GLBCC: Cognitive Conflict-Based Generative Learning

• The paper’s main contribution is demonstrating that integrating cognitive conflict with generative learning enhances science literacy in physics through a structured six-stage model.
• GLBCC is defined by deliberately creating discrepancies that trigger active reconstruction of concepts, thus fostering deeper critical thinking and conceptual change.
• Applications include AI-mediated deliberation frameworks and interactive multimedia designs that use structured conflict to promote reflective reasoning and effective learning.

Searching arXiv for the specified GLBCC-related papers to ground the article in current research. The Cognitive Conflict-Based Generative Learning Model (GLBCC) is a learning model that combines cognitive conflict and generative learning so that learners do not merely receive explanations but are induced to confront discrepancies between prior conceptions and new evidence, articulate their reasoning, reconstruct concepts, and apply them in new contexts. Across recent work, GLBCC appears in two closely related forms: as a six-syntax constructivist instructional model for physics learning oriented toward science literacy (Akmam et al., 1 Sep 2025), and as a GLBCC-style AI-mediated deliberation framework in which a Peer Agent intentionally generates socio-cognitive conflict to deepen reflective reasoning in dilemma-based discussion (Kim et al., 9 Aug 2025). A further line of work uses learners’ critical thinking profiles, learning styles, and media needs as the basis for designing interactive multimedia grounded in a generative learning model with cognitive conflict strategy, though without yet reporting a completed product or intervention test (Ahzari et al., 24 Oct 2025).

1. Conceptual definition and scope

GLBCC is presented most explicitly as the “Cognitive Conflict-Based Generative Learning Model,” a physics teaching model intended to improve science literacy by integrating two mechanisms: first, cognitive conflict, in which students are confronted with phenomena or information that contradict their existing misconceptions; second, generative learning, in which students actively construct meaning by connecting prior knowledge with new information rather than passively receiving explanations (Akmam et al., 1 Sep 2025). In that formulation, GLBCC is rooted in constructivism and organized as a six-syntax learning model.

The same underlying principle is instantiated in a different domain by the “Peer Agent” system described in “Your Thoughtful Opponent: Embracing Cognitive Conflict with Peer Agent” (Kim et al., 9 Aug 2025). There, the emphasis is less on the formal GLBCC label and more on a system that operationalizes the same logic: learning becomes generative when a thoughtful opponent provokes, mirrors, and strategically challenges learners’ views so that they must elaborate, justify, compare, and possibly revise their reasoning. In this implementation, socio-cognitive conflict is intentionally generated by an AI peer to deepen reflective thinking in collaborative deliberation (Kim et al., 9 Aug 2025).

A third paper extends the conceptual family of GLBCC into digital media design. “Analyzing Students Critical Thinking as a Basis for Developing Interactive Physics Multimedia with Generative Learning and Cognitive Conflict Strategies” treats a generative learning model integrated with a cognitive conflict strategy as the basis for future interactive physics multimedia development (Ahzari et al., 24 Oct 2025). That study is explicitly a needs-analysis and foundational study rather than a report of a finished multimedia product or an experimental effectiveness test.

Taken together, these works define GLBCC not as a single immutable procedure but as a broader learning-design logic in which cognitive tension functions as the trigger for active knowledge generation, conceptual reconstruction, and higher-order reasoning.

2. Theoretical basis: cognitive conflict and generative learning

The core of GLBCC is the pairing of cognitive conflict with generative learning. In the physics-learning study, cognitive conflict is described as the deliberate presentation of discrepant events, anomalies, or real-life physics problems that do not fit learners’ existing ideas, thereby triggering curiosity, exposing misconceptions, creating dissatisfaction with incorrect explanations, and motivating conceptual reorganization (Akmam et al., 1 Sep 2025). Generative learning then provides the knowledge-construction process through which students explain, discuss, test, revise, and apply ideas until a coherent scientific explanation is built (Akmam et al., 1 Sep 2025).

The multimedia-oriented study frames the same combination in closely aligned terms. It describes cognitive conflict strategy as creating an “intellectual mismatch” between what students expect and what they observe, especially in physics where abstract concepts are prone to misconception (Ahzari et al., 24 Oct 2025). Generative learning is treated as an active knowledge-construction process in which students organize, relate, and integrate new information with prior knowledge, infer, analyze, evaluate, and explain, and then reorganize understanding based on feedback (Ahzari et al., 24 Oct 2025). The proposed multimedia design therefore uses generative learning as the learning pathway and cognitive conflict as the trigger for deeper processing.

The deliberative AI work extends the mechanism from conceptual conflict in science learning to socio-cognitive conflict in value-laden discussion. Drawing on classic findings from Doise and Mugny, it treats exposure to differing viewpoints as a source of cognitive tension that promotes cognitive restructuring and deeper understanding (Kim et al., 9 Aug 2025). In that setting, the conflict is engineered by the system rather than left to chance: the agent adopts an opposing stance when human discussants agree, or aligns with the player with lower confidence when they disagree, thereby amplifying the minority viewpoint (Kim et al., 9 Aug 2025). This produces epistemic tension, productive disagreement, and deeper engagement during deliberation.

A plausible implication is that GLBCC functions as a general design pattern for learning environments in which disequilibrium is not regarded as a breakdown to be minimized, but as a productive condition for reflection, conceptual change, and meaning-making.

3. Instructional structure in physics education

In its most formal educational expression, GLBCC is a six-stage framework for physics instruction (Akmam et al., 1 Sep 2025). The sequence is presented as:

$$\text{Orientation} \rightarrow \text{Cognitive Conflict} \rightarrow \text{Disclosure} \rightarrow \text{Construct} \rightarrow \text{Application} \rightarrow \text{Reflection and Evaluation}$$

Each stage has a specific instructional function. Orientation activates prior knowledge, introduces the topic, asks opening questions, and orients students to objectives and supporting materials. Cognitive Conflict presents discrepant events, anomalies, or real-life problems that do not fit prior ideas. Disclosure is the stage in which students express ideas, propose explanations, and consider possible solutions. Construct supports the reorganization of prior knowledge into scientifically correct concepts through teacher guidance and learning resources. Application requires students to test and transfer newly constructed concepts to experiments, problem solving, or other contexts. Reflection and Evaluation consolidates learning, improves metacognition, provides feedback, and reinforces correct conceptual structures (Akmam et al., 1 Sep 2025).

The model is described not merely as a sequence of activities but as a system involving syntax or stages, a support system such as teaching modules, worksheets, materials, kits, and evaluation tools, a reaction principle emphasizing student-centered and process-oriented science literacy learning, and intended instructional impacts focused on science literacy competencies (Akmam et al., 1 Sep 2025).

This six-stage form of GLBCC was developed to address three connected problems in physics learning: persistent misconceptions, weak science literacy, and limited engagement in meaningful scientific thinking (Akmam et al., 1 Sep 2025). Conventional expository teaching is described as often emphasizing memorization and teacher explanation, which are insufficient for helping students explain phenomena scientifically, interpret data, or apply scientific reasoning in real-life contexts (Akmam et al., 1 Sep 2025). GLBCC therefore attempts to make learning more contextual and cognitively demanding.

The multimedia-design study does not provide an equivalent finalized GLBCC instructional syntax, but its intended multimedia sequence aligns with the same logic: visual and contextual presentation of a physics phenomenon, initial student understanding or prediction, presentation of a conflicting phenomenon or simulation result, reflection and analysis, and feedback-guided reconstruction of the correct concept (Ahzari et al., 24 Oct 2025). This suggests continuity between classroom pedagogy and digital learning environments in the use of conflict-triggered generative processing.

4. AI-mediated GLBCC-style deliberation

The Peer Agent system constitutes a GLBCC-style framework for dilemma-based learning rather than a classroom physics model (Kim et al., 9 Aug 2025). Its educational problem is the cultivation of democratic deliberation skills in contexts defined by conflicting values, competing stakeholder perspectives, and the absence of a simple correct answer. The paper specifically targets AI ethics dilemmas in a game-like setting, including the example “should we allow the development of AI killer robots?” (Kim et al., 9 Aug 2025).

In this framework, learners state both a stance and how strongly they hold it on a 1–5 scale. The system then uses stance and opinion strength to shape the agent’s behavior and generate meaningful discussion tension (Kim et al., 9 Aug 2025). The learning goal is reflective deliberation rather than content memorization. Learners are expected to consider multiple perspectives, engage in collaborative decision-making, defend and revise positions, and reach socially acceptable consensus (Kim et al., 9 Aug 2025).

The Peer Agent is intended to function as a socially coherent peer rather than an instructor or authority. It has two main roles: participating as a peer in voice-based multi-party deliberation and dynamically intervening to stimulate deliberation by adopting or shifting stance strategically, nudging human players to elaborate and defend positions, and using inner thoughts to decide when and how to speak (Kim et al., 9 Aug 2025). This role design is significant because the effectiveness of socio-cognitive conflict is tied to the perceived legitimacy of the interlocutor as a conversational partner rather than an external evaluator.

The system architecture consists of five modules: Context Interpreter, Agent State Manager, Thought Generator, Thought Evaluator, and Thought Articulator (Kim et al., 9 Aug 2025).

Module	Function
Context Interpreter	Stores each player’s position and confidence score, analyzes dialogue using basic human values theory, and tracks discussion phase
Agent State Manager	Tracks Position, Opinion Strength, Long-Term Memory, and Agent Persona
Thought Generator	Creates candidate General Thoughts and Strategic Thoughts from utterances, extracted values, and retrieved memory
Thought Evaluator	Assigns a motivation score from 1 to 5 using heuristic criteria and rule-based logic
Thought Articulator	Selects and expresses the final utterance in spoken or textual peer-like form

The conflict-generation rule is central. If both human players agree, the agent adopts the opposing stance. If the human players disagree, the agent aligns with the player who has lower opinion strength (Kim et al., 9 Aug 2025). This makes the agent a conflict catalyst rather than a neutral responder. Strategic thoughts are preferred when their motivation score exceeds a threshold; otherwise, general thoughts are selected probabilistically to keep the conversation fluid (Kim et al., 9 Aug 2025). The agent’s initial opinion strength is the average of the two human players’ confidence scores and may be updated dynamically when persuasive arguments are detected (Kim et al., 9 Aug 2025).

Theoretical grounding is provided by the Inner Thoughts framework by Liu et al., which supports proactive agent interventions through turn-taking driven by intrinsic motivation, and by value-sensitive discourse analysis using Schwartz’s theory of basic human values to detect value-laden content (Kim et al., 9 Aug 2025). This places the deliberative GLBCC-style instantiation at the intersection of value analysis, argumentation, and socio-cognitive conflict.

5. Empirical evidence and reported outcomes

The strongest direct empirical evidence for GLBCC in the supplied literature comes from the experimental physics study (Akmam et al., 1 Sep 2025). That study employed a quasi-experimental pretest-posttest control group design involving 167 Grade XI high school students from three schools. The experimental group comprised 83 students receiving GLBCC, while the control group comprised 84 students receiving Expository Learning (Akmam et al., 1 Sep 2025). Science literacy was measured using a validated instrument covering six indicators: ability to describe scientific inquiry and apply it to investigating problems; ability to describe and carry out experimental procedures; ability to perform appropriate laboratory tasks; ability to interpret and communicate scientific information in written, oral, and graphical forms; ability to describe science-technology-society relationships and daily-life applications; and ability to demonstrate logical reasoning in explaining phenomena and scientific or technological applications (Akmam et al., 1 Sep 2025). Reported reliability coefficients exceeded 0.87 for all subscales.

The statistical analysis used ANOVA, Tukey HSD post-hoc tests, factor analysis, KMO test, Bartlett’s test of sphericity, and Principal Component Analysis with varimax rotation, implemented in SPSS 25 (Akmam et al., 1 Sep 2025). The central finding was that students receiving GLBCC obtained higher posttest science literacy scores than those in the expository condition, with differences statistically significant at $p < 0.001$ (Akmam et al., 1 Sep 2025). The paper highlights a reported mean difference of 20.892 with $p = .000$ for one GLBCC group relative to an EL group. The authors interpret these findings as evidence that GLBCC actively engages students in scientific and logical reasoning, confronts misconceptions directly, supports conceptual restructuring, and strengthens the ability to apply physics ideas in practical contexts (Akmam et al., 1 Sep 2025).

The needs-analysis study does not report intervention effectiveness, but it provides the diagnostic basis for why a GLBCC-like multimedia intervention was considered necessary (Ahzari et al., 24 Oct 2025). It involved 125 eleventh-grade students from three public high schools in Lima Puluh Kota Regency, Indonesia, selected to represent high, medium, and low accreditation settings (Ahzari et al., 24 Oct 2025). The study used five validated instruments: a teacher questionnaire, an interactive multimedia needs analysis questionnaire, a student attitude questionnaire adapted from CLASS, a learning style questionnaire, and a critical thinking skills test consisting of five contextual physics-phenomenon narrative questions (Ahzari et al., 24 Oct 2025).

The reported results show that current instruction relied mainly on PowerPoint, learning videos, audio, and real objects, which the authors characterize as limited for supporting varied learning styles and deeper critical reasoning (Ahzari et al., 24 Oct 2025). Visual learning style was dominant, with reported proportions of 53.06% in SHS A, 50.00% in SHS B, and 40.00% in SHS C (Ahzari et al., 24 Oct 2025). Students’ attitudes toward physics were moderately positive overall at 65.69%, with learning attitude at 71.46%, material attitude at 63.57%, and learning effort at 62.02% (Ahzari et al., 24 Oct 2025). Critical thinking skills were reported as critically low, averaging 27.80%, with indicator averages of 27.67% for interpretation, 28.33% for analysis, 28.00% for evaluation, 24.00% for inference, and 31.00% for explanation (Ahzari et al., 24 Oct 2025). The study explicitly concludes that these findings justify the development of visual, interactive multimedia that facilitates cognitive conflict and deeper reflection.

By contrast, the Peer Agent paper is primarily architectural and conceptual. It describes operational rules, workflow, and theoretical grounding for a GLBCC-style system, but the supplied material does not report comparable controlled effectiveness metrics (Kim et al., 9 Aug 2025).

6. Critical implementation factors, design requirements, and limitations

The physics experiment also examined implementation conditions through factor analysis (Akmam et al., 1 Sep 2025). Preliminary screening yielded an initial KMO of 0.637 and a significant Bartlett’s test with $\chi^2 = 151.018$, $df = 78$, $p < 0.001$. After excluding weak variables, the filtered model had KMO = 0.772 and a still-significant Bartlett’s test with $\chi^2 = 236.556$, $p < 0.001$ (Akmam et al., 1 Sep 2025). Variables with MSA below 0.5 were removed, including GLBCC implementation guidelines, reflection and evaluation guidelines, and time adequacy. The rotated component analysis then yielded four critical factors: science literacy development components; learning stages and orientation; motivation and objectives; and knowledge construction processes (Akmam et al., 1 Sep 2025).

These factors are significant because they indicate that successful GLBCC implementation depends on more than the mere presence of discrepant events. The model requires explicit cultivation of literacy behaviors, careful sequencing of stages, clear learning motivation and objectives, and robust support for knowledge construction (Akmam et al., 1 Sep 2025). A plausible implication is that poorly structured conflict may produce confusion rather than conceptual change if reconstructive phases are weak.

The multimedia-needs study identifies additional practical constraints relevant to GLBCC deployment in digital environments. These include time constraints, limited school technology, and teachers’ limited technical skills (Ahzari et al., 24 Oct 2025). At the same time, the learner profile reported in that study favors strong visual and interactive representations, suggesting a need for multimedia systems that are not only cognitively challenging but also compatible with local resource conditions and teacher capacity (Ahzari et al., 24 Oct 2025).

The Peer Agent framework has its own implementation requirements. Its design depends on accurate detection of position and confidence, value-sensitive discourse analysis, discussion-phase tracking, thought generation from utterances and memory, motivation-based evaluation, and threshold-governed articulation (Kim et al., 9 Aug 2025). Because the agent is intended to function as a socially coherent peer, the quality of turn-taking, the tone of articulation, and the timing of interventions are integral to the model rather than peripheral interface choices. This suggests that the effectiveness of AI-mediated GLBCC-style systems may depend on preserving both conversational realism and pedagogical intent.

One common misconception would be to treat GLBCC as identical across all domains. The available literature does not support that simplification. In physics education, GLBCC is a formal six-stage instructional model directly tied to science literacy outcomes (Akmam et al., 1 Sep 2025). In AI-supported deliberation, it appears as a style of framework in which socio-cognitive conflict is intentionally generated by a peer-like agent within value-laden discussion (Kim et al., 9 Aug 2025). In multimedia design research, it is a conceptual basis for future intervention development rather than a completed validated product (Ahzari et al., 24 Oct 2025).

7. Significance and research trajectory

GLBCC occupies a position at the convergence of constructivist pedagogy, conceptual change research, science literacy development, critical thinking, and AI-mediated deliberation. In the physics study, its significance lies in showing that a structured synthesis of cognitive conflict and generative learning can outperform expository learning in improving science literacy among Grade XI students (Akmam et al., 1 Sep 2025). In the multimedia-design study, its significance lies in providing a pedagogical rationale for interactive environments that are visually rich, cognitively demanding, and explicitly oriented toward conceptual restructuring and critical thinking (Ahzari et al., 24 Oct 2025). In the Peer Agent work, its significance lies in extending the logic of cognitive conflict from subject-matter misconception correction to democratic deliberation on controversial issues, where conflict is used to promote reflective reasoning rather than mere opposition (Kim et al., 9 Aug 2025).

Several lines of development are already implied by the three papers. One is the translation of GLBCC principles into multimodal and interactive digital systems, especially for abstract physics concepts and varied learner profiles (Ahzari et al., 24 Oct 2025). Another is the extension of conflict-based generative learning from classroom science to socially embedded, value-sensitive discussion supported by conversational AI (Kim et al., 9 Aug 2025). A third is the refinement of implementation conditions, such as stage structure, motivational design, and support for knowledge construction, which the factor analysis identifies as central to successful practice (Akmam et al., 1 Sep 2025).

The present literature also leaves clear boundaries. The multimedia study explicitly states that future research should develop prototypes, test them experimentally or with mixed methods, and measure learning outcomes (Ahzari et al., 24 Oct 2025). The Peer Agent paper, as summarized here, primarily contributes a system-level architecture and decision rules rather than an experimental evaluation of learning gains (Kim et al., 9 Aug 2025). Consequently, the strongest empirical support currently documented is for classroom-based GLBCC in physics learning, while AI-mediated and multimedia-based variants remain more architectural or developmental.

In aggregate, the research portrays GLBCC as a model family in which deliberate contradiction is not an end in itself but the initiating condition for articulation, comparison, reconstruction, application, and reflection. Whether enacted through teacher-guided classroom syntax, interactive multimedia, or a peer-like deliberative agent, the model’s defining principle is that learning deepens when cognitive conflict is structured so that learners must generate understanding rather than inherit it ready-made.