Abstract
This study aimed to design a framework for constructive interaction between preservice teachers and artificial intelligence (AI). This study employed a mixed-methods approach using an exploratory sequential design. In the qualitative phase, data were collected through in-depth semi-structured interviews and analyzed using Attride-Stirling’s (2001) thematic network analysis. Participants were selected using purposive sampling with a criterion-based sampling technique. The sampling and interview process continued until theoretical saturation was achieved. In the content validation phase, the statistical population consisted of experts in educational sciences, psychology, and technology at Farhangian University in 2025. A sample of 15 experts was selected using convenience sampling, and data were collected through a researcher-made questionnaire. Content validity was assessed using the Content Validity Ratio (CVR) and Content Validity Index (CVI). The qualitative findings resulted in the identification of one overarching theme entitled “Framework for Constructive Interaction between Preservice Teachers and Artificial Intelligence,” five organizing themes, and 23 basic themes. The organizing themes included “Problem Ownership” with five basic themes, “Interaction Refinement” with five basic themes, “Critical Evaluation” with four basic themes, “Verification” with four basic themes, and “Knowledge Reconstruction” with five basic themes. The results of the content validation indicated that all basic themes demonstrated acceptable content validity (CVR > 0.49 and CVI > 0.79). The findings suggest that AI education at Farhangian University should extend beyond workshops and the acquisition of technical skills and place greater emphasis on developing preservice teachers’ abilities to engage in constructive interaction with AI.

Introduction
The rapid expansion of artificial intelligence (AI) in education has fundamentally reshaped the relationship between learners, content, and instructional processes. AI-based tools—including large language models such as ChatGPT, Gemini, Claude, and Copilot, intelligent tutoring systems, recommendation engines, chatbots, and content-generation platforms—offer new opportunities for personalized learning, immediate feedback, adaptive assessment, and learner support. At the same time, their increasing presence in educational settings raises serious questions about the nature of human–machine interaction, learner agency, critical thinking, and ethical responsibility. Passive or uncritical use of AI can encourage superficial thinking, over-reliance on generated content, reduced engagement with lived experience, and exposure to inaccurate or biased information. Consequently, the issue is not merely whether student-teachers have access to AI tools or receive technical training in their use, but how their interaction with AI can be guided in a purposeful, critical, and learning-oriented manner.
This concern is particularly significant for Farhangian University, the principal institution responsible for teacher education in Iran. Student-teachers are future teachers who will enter classrooms where AI-based tools are increasingly part of learning activities. If they use AI without a coherent framework for constructive interaction, they may reproduce passive and instrumental patterns of use in their own teaching. Previous research has largely focused on identifying opportunities, challenges, or technical skills related to AI in education, but less attention has been paid to how student-teachers’ interaction with AI should be organized and directed within the curriculum. The present study addresses this gap by designing a framework for student-teachers’ constructive interaction with AI in the curriculum of Farhangian University. The central research questions are: What are the components of student-teachers’ constructive interaction with AI in the curriculum of Farhangian University? And do the identified components possess content validity? The study draws on constructivist and interactional perspectives, including Dewey’s emphasis on interaction as central to education and the view that learners construct knowledge by transforming information in light of their prior experience and values. Constructive interaction is thus understood as purposeful, active, and critical engagement with AI in the process of learning and knowledge construction, rather than passive consumption of AI outputs.

Research Methodology
The study employed an exploratory sequential mixed-methods design. In the qualitative phase, data were collected through in-depth, semi-structured interviews with 16 participants affiliated with Farhangian University, including faculty members in educational sciences and psychology and technology specialists. Participants were selected through criterion-based purposive sampling. Inclusion criteria included willingness to participate, at least a master’s degree in relevant fields, more than five years of teaching or administrative experience in modern educational technologies, and practical experience with AI in education, including workshop delivery. Interviews continued until theoretical saturation; no new codes or themes emerged after interviews 15 and 16. Interviews were conducted in person between October and November 2025, averaged about 50 minutes, were audio-recorded with consent, and were transcribed. Qualitative data were analyzed using Attride-Stirling’s (2001) thematic network analysis, proceeding from basic themes to organizing themes and a global theme. To reduce bias, coding was performed by three researchers, disagreements were resolved through discussion, and findings were reviewed by participants and an external qualitative expert. Inter-coder agreement was 0.87. In the content-validation phase, a researcher-developed questionnaire containing 23 items based on the basic themes was distributed to 15 experts in educational sciences, psychology, and technology at Farhangian University in 2025, selected through convenience sampling. Content validity was assessed using the Content Validity Ratio (CVR) and Content Validity Index (CVI). With 15 experts, the minimum acceptable CVR was 0.49, and the acceptable CVI threshold was 0.79. Ethical considerations included informed consent, confidentiality, anonymity, and academic integrity.

Discussion
Problem Ownership as the Starting Point of Constructive Interaction
The first organizing theme, “problem ownership,” comprises five basic themes: determining the learning goal, defining the problem, recognizing prior experiences, setting expectations, and responsibility-taking. The findings indicate that constructive interaction with AI begins before the learner enters the interaction itself. Student-teachers must first clarify what they intend to learn and why. They must be able to formulate the problem in their own language, which signals that they have an initial understanding of the subject rather than merely forwarding a vague request to an AI system. Recognizing prior knowledge and experiences is equally important: participants emphasized that AI should not be the starting point of learning but should be used against a background of existing knowledge so that its outputs can be evaluated meaningfully. Setting expectations involves understanding what the technology can and cannot do and what responsibilities remain with the learner. Responsibility-taking means that the student-teacher remains accountable for the educational and ethical consequences of using AI-generated content. This theme aligns with constructivist and self-regulated learning perspectives, which stress learner agency, goal-setting, and ownership of the learning process. It also responds to the risk that AI may replace thinking rather than support it.
Interaction Refinement as a Dialogic and Iterative Process
The second organizing theme, “interaction refinement,” includes five basic themes: designing precise questions, providing context, revising requests, multi-stage interaction, and managing the dialogue process. The findings show that the quality of AI outputs depends heavily on the quality of the learner’s interaction. Many inappropriate responses arise not from the AI system alone but from poorly formulated questions. Student-teachers need to learn how to ask clear, precise, and goal-relevant questions, provide sufficient contextual information, and revise their requests iteratively. Constructive interaction is not a one-shot transaction; the first response is rarely the final answer. Learners must manage a multi-stage dialogue, examining the problem from different angles and guiding the conversation toward their learning objectives. This theme reflects a dialogic and constructivist understanding of learning: knowledge is gradually refined through interaction, questioning, and revision. It also highlights the active role of the learner in directing AI rather than being directed by it.
Critical Evaluation of AI Outputs
The third organizing theme, “critical evaluation,” comprises four basic themes: identifying assumptions, examining logical coherence, seeking alternative perspectives, and analyzing strengths and weaknesses. The findings indicate that constructive interaction requires learners to evaluate AI-generated responses rather than accept them uncritically. Student-teachers must identify explicit and implicit assumptions underlying AI outputs, assess the logical coherence of arguments, and search for alternative viewpoints. They must also analyze the strengths and limitations of responses. This theme is consistent with critical thinking and critical pedagogy, which emphasize questioning, analysis, and reflective judgment. In the context of AI, critical evaluation is essential because generated content may appear polished and persuasive while containing logical gaps, factual errors, or hidden biases. The theme thus positions the learner as an active critic rather than a passive recipient of machine-generated knowledge.
Verification as a Necessary Safeguard
The fourth organizing theme, “verification,” includes four basic themes: trustworthiness of information, credibility of information, identifying possible errors, and detecting algorithmic biases. Participants regarded verification as one of the most important stages of interaction with AI. Since AI systems may produce inaccurate, incomplete, or biased information, learners must check outputs against credible sources, distinguish scientific sources from non-specialist content, identify factual and reasoning errors, and recognize algorithmic biases rooted in training data. Verification is therefore a core component of AI literacy and a safeguard against uncritical acceptance. It also reflects the ethical dimension of AI use: learners are responsible for the accuracy and credibility of the knowledge they adopt and later use in educational contexts.
Knowledge Rewriting as the Final Stage of Constructive Interaction
The fifth organizing theme, “knowledge rewriting,” comprises five basic themes: rewriting content in one’s own language, adding personal interpretation, integrating new information with prior knowledge, application in practice, and metacognition. The findings show that the ultimate goal of constructive interaction is not merely to receive an answer but to reconstruct knowledge in light of the learner’s own understanding and experience. Student-teachers should rewrite AI-generated content in their own language, add personal interpretation, integrate new information with prior knowledge, apply it in practical contexts, and reflect metacognitively on how the interaction affected their learning. This theme transforms AI use from consumption into knowledge construction. It aligns with Bloom’s higher-order cognitive processes, particularly evaluation, synthesis, and creation, and with the constructivist principle that meaningful learning occurs when new information is integrated into existing cognitive structures. The framework thus presents constructive interaction as a five-stage process—problem ownership, interaction refinement, critical evaluation, verification, and knowledge rewriting—that guides student-teachers from initial ownership of the learning problem to the reconstruction of knowledge.
Content Validity of the Framework
The content-validation results indicated that all 23 basic themes had acceptable content validity. CVR values were greater than 0.49, and CVI values were greater than 0.79. These findings indicate expert consensus on the necessity, relevance, clarity, and appropriateness of the framework’s components. The framework was therefore confirmed for the final version. The study’s findings are consistent with previous research on student–AI interaction patterns, critical AI literacy, and the importance of verification and knowledge reconstruction. They also extend existing work by organizing these components into a coherent five-stage framework specifically designed for student-teachers in the curriculum of Farhangian University.

Conclusion
This study designed and validated a framework for student-teachers’ constructive interaction with AI in the curriculum of Farhangian University. The framework consists of one global theme, five organizing themes, and 23 basic themes: problem ownership, interaction refinement, critical evaluation, verification, and knowledge rewriting. All basic themes demonstrated acceptable content validity based on expert judgment using CVR and CVI. The findings suggest that curriculum planners and educational stakeholders at Farhangian University should move beyond workshops and courses focused primarily on technical AI skills and instead emphasize the development of learners’ competencies for constructive interaction with AI. Specifically, specialized workshops should address information refinement, critical thinking, verification, and knowledge rewriting. Educational practitioners and curriculum designers should provide the conditions for implementing the five-stage framework as a guide for purposeful AI use in the curriculum. The study’s limitations include the limited number of participants and experts, the focus on a single educational and university context, the interpretive nature of participant and researcher perspectives, and the absence of testing the framework’s effectiveness in real educational settings. Future research should test the framework in actual teaching and learning environments and examine its impact on student-teachers’ AI literacy, critical thinking, and professional development.