engage in continuous self-reflection and reflexive analysis throughout the research process, and to disclose any information that may influence data interpretation[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Collaborative coding is recognized as a method to mitigate oversight or bias in theory development and to introduce diverse perspectives[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e][\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. During this process, interrater reliability (IRR) or interrater agreement (IRA) are widely used to assess the reliability and consistency of the coding process[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e][\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Furthermore, mixed methods approaches, such as triangulation involving multiple data sources and coders, have been proposed to reduce subjectivity and enhance the credibility of findings[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. The use of computer-assisted qualitative data analysis software (CAQDAS), such as NVivo, Atlas.ti,and QDA Miner, facilitates data management and analysis for research teams[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. While these strategies collectively contribute to enhancing the rigor of qualitative research, they largely rely on post-hoc validation to refine coding outcomes, which can be time-consuming and labor-intensive. This paper argues that enhancing researchers' theoretical processing capabilities to a consistently high level is pivotal. These capabilities encompass analytical thinking, logical reasoning, and memory. Therefore, this study proposes a computational grounded theory method based on knowledge graphs (CGT-KG), the integration of knowledge graphs as a knowledge visualization and data storage tool, leveraging computational techniques to assist in enhancing researchers' theoretical processing capabilities.\u003c/p\u003e\u003cp\u003eThe research object of grounded theory is social phenomena. Any social phenomenon has some kind of association. There is no completely isolated phenomenon or theory, and the complex relationship that connects concepts or categories is the theory that researchers expect to obtain[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e], and these relationships are breeding the generation of new knowledge [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Knowledge graph is a powerful tool that can display the complex and intertwined relationships between knowledge units or knowledge groups. Knowledge graph uses nodes and edges to show entities and the relationships between entities, which can transform the implicit and abstract analysis process of researchers into an explicit and concrete knowledge network[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Through the visualized knowledge network, researchers can more intuitively understand and analyze the relationships between various concepts, improve the depth and breadth of analysis and reasoning, reduce communication costs, and make research results traceable and reproducible. Compared with traditional relational databases, the graph data structure adopted by knowledge graph has stronger knowledge representation ability and scalability, and can retain researchers' analysis ideas and the process of constructing theories. In addition, by constructing the connection between concepts and theories from a macro perspective, knowledge graph assists researchers to more efficiently observe the theoretical framework and accelerate the discovery of new theories.\u003c/p\u003e"},{"header":"3. A Knowledge Graph-Based Approach to Grounded Theory Research","content":"\u003cp\u003eThis paper proposes a knowledge graph-based approach to grounded theory research, introducing graph visualization into grounded theory, with knowledge graph serving as a visual tool to strengthen the results of multiple steps. Compared with traditional grounded theory research methods, the improved method adds the step of constructing a conceptual knowledge graph in the original coding stage, and also improves the axial coding and selective coding, theoretical saturation test and theoretical sampling. The specific implementation ideas are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e: (1) Defining the problem and literature review, reviewing the literature in related fields, and determining the problem to be studied; (2) Initial sampling, preliminary selection of research questions and data collection; (3) Open coding, analyzing the data and extracting concepts from it; (4) Constructing a conceptual knowledge graph, extracting the relationships between concepts and theories based on the literature, and constructing a knowledge graph based on this; (5) Axial coding and selective coding, further deepening and summarizing according to the clear presentation of relationships in the conceptual knowledge graph to form multiple categories and paradigms; (6) Theoretical saturation, if there are ambiguous relationships or independent hyperedges in the knowledge graph, it indicates that the theory is not saturated and needs to continue to supplement data; (7) Theoretical sampling, supplementing literature data for ambiguous relationships and independent hyperedges, and then returning to step 3 until a complete conceptual knowledge graph is obtained; (8) Theoretical development, combing the relationship between concepts and theories and the implicit information of the graph structure based on the conceptual knowledge graph, and mining and constructing new theories; (9) Conclusions and suggestions, comparing the newly constructed theory with existing theories, analyzing the differences and reasons, and summarizing the advantages and contributions of the new theory, etc.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Constructing a Conceptual Knowledge Graph\u003c/h2\u003e\u003cp\u003eIn 1966, Kerlinger (1966) [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] proposed that \"a theory is a set of related constructs (concepts), definitions, and propositions, which expresses a systematic understanding of phenomena through specific relationships between variables.\" This indicates that theory can be regarded as a relational expression of a set of related concepts, corresponding to independent variables, dependent variables, and the relationships between variables in the literature. This paper uses hypergraph to construct a conceptual knowledge graph to describe the above relationships, and uses the graph as a data support environment for research to participate in key links to achieve a strengthening effect. The edges in ordinary graphs can only connect two nodes, but many theories involve more than two concepts, so ordinary graphs often fail to completely describe and explain the structure and connotation of each theory in the theoretical framework. The new concept derived from hypergraph - hyperedge can connect any number of nodes, which means that hypergraph can describe more complex data relationships [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e][\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], express richer theoretical connotations, and can be used to solve the problems of easy edge explosion and easy loss of high-order structural information in ordinary graphs [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. It has been proven to be able to effectively solve many complex real-world problems[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eTaking the \"Three-Component Model of Working Memory\" proposed by Baddeley and Hitch (1992) [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]in 1974 as an example, this model proposes that working memory includes three parts: the central executive system, the visuospatial sketchpad, and the phonological loop. Subsequently, Baddeley (2000) [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e] added a fourth element - the episodic buffer - to the original model, expanding it into a \"Four-Component Model of Working Memory\". The analysis of the theoretical framework is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(a): the foundation of the theoretical framework is the four atomic concepts of phonological loop, central executive system, visuospatial sketchpad, and episodic buffer. Therefore, concept nodes are created in the bottom layer plane. The three-component model of working memory and the four-component model of working memory respectively generate the composite concepts \"working memory 1\" and \"working memory 2\". A composite concept refers to a concept whose meaning is explained by atomic concepts, and should be distributed in different layers of the graph from the atomic concepts to distinguish the hierarchical relationship. Therefore, the upper layer plane is constructed and the corresponding nodes A and B of the composite concepts are created on this plane. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(a), the traditional graph structure can easily describe the relationships between concepts, but it lacks the depiction of the theory itself, and cannot express the fact that a theory is composed of two or more concepts, let alone present the association between concepts and theories, and between theories. Moreover, as the facts continue to enrich, the number of edges and theoretical levels will continue to increase, which is very unfavorable for observing and analyzing the implicit information in the graph.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThis paper uses hypergraph to model the theoretical framework, mapping atomic concepts to nodes, and mapping composite concepts and theories to connectable hyperedges. In fact, in many cases, the composite concept and the theory that generates the concept are equivalent, so the corresponding hyperedges and nodes can be transformed into each other. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(b), the hypergraph structure simplifies the multiple edges between concept nodes in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e(a) into polygons to represent the connotation of the theory and its generated composite concepts, and displays the concepts and theories on the same plane. In addition, through the intuitive graphical structure of hyperedges, the hierarchical relationship between theories and their interrelationships are clearly shown. For example, the coupling part and the difference part between theory A and theory B are clear at a glance, which promotes researchers' understanding and communication of theories.\u003c/p\u003e\u003cp\u003eThe introduction of hypergraph into theory is of great significance. It enables grounded theory research to no longer be limited to the level of connections between concepts, but to rise to the theoretical level, and can intuitively observe the existing theories, the subordinate relationship between concepts and theories, and the affinity relationship between theories from the knowledge graph.\u003c/p\u003e\u003cp\u003eThe construction of the conceptual knowledge graph is carried out synchronously with open coding. Open coding refers to the process by which researchers interpret original text data, extract assertions, and refine concepts, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. When constructing the conceptual knowledge graph, researchers need to further refine the relationship types between concepts, such as correlation, causality, time relationship, context relationship, etc., and record them in the form of triples, and then represent the theory and its composition in the form of hyperedges. This step makes the complex thinking process in the researcher's mind explicit and structured. Open coding provides conceptual entities for the conceptual knowledge graph, while the conceptual knowledge graph visualizes the coding results and complements the relationships between concepts, making the information richer and more intuitive, and easy to share. This not only helps the research team record and verify research results, but also reduces the adverse effects of individual subjective thinking differences, facilitates the team to trace and reproduce the theory construction process, and makes up for the differences in theoretical processing ability among members of the research team.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePartial results of open coding on original literature\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOriginal Literature Data\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAssertion Triple\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eInvolved Concept\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGeneral self-efficacy is correlated with and predictive of depression levels; higher self-efficacy is associated with lower depression levels\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e(Self-efficacy, related, depression level)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSelf-efficacy,\u003c/p\u003e\u003cp\u003edepression leve\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAmong personality traits, extraversion, conscientiousness, and agreeableness have significant effects on mathematics achievement and are good predictors of mathematics achievement\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e(Personality traits, related, mathematics achievement)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePersonality and personality traits, mathematics achievement\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDifferences in ability attribution and cognitive strategies exist between students of different genders, but have not reached a significant level; there are no significant differences in other aspects\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e(Gender, unknown, ability attribution)(Gender, unknown, cognitive strategies)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLearning attribution, gender, cognitive strategies\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Axial Coding and Selective Coding\u003c/h2\u003e\u003cp\u003eIn the new research method, the conceptual knowledge graph, as a data support environment, replaces the traditional open coding data foundation. In the axial coding stage, researchers need to divide numerous concepts into systematic category structures based on the hierarchical relationship between concepts in the conceptual knowledge graph. In the selective coding phase, researchers build upon the axial coding structure, integrating the research theme to analyze the intrinsic connections among categories, refine the core category, and ultimately identify a comprehensive theoretical framework that can integrate all categories.\u003c/p\u003e\u003cp\u003eCompared with using open coding data, the conceptual knowledge graph will provide richer and more intuitive information. First, the main task of axial coding is to discover and establish various connections between concept categories to express the organic connection between various parts of the data [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. The conceptual knowledge graph records the verifiable relationships between concepts based on literature data, providing a solid theoretical basis for axial coding to classify categories; and the graphical presentation method helps researchers to clarify complex conceptual relationships, making axial coding more systematic and clear. Secondly, in the selective coding stage, the conceptual knowledge graph further helps researchers analyze the internal connections between categories, identify core categories, and verify whether these core categories can统领 all other categories, ensuring the completeness and consistency of the theoretical framework. In addition, the conceptual knowledge graph, through its visualization characteristics, enables researchers to intuitively display and explain complex conceptual relationships and theoretical frameworks, reducing communication costs and improving team collaboration efficiency.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Theoretical Saturation Test and Theoretical Sampling\u003c/h2\u003e\u003cp\u003eTheoretical saturation test and theoretical sampling is a process of continuous iteration, dynamically adjusting the direction of researchers' data collection to ensure the depth and breadth of research results. The founders of grounded theory, Strauss and Glaser (2017) [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e], first proposed the concept of theoretical saturation and defined it as \"if sampling continues, no new categories or related themes will appear.\" The \"sampling\" here refers to theoretical sampling, and its connotation is that the concepts, categories, or theories being formed guide researchers in the next step of what data to collect, where to collect, etc. [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e][\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. Although the traditional theoretical sampling method is systematic and comprehensive, it often faces problems such as complex data management, difficult relationship identification, and cumbersome data updates when dealing with large amounts of data and complex relationships. These challenges limit the efficiency and quality of theoretical sampling to a certain extent.\u003c/p\u003e\u003cp\u003eIn addition, color coding is also applicable to hyperedges. To distinguish composite concepts, known theories, and theories to be tested, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e uses green, blue, and orange for identification respectively. Among them, the composite concept hyperedge is a special case of the known theory hyperedge, and the key to discovering the theory to be tested lies in finding competitive theories. Competitive theories refer to the existence of multiple theories for the same phenomenon, and researchers need to choose among these theories. The relationships between these theories can be further subdivided into three situations: mutually exclusive, complementary, and the same; relatively, there is no need for selection between non-competitive theories. In Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, theory C and theory D constitute a set of competitive theories, and they give different explanations for the relationship between node 9 and node 11. Therefore, the union of the two theories, that is, the new hyperedge E, is the theory to be tested. Another advantage of the knowledge graph is that it can dynamically manage the relationship types between concepts and adjust the color coding of relationships in real time to share research progress information, which further enhances the flexibility and scalability of grounded theory.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe underlying logic of the theoretical saturation test is based on the idea proposed by Strauss and Glaser that \"various concepts should be closely intertwined to form a unified and internally connected whole.\" That is, there are common concepts among various theories in the same research field, and various theories rely on the common concepts as intersection points to form a concept relationship network. It is impossible for a single concept or theory to be separated from the theoretical system and become its own school. Therefore, it is possible to judge whether there is a lack of theory through the graph structure of the hypergraph. A theoretically saturated hypergraph should conform to the characteristic that \"mutually connected hyperedges cover all concept nodes.\" Combined with the definition of theoretical saturation mentioned above, it can be known that theoretical saturation needs to satisfy two conditions: there are no ambiguous relationships in the conceptual knowledge graph and there are no independent nodes and hyperedges.\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Experiment and Analysis","content":"\u003cp\u003eThis study selected a highly cited paper published in 2014, \u0026quot;The Psychological Structures Model of Highly Effective Mathematics Learning Based on the NVivo10 Qualitative Analysis\u0026quot; [40], as the benchmark. It adopts the procedural grounded theory method, which can ensure the validity of the comparison results of this study. This paper screened 47 literatures from 2000 to 2013 (hereinafter referred to as \u0026quot;specified years\u0026quot;) as data according to the research theme, and used NVivo10 tool for coding; this study also used the 47 literatures mentioned in this paper as the basic data, and used knowledge graph to visualize the coding process, and finally compared the theoretical results of the two research methods.\u003c/p\u003e\n\u003ch3\u003e4.2 Grounded Theory Research Example Based on Conceptual Knowledge Graph\u003c/h3\u003e\n\u003cp\u003eThis study constructed a conceptual knowledge graph based on hypergraph, and used the relationships between concepts in it to sort out and summarize to construct theories. The results are shown in Table 2. Comparing the research results of the traditional method and the new method, it is found that the new method largely reproduces the results of the traditional method, indicating that the addition of the knowledge graph is practical and feasible. It is worth noting that for the differences in conclusions between the two methods, the new method provides a visual process of theory construction, helping researchers to trace back and analyze the reasons for these differences, making the theoretical framework more reliable and reusable.\u003c/p\u003e\n\u003cp\u003eTable 2 Comparison of Psychological Models Based on NVivo and Knowledge Graph\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" style=\"width: 352px;\"\u003e\n \u003cp\u003eNVivo-based psychological model (traditional approach)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\" style=\"width: 360px;\"\u003e\n \u003cp\u003eKnowledge graph-based psychological model (novel approach)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003eMain Category-\u003c/p\u003e\n \u003cp\u003eFirst-level nodes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eSubcategory-\u003c/p\u003e\n \u003cp\u003eSecond-level nodes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003eReference points\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 129px;\"\u003e\n \u003cp\u003eMain Category-\u003c/p\u003e\n \u003cp\u003eFirst-level nodes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eSubcategory-\u003c/p\u003e\n \u003cp\u003eSecond-level nodes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003eReference points\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" style=\"width: 130px;\"\u003e\n \u003cp\u003ePsychological Mechanism\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eSelective Attention\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" style=\"width: 129px;\"\u003e\n \u003cp\u003ePsychological Mechanism\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eSelective Attention\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eWorking memory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eCognitive Processing\u003cstrong\u003e（\u003c/strong\u003e\u003cstrong\u003eadd\u003c/strong\u003e\u003cstrong\u003e）\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003e\u0026nbsp;Implicit learning\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eMemory Capacity\u003cstrong\u003e（\u003c/strong\u003e\u003cstrong\u003eupdate\u003c/strong\u003e\u003cstrong\u003e）\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e23\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" style=\"width: 130px;\"\u003e\n \u003cp\u003eLearning strategies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eCognitive strategies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003e\u0026nbsp;Implicit Cognition\u003cstrong\u003e（\u003c/strong\u003e\u003cstrong\u003eupdate\u003c/strong\u003e\u003cstrong\u003e）\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eMetacognitive strategies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" style=\"width: 129px;\"\u003e\n \u003cp\u003eLearning strategies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eCognitive strategies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eResource management strategies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eMetacognitive strategies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"3\" style=\"width: 130px;\"\u003e\n \u003cp\u003eMathematical learning literacy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eMathematical ability\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eResource management strategies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eMathematical view\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eSelf-regulated learning\u003cstrong\u003e（\u003c/strong\u003e\u003cstrong\u003eadd\u003c/strong\u003e\u003cstrong\u003e）\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eLearning habit\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"3\" style=\"width: 129px;\"\u003e\n \u003cp\u003eMathematical learning literacy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eMathematical ability\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 130px;\"\u003e\n \u003cp\u003eMetacognition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eMetacognitive knowledge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eMathematical view\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eMetacognitive monitoring\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eLearning habit\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"8\" style=\"width: 130px;\"\u003e\n \u003cp\u003eNon-intellectual factors\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eWillpower\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 129px;\"\u003e\n \u003cp\u003eMetacognition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eMetacognitive knowledge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eAcademic emotions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eMetacognitive monitoring\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003ePersonality traits\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" style=\"width: 129px;\"\u003e\n \u003cp\u003eNon-intellectual factors\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eWillpower\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eLearning attitude\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eAcademic emotions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eCognitive style\u003cstrong\u003e（\u003c/strong\u003e\u003cstrong\u003eremove\u003c/strong\u003e\u003cstrong\u003e）\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003ePersonality traits\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eLearning motivation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eLearning attitude\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eMathematical achievement motivation\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e（\u003c/strong\u003e\u003cstrong\u003eremove\u003c/strong\u003e\u003cstrong\u003e）\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eLearning motivation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e47\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eInfluence of external factors\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eSelf-awareness\u003cstrong\u003e（\u003c/strong\u003e\u003cstrong\u003eadd\u003c/strong\u003e\u003cstrong\u003e）\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e23\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 130px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 78px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 164px;\"\u003e\n \u003cp\u003eInfluence of external factors\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eTaking the secondary node implicit cognition as an example, according to the concept-relationship triples in the conceptual knowledge graph, it can be traced back to the original text of the literature: implicit cognition refers to a kind of cognitive processing activity that does not require conscious effort to complete, mainly including unconscious perception, implicit memory and implicit learning, etc. Then, find the triples corresponding to this material, namely (unconscious perception, subclass, implicit cognition), (implicit memory, subclass, implicit cognition), (implicit learning, subclass, implicit cognition). The axial coding link to summarize the three under the implicit cognition concept is a natural result. By introducing the conceptual knowledge graph into grounded theory research, researchers can make the research process transparent, so as to achieve process traceability and result reproducibility.\u003c/p\u003e\n\u003ch3\u003e4.3 Theoretical Function of Hypergraphs: Explanation and Prediction of Theories\u003c/h3\u003e\n\u003cp\u003eInformation visualization tools have transformed the way humans perceive the world [27]. Knowledge hypergraphs, by visualizing the theoretical frameworks of specific domains, employ higher-level abstractions and intuitive graphical representations to provide in-depth insights into theoretical outcomes. They democratize access to knowledge repositories, offering ordinary scholars capabilities comparable to those of experts. This not only empowers researchers to efficiently organize and interpret existing theoretical findings but also facilitates the exploration of novel theories through the analysis of theoretical network structures.\u003c/p\u003e\n\u003cp\u003eAs shown in the previous educational psychology case, the hypergraph model can parse complex theoretical relationships in an intuitive and efficient way. Figure 4 shows a partial structure of the memory-related theoretical model. In this figure, each node represents an atomic concept (such as long-term memory), and each hyperedge corresponds to a theory (such as the multi-store model) or a composite concept (such as working memory). The labels on the edges indicate the relationships between nodes (such as function). The advantage of the hypergraph lies in its ability to clearly present the hierarchical, inclusive, and associative relationships between concepts and theories:\u003c/p\u003e\n\u003cp\u003eHierarchical relationship of concepts (or theories): Memory is a sub-concept of a psychological mechanism, while working memory 1, working memory 2, etc. are sub-concepts of memory.\u003c/p\u003e\n\u003cp\u003eInclusive relationship of concepts (or theories): The composite concept working memory 2 includes the central executive system, the visuospatial sketchpad, the phonological loop, and the episodic buffer. The multi-store model divides the memory system into long-term memory, short-term memory, and sensory memory.\u003c/p\u003e\n\u003cp\u003eGeneral association between concepts (or theories): The working memory multi-component model connects working memory 2 and long-term memory, connecting working memory with the multi-store model.\u003c/p\u003e\n\u003cp\u003eThrough such a concise hypergraph, the multi-faceted relationships between 11 atomic concepts, 3 composite concepts, and 2 theories are successfully visualized, making the originally complex and independent concepts (or theories) simple and clear. This fully demonstrates the powerful ability of hypergraphs in theoretical explanation.\u003c/p\u003e\n\u003cp\u003eAnalogical reasoning is an important way to generate theoretical hypotheses [41][42]. Kant once said [43]: \u0026quot;Whenever reason lacks a reliable line of argument, the method of analogy can often guide us forward.\u0026quot; Therefore, this paper summarizes four methods of generating new theoretical hypotheses based on the idea of analogical reasoning, as shown in Figure 5:\u003c/p\u003e\n\u003col\u003e\n \u003cli\u003e\u003cstrong\u003eAnalogical transfer:\u003c/strong\u003e This is the most basic and intuitive application of the analogical reasoning method, which refers to analogy with the structure of existing theories, and trying to add new elements to construct new theories. Taking the multi-component model of working memory as an example, Baddeley proposed a new model \u0026quot;four-component model of working memory\u0026quot; by adding new components to the original model. Analogizing the graph structure reasoning and development mode of this theory, there may be new components that can develop into new theories.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eDeductive refinement:\u003c/strong\u003e It is assumed that lower-level concepts will inherit some attributes of higher-level concepts, so lower-level concepts can inherit the research framework applicable to higher-level concepts. Taking long-term memory as an example, as a component of memory, long-term memory should have some basic attributes of memory, so it can inherit the analysis dimensions of memory. In this model, the analysis dimensions of memory include time limit, function, encoding and retrieval mechanism, decay and forgetting mode, neural basis, etc., then the above dimensions can be directly used to explore the various attributes of long-term memory. By searching relevant literature for different attributes and checking for research directions that are still blank, the possibility of new theories can be explored. Based on the relevant research on memory alone, the analysis of long-term memory supplements content such as classical conditional reflex effect and priming effect. This process is the formation of a new theory, and the specific results are shown in Table 3, with the bold content being the knowledge not added in the original graph. The duration dimension is temporarily empty, and there are two possibilities: First, this dimension is not applicable to the research of long-term memory; second, it is still in the theoretical exploration stage and needs further in-depth exploration.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eTable 3 Applying a memory research framework to long-term memory\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eDimension\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 282px;\"\u003e\n \u003cp\u003eMemory capacity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eLong-term memory\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eDuration\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 282px;\"\u003e\n \u003cp\u003eMulti-store model (sensory memory, short-term memory, and long-term memory)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e-\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eFunction\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 282px;\"\u003e\n \u003cp\u003eWorking memory theory (working memory 1, working memory 2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eExplicit memory（Episodic memory、Semantic memory）、Implicit memory (procedural memory, classical conditioning effects, priming effects)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eEncoding and retrieval\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 282px;\"\u003e\n \u003cp\u003eRe-encoding、Levels of processing theory、Encoding specificity principle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eSemantic encoding\u003cstrong\u003e、\u003c/strong\u003eAssociative encoding、Levels of processing theory、\u003c/p\u003e\n \u003cp\u003eDual-coding theory、Recoding\u003cstrong\u003e、\u003c/strong\u003eFalse memory\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eDecay and forgetting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 282px;\"\u003e\n \u003cp\u003eForgetting curve、Interference theory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eForgetting curve、Interference theory、Retrieval-induced forgetting\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eNeural basis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 282px;\"\u003e\n \u003cp\u003eHippocampus、Prefrontal cortex、Amygdala、Cerebellum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003ePrefrontal cortex、Amygdala、Frontal lobe、Medial temporal lobe\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eStorage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 282px;\"\u003e\n \u003cp\u003eMemory trace、Neural plasticity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eUnlimited capacity\u003cstrong\u003e、\u003c/strong\u003eDurable\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 90px;\"\u003e\n \u003cp\u003eRelated Concepts\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 282px;\"\u003e\n \u003cp\u003eRelearning\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 312px;\"\u003e\n \u003cp\u003eShort-term memory、Sensory memory、Age\u003cstrong\u003e、\u003c/strong\u003eDisease\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eIn addition, you can also choose the general system analysis method 5W1H to conduct a preliminary analysis of long-term memory. The 5W1H method provides a systematic and comprehensive framework by clarifying the six dimensions of Why (why), Who (research subject), What (what), When (when), Where (where), and How (how), which helps to explore the potential mechanisms of long-term memory from multiple perspectives. The specific results are shown in Table 4. Through this general analysis framework, extensive exploration can be carried out in the early stage of research to accumulate multi-dimensional data about long-term memory.\u003c/p\u003e\n\u003cp\u003eTable 4 Examining the Research Gap in Working Memory Using the 5W1H Analysis Method\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" align=\"\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003e5W1H\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 472px;\"\u003e\n \u003cp\u003eLong-Term Memory\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003eWHY (Purpose)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 472px;\"\u003e\n \u003cp\u003eTo store and retrieve information, supporting knowledge accumulation, experience summarization, and skill formation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003eWHO(Related Concepts)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 472px;\"\u003e\n \u003cp\u003eMulti-store model (sensory memory, short-term memory, and long-term memory)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003eWHAT (Content)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 472px;\"\u003e\n \u003cp\u003eImplicit memory、Explicit memory\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003eWHEN (Timeline)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 472px;\"\u003e\n \u003cp\u003eWhen information transfers from short-term memory to long-term memory\u003cstrong\u003e、\u003c/strong\u003eDuring learning and recall\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003eWHERE (Location)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 472px;\"\u003e\n \u003cp\u003eHippocampus、Cerebral cortex、Amygdala、Cerebellum, etc\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 194px;\"\u003e\n \u003cp\u003eHOW (Mechanism)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 472px;\"\u003e\n \u003cp\u003eThrough deep processing, review and consolidation, and emotional association to consolidate and retrieve information\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eThese two methods have their own unique advantages. The method of inheriting the memory analysis framework can conduct in-depth analysis based on existing theories, ensuring that the results are closely aligned with the research theme; while 5W1H, as a general analysis framework, is more suitable for the initial stage of research, and helps to explore various potential problems and characteristics of long-term memory from multiple perspectives. Choosing an appropriate research framework according to the development stage of the research object\u0026apos;s field, or combining the two to complement each other, helps to discover new research dimensions of the research object, fill existing research gaps, and promote further development of the field.\u003c/p\u003e\n\u003cp\u003eTable 5 Generalizing the Research Framework to the Remaining Three Concepts\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" class=\"fr-table-selection-hover\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Framework\u003c/p\u003e\n \u003cp\u003eSubject\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 349px;\"\u003e\n \u003cp\u003eMechanism\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003eDuration\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eMemory capacity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 349px;\"\u003e\n \u003cp\u003eMulti-component model of working memory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003eMulti-store model\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eSelective attention\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 349px;\"\u003e\n \u003cp\u003eFilter theory、Multiple-channel model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003eEarly selection model、\u003c/p\u003e\n \u003cp\u003eLate selection model\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eImplicit memory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 349px;\"\u003e\n \u003cp\u003eIncubation effect、Implicit learning theory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003eImplicit temporal representation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 144px;\"\u003e\n \u003cp\u003eCognitive processing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 349px;\"\u003e\n \u003cp\u003eInformation processing theory、Levels of processing theory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 180px;\"\u003e\n \u003cp\u003eCognitive load theory\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003col start=\"3\" type=\"1\"\u003e\n \u003cli\u003e\u003cstrong\u003eUniversal framework:\u0026nbsp;\u003c/strong\u003eIt is assumed that many lower-level concepts of the same higher-level concept have similar local graph structures. Therefore, the existing research framework is extended to other concepts to discover new theories. For example, selective attention, implicit cognition, and cognitive processing and memory all belong to the category of psychological mechanisms. Therefore, analogical reasoning is performed, and the research framework of memory is applied to the other three nodes. In fact, the feasibility of this method has been confirmed. As shown in Table 5, there are indeed corresponding theories for studying the other three concepts based on the research framework of operation mechanism and time limit (time). If it is found that there is no theory when generalizing the theoretical framework, there may be two reasons: First, it does discover a new direction of research in the field; second, this attribute is not suitable for the research object.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eShifted objectives:\u003c/strong\u003e It is assumed that many lower-level concepts of the same higher-level concept have similar associations with the research subject. For example, long-term memory, sensory memory, and short-term memory belong to the same theory - the multi-store theory. There are also theories describing the association between working memory and long-term memory. Therefore, the association between working memory and the above elements can be discussed.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"5. Discussion and Conclusion","content":"\u003ch3\u003e5.1 Discussion\u003c/h3\u003e\n\u003cp\u003eLogical reasoning is crucial to the development of human intelligence. With the increasing penetration of data-driven AI models in various fields, their \u0026quot;black box\u0026quot; problem has become increasingly prominent. To ensure the reliability of AI, there is an urgent need to constrain and test the calculation process of AI through rigorous logical reasoning. Therefore, the demand for logical reasoning has also increased, and theory, as the cornerstone of logical reasoning, guides the direction of reasoning [44]. Rapidly building a solid theoretical foundation has become an urgent problem to be solved. Computational grounded theory is an efficient method for constructing a theoretical system. It integrates explanatory and predictive thinking [45], providing solid theoretical support for the development of scientific research, thereby solving the \u0026quot;black box\u0026quot; problem of AI and promoting its application in more fields.\u003c/p\u003e\n\u003cp\u003eTo break through the limitations of inefficient theory construction, more and more researchers have begun to turn their attention to the field of computational grounded theory. Brailas et al. (2014) [46] proposed networked grounded theory, introducing community discovery technology to cluster the coding network and automatically select the coding process; Nelson (2020)[17] formally proposed the idea of computational grounded theory, and provided a methodological framework that has both the excellent interpretive ability of human beings and the rigorous, reliable and reproducible characteristics of computer technology, which has received widespread attention from scholars in the field. If grounded research is divided into three stages: coding, inductive reasoning, and theory construction, then existing research results have significantly accelerated the tedious preparatory work in the early stage. However, how to improve the efficiency of the inductive reasoning and theory construction stages is still an urgent problem to be solved.\u003c/p\u003e\n\u003cp\u003eGrounded theory automation is one way to achieve this goal, and scholars have already tried to integrate methods such as large models [47][48] and machine learning [49] into grounded theory. The knowledge graph-based computational grounded method proposed in this paper aims to use the idea of knowledge visualization to break through the limitations of existing research. With the powerful functions of analyzing and mining relationships of knowledge graph, grounded research has the ability of complex reasoning and knowledge discovery, which not only improves the efficiency of inductive sorting theory, but also finds the pattern of discovering theoretical hypotheses, and enhances the explanatory and predictive nature of grounded research theoretical results. With the conceptual knowledge graph as a data support environment, subsequent research can use natural language processing technology to further realize grounded research automation. For example, using natural language processing technology to extract assertions and process them into triples, and then visually display complex concepts and relationships in the form of hypergraphs, thereby improving coding efficiency and consistency; by introducing a reasoning engine into the knowledge graph, complex causal relationships and associations hidden in the data can be identified, thereby expanding the depth and breadth of research; the knowledge graph provides unique advantages in relationship analysis and mining, and can discover patterns and implicit relationships in the data through network analysis algorithms, thereby realizing the development and innovation of the theoretical system.\u003c/p\u003e\n\u003ch3\u003e5.2 Conclusion\u003c/h3\u003e\n\u003cp\u003eManagement theory has long faced problems such as theory lagging behind practice and uneven levels of theory construction [50]. On the one hand, the rapidly changing and complex environment has placed higher demands on management theory; on the other hand, theoretical innovation highly depends on the personal experience of researchers, resulting in long research cycles and strong subjectivity of results. The knowledge graph-based computational grounded theory method proposed in this paper provides a more efficient, transparent, and explanatory new approach for theory construction and prediction.\u003c/p\u003e\n\u003cp\u003eThis study selected a traditional grounded research with scientific literature as qualitative data as a benchmark, and verified the performance improvement of grounded research after introducing the knowledge graph through comparative experiments. The results show that, by using the graph structure to describe the relationship between concepts and theories, the new method strengthens the research process of grounded theory in several links. In the data analysis link, the introduction of the knowledge hypergraph realizes the construction of multi-faceted relationships between theory and concepts, successfully breaking through the limitations of binary relationships. Moreover, the visualized research results facilitate the dissemination, sharing, and reuse of knowledge, enhance the consistency of researchers\u0026apos; interpretation of research and analysis, and improve communication efficiency in multi-person collaborative research. In the theory construction link, different types of edges and nodes in the graph are distinguished by color coding for targeted theoretical sampling, thereby improving the validity and reliability of supplementary data and information. In the conclusion review link, the complete construction process of the theory is visually displayed, which significantly improves the explanatory power of the theory. The reproducibility and traceability also confirm the improvement of the objectivity of the research conclusions. This series of improvements has enhanced the validity, completeness, and reliability of grounded theory.\u003c/p\u003e\n\u003cp\u003eAs Hofman et al. (2021)[51] said, research that integrates explanatory and predictive thinking will inject new vitality into the development of disciplines. The conceptual knowledge hypergraph proposed in this study can not only deeply explain existing theories and reveal the deep reasons behind phenomena, but also predict scientific hypotheses based on four new theory discovery paths, providing direction for future research. This method is general and not only applicable to the current research field, but also expected to be extended to other disciplines.With the continuous advancement of technology and the deepening of research, computational grounded theory will surely play an increasingly important role in promoting the development of multiple disciplines\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAuthor contributions: Song. proposes research topics based on knowledge graphs for grounded theory, design experimental plans, revise papers and finalize them.Wu. Participated in topic selection and experimental method design, managed the data, modeled the psychological structure of efficient mathematics learning using knowledge graph-based grounded theory, validated the method's effectiveness, visualized the findings, and drafted the initial manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col class=\"decimal_type\"\u003e\n \u003cli\u003eCharmaz, K. (2000). Constructivist and objectivist grounded theory.\u0026nbsp;Handbook of qualitative research,\u0026nbsp;2, 509-535.\u003c/li\u003e\n \u003cli\u003eTurner, C., \u0026amp; Astin, F. (2021). Grounded theory: What makes a grounded theory study? European Journal of Cardiovascular Nursing, 20(3), 285\u0026ndash;289.\u003c/li\u003e\n \u003cli\u003eColeman, G., O\u0026rsquo;Connor, R. (2007). Using grounded theory to understand software process improvement: A study of Irish software product companies. Information and Software Technology, 49(6), 654\u0026ndash;667.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eBirks, M., \u0026amp; Mills, J. (2015).\u0026nbsp;Grounded theory: A practical guide. Sage.\u003c/li\u003e\n \u003cli\u003eSuddaby, R. (2006). From the editors: What grounded theory is not.\u0026nbsp;Academy of management journal,\u0026nbsp;49(4), 633-642.\u003c/li\u003e\n \u003cli\u003eGlaser, Barney G., Anselm L. Strauss, and Elizabeth Strutzel. \u0026quot;The discovery of grounded theory; strategies for qualitative research.\u0026quot; [J],Nursing research,1968,17(4): 364.\u003c/li\u003e\n \u003cli\u003eWagner S M , Lukassen P , Mahlendorf M .Misused and missed use \u0026mdash; Grounded Theory and Objective Hermeneutics as methods for research in industrial marketing[J].Industrial Marketing Management, 2010, 39(1):5-15.\u003c/li\u003e\n \u003cli\u003eSaldana, Johnny. 2015. The Coding Manual for Qualitative Researchers. 3rd ed. Los Angeles, CA: Sage\u003c/li\u003e\n \u003cli\u003eBryant, A., \u0026amp; Charmaz, K. (2010).\u0026nbsp;The SAGE handbook of grounded theory: Paperback edition. Sage publications.\u003c/li\u003e\n \u003cli\u003eChun Tie, Y., Birks, M., \u0026amp; Francis, K. (2019). Grounded theory research: A design framework for novice researchers. SAGE Open Medicine, 7, 2050312118822927.\u003c/li\u003e\n \u003cli\u003eStrauss, Anselm, and Juliet Corbin.\u0026quot; Basics of qualitative research techniques.\u0026quot; (1998).\u003c/li\u003e\n \u003cli\u003eCharmaz, Kathy.\u0026quot; Constructivist and objectivist grounded theory.\u0026quot; ?Handbook of qualitative research?2 (2000): 509-535.\u003c/li\u003e\n \u003cli\u003eVollstedt, M., \u0026amp; Rezat, S. (2019). An introduction to grounded theory with a special focus on axial coding and the coding paradigm. Compendium for early career researchers in mathematics education, 13(1), 81-100.\u003c/li\u003e\n \u003cli\u003eJonsen, K., \u0026amp; Jehn, K. A. (2009). Using triangulation to validate themes in qualitative studies. Qualitative research in organizations and management: an international journal, 4(2), 123-150.\u003c/li\u003e\n \u003cli\u003eGuest, G., \u0026amp; MacQueen, K. M. (Eds.). (2008). Handbook for team-based qualitative research. Rowman Altamira.\u003c/li\u003e\n \u003cli\u003eBerente, N., \u0026amp; Seidel, S. (2014). Big data \u0026amp; inductive theory development: Towards computational Grounded Theory?.\u003c/li\u003e\n \u003cli\u003eNelson L K.Computational grounded theory: A methodological framework[J].Sociological methods \u0026amp; research, 2020, 49(1):3-42\u003c/li\u003e\n \u003cli\u003eCofie, N., Braund, H., \u0026amp; Dalgarno, N. (2022). Eight ways to get a grip on intercoder reliability using qualitative-based measures. Canadian Medical Education Journal, 13(2), 73-76.\u003c/li\u003e\n \u003cli\u003eJones, M., \u0026amp; Alony, I. (2011). Guiding the use of Grounded Theory in Doctoral studies\u0026ndash;an example from the Australian film industry.\u003c/li\u003e\n \u003cli\u003eMacQueen, K. M., McLellan-Lemal, E., Bartholow, K., \u0026amp; Milstein, B. (2008). Team-based codebook development: Structure, process, and agreement. Handbook for team-based qualitative research, 119, 119-135.\u003c/li\u003e\n \u003cli\u003eSchreier, M. (2012). Qualitative content analysis in practice. Jacobs University Bremen.\u003c/li\u003e\n \u003cli\u003eMcDonald, N., Schoenebeck, S., \u0026amp; Forte, A. (2019). Reliability and inter-rater reliability in qualitative research: Norms and guidelines for CSCW and HCI practice.\u0026nbsp;Proceedings of the ACM on human-computer interaction,\u0026nbsp;3(CSCW), 1-23.\u003c/li\u003e\n \u003cli\u003eD\u0026iacute;az, J., P\u0026eacute;rez, J., Gallardo, C., \u0026amp; Gonz\u0026aacute;lez-Prieto, \u0026Aacute;. (2023). Applying inter-rater reliability and agreement in collaborative grounded theory studies in software engineering.\u0026nbsp;Journal of Systems and Software,\u0026nbsp;195, 111520.\u003c/li\u003e\n \u003cli\u003eJonsen, K., \u0026amp; Jehn, K. A. (2009). Using triangulation to validate themes in qualitative studies. Qualitative research in organizations and management: an international journal, 4(2), 123-150.\u003c/li\u003e\n \u003cli\u003eHoughton, C., Murphy, K., Meehan, B., Thomas, J., Brooker, D., \u0026amp; Casey, D. (2017). From screening to synthesis: using nvivo to enhance transparency in qualitative evidence synthesis. Journal of clinical nursing, 26(5-6), 873-881.\u003c/li\u003e\n \u003cli\u003eLi, Z. G. (2007). The study of grounded theory in business research. In\u0026nbsp;Oriental Forum\u0026nbsp;(Vol. 4, pp. 90-94).\u003c/li\u003e\n \u003cli\u003eChen, Y., Chen, C. M., Liu, Z. Y., Hu, Z. G., \u0026amp; Wang, X. W. (2015). The methodology function of CiteSpace map** knowledge domains.\u0026nbsp;Studies in Science of Science,\u0026nbsp;33(2), 242-253.\u003c/li\u003e\n \u003cli\u003eJi, S., Pan, S., Cambria, E., Marttinen, P., \u0026amp; Philip, S. Y. (2021). A survey on knowledge graphs: Representation, acquisition, and applications. IEEE transactions on neural networks and learning systems, 33(2), 494-514.\u003c/li\u003e\n \u003cli\u003eKerlinger, Fred Nichols.\u0026quot; Foundations of behavioral research.\u0026quot; [J].1966\u003c/li\u003e\n \u003cli\u003eAntelmi, A., Cordasco, G., Polato, M., Scarano, V., Spagnuolo, C., \u0026amp; Yang, D. (2023). A survey on hypergraph representation learning. ACM Computing Surveys, 56(1), 1-38.\u003c/li\u003e\n \u003cli\u003eZhou, D., Huang, J., \u0026amp; Sch\u0026ouml;lkopf, B. (2006). Learning with hypergraphs: Clustering, classification, and embedding. Advances in neural information processing systems, 19.\u003c/li\u003e\n \u003cli\u003eTian, L., Zhou, X., Wu, Y. P., Zhou, W. T., Zhang, J. H., \u0026amp; Zhang, T. S. (2022). Knowledge graph and knowledge reasoning: A systematic review. Journal of Electronic Science and Technology, 20(2), 100159.\u003c/li\u003e\n \u003cli\u003eEstrada, E., \u0026amp; Rodriguez-Velazquez, J. A. (2005). Complex networks as hypergraphs. arxiv preprint physics/0505137.\u003c/li\u003e\n \u003cli\u003eBaddeley A.Working memory[J].Science, 1992, 255(5044):556-559\u003c/li\u003e\n \u003cli\u003eBaddeley, Alan.\u0026quot; The episodic buffer: a new component of working memory?.\u0026quot;[J]. Trends in cognitive sciences ,2000,4 (11): 417-423.\u003c/li\u003e\n \u003cli\u003eChen, X. (2008). The New Development of Qualitative Research and Its Significance to Social Science Research. Educational Research and Experiment, No. 2, 14-18.\u003c/li\u003e\n \u003cli\u003eGlaser, Barney, and Anselm Strauss.Discovery of grounded theory: Strategies for qualitative research.[M] Routledge, 2017.\u003c/li\u003e\n \u003cli\u003eGlaser, B. G., \u0026amp; Strauss, A. (2012). The discovery of grounded theory. Aldine Transaction. (Original work published 1967)\u003c/li\u003e\n \u003cli\u003eConlon, C., Timonen, V., Elliott-O\u0026rsquo;Dare, C., O\u0026rsquo;Keeffe, S., \u0026amp; Foley, G. (2020). Confused about theoretical sampling? Engaging theoretical sampling in diverse grounded theory studies.\u0026nbsp;Qualitative health research,\u0026nbsp;30(6), 947-959.\u003c/li\u003e\n \u003cli\u003eGuangming, W., Wenjuan, S., \u0026amp; ****, S. (2014). The psychological structures model of highly effective mathematics learning based on the NVivo10 qualitative analysis.\u0026nbsp;Studies of Psychology and Behavior,\u0026nbsp;1(1), 1-13.\u003c/li\u003e\n \u003cli\u003eGentner, D., \u0026amp; Maravilla, F. (2017). Analogical reasoning. In\u0026nbsp;International handbook of thinking and reasoning\u0026nbsp;(pp. 186-203). Routledge.\u003c/li\u003e\n \u003cli\u003eChen, P., \u0026amp; Garcia, W. (2010, July). Hypothesis generation and data quality assessment through association mining. In\u0026nbsp;9th IEEE International Conference on Cognitive Informatics (ICCI\u0026apos;10)\u0026nbsp;(pp. 659-666). IEEE.\u003c/li\u003e\n \u003cli\u003eKant, I. (1797). Allgemeine Naturgeschichte und Theorie des Himmels, oder Versuch von der Verfassung und dem mechanischen Ursprunge des ganzen Weltgeb\u0026auml;udes: nach Newtonschen Grunds\u0026auml;tzen abgehandelt.\u003c/li\u003e\n \u003cli\u003eLiu, H., Ning, R., Teng, Z., Liu, J., Zhou, Q., \u0026amp; Zhang, Y. (2023). Evaluating the logical reasoning ability of chatgpt and gpt-4.\u0026nbsp;arxiv preprint arxiv:2304.03439.\u003c/li\u003e\n \u003cli\u003eChen, Z., \u0026amp; Chen, Y. (2024). Computing grounded theory: a quantitative method to develop theories.\u0026nbsp;The Journal of Chinese Sociology,\u0026nbsp;11(1), 17.\u003c/li\u003e\n \u003cli\u003eBrailas, A. V. (2014). Networked grounded theory. The Qualitative Report, 19(8), 1-16.\u003c/li\u003e\n \u003cli\u003e\u0026Uuml;bellacker, T. (2024). AcademiaOS: Automating Grounded Theory Development in Qualitative Research with Large Language Models.\u0026nbsp;arxiv preprint arxiv:2403.08844.\u003c/li\u003e\n \u003cli\u003eZhou, Y., Yuan, Y., Huang, K., \u0026amp; Hu, X. (2024). Can ChatGPT perform a grounded theory approach to do risk analysis? An empirical study. Journal of Management Information Systems, 41(4), 982-1015.\u003c/li\u003e\n \u003cli\u003eMuller, M., Guha, S., Baumer, E. P., Mimno, D., \u0026amp; Shami, N. S. (2016, November). Machine learning and grounded theory method: convergence, divergence, and combination. In\u0026nbsp;Proceedings of the 2016 ACM International Conference on Supporting Group Work\u0026nbsp;(pp. 3-8).\u003c/li\u003e\n \u003cli\u003eVan de Ven, A. H. (2007). Engaged scholarship: A guide for organizational and social research. Oxford University Press.\u003c/li\u003e\n \u003cli\u003eHofman, J. M., Watts, D. J., Athey, S., Garip, F., Griffiths, T. L., Kleinberg, J., ... \u0026amp; Yarkoni, T. (2021). Integrating explanation and prediction in computational social science. Nature, 595(7866), 181-188.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Computational Grounded Theory, Theory Development, Knowledge Graph, Knowledge Visualization, Theory Construction","lastPublishedDoi":"10.21203/rs.3.rs-6433331/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6433331/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTraditional grounded theory (GT) heavily depends on researchers' cognitive abilities, making theoretical construction highly subjective and inconsistent. This reliance on individual expertise results in significant variations in research quality, limiting the interpretability, reproducibility, and scalability of GT. Recent advancements in knowledge extraction technologies have demonstrated the potential to bridge this gap by enabling non-expert researchers to achieve expert-level analytical capabilities. To address these challenges, this study proposes a computational grounded theory method based on knowledge graphs (CGT-KG), integrating knowledge graph techniques to enhance theory construction. By systematically representing multi-dimensional concept-theory relationships, CGT-KG reduces subjectivity, improves transparency, and strengthens theoretical validation. Taking the construction of a psychological structure model for efficient mathematics learning as a case, the paper verifies that the knowledge hypergraph strengthens grounded theory in three aspects: ① multi-dimensional concept-theory relationship, the knowledge hypergraph overcomes the limitations of binary relationships, making theory construction richer and more structured, and the visualized results promote knowledge sharing among researchers; ② enhanced theory verification, the color coding of nodes and edges provides a new perspective for theoretical saturation test and theoretical sampling; ③ automatic hypothesis discovery, a new theory generation framework based on analogical reasoning is proposed, making systematic exploration of new theories possible.\u003c/p\u003e","manuscriptTitle":"A Knowledge Graph-Based Approach to Enhancing Grounded Theory Research","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-05 12:22:06","doi":"10.21203/rs.3.rs-6433331/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a43c72e6-7756-4c71-a903-83524432d09c","owner":[],"postedDate":"August 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":52341234,"name":"Business and commerce/Information systems and information technology"},{"id":52341235,"name":"Social science/Complex networks"},{"id":52341236,"name":"Social science/Education"}],"tags":[],"updatedAt":"2025-11-25T11:53:50+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-05 12:22:06","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6433331","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6433331","identity":"rs-6433331","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}