Key points are not available for this paper at this time.
Methodological Approach This article is an opinion-based conceptual piece that draws on a targeted selection of peer-reviewed sources to develop a conceptual discussion on digital anthropomorphism in generative AI tutors. To ground our argument in current scholarship, we searched Google Scholar, Scopus, and Web of Science for literature published between 2019 and 2025, using terms such as "AI trust," "digital anthropomorphism," and "generative AI in education." We focused on works that explicitly addressed human–AI interaction, trust psychology, or anthropomorphism in educational contexts, and excluded purely technical studies and noneducational applications. Approximately 45 relevant papers were identified. Rather than conducting a systematic review, we engaged in an informal thematic grouping of recurring ideas— such as perceived authority, emotional reassurance, automation bias, and epistemic vigilance— which informed the structure of this article. The aim here is not to provide exhaustive coverage, but to integrate converging insights from cognitive psychology, human–computer interaction, and educational technology into a coherent, opinion-driven perspective on trust calibration in AI-mediated learning. Introduction: When the Machine Feels Human Today's students interact more with generative AI tools like ChatGPT, Claude, and Google Gemini as conversational partners rather than as disembodied software. When these systems respond with fluency, politeness, and encouragement, they create a subtle but potent illusion: the AI appears to "understand" the user (Cohn et al., 2024; Karimova Placani, 2024). This article offers a conceptual, opinion-based synthesis of recent peer-reviewed literature on this topic, drawing on insights from cognitive psychology, human–computer interaction, and educational technology. Our aim is not to provide an exhaustive or systematic review but to integrate converging findings into a coherent framework for understanding trust calibration in AI-mediated education. We structure the discussion around the conceptual pathway illustrated in Figure 1, which traces how anthropomorphic design cues may foster affective trust, reduce epistemic vigilance, and influence learner dependency, while also considering contexts in which anthropomorphism can enhance engagement and confidence when ethically designed. The Cognitive Basis of Digital Anthropomorphism Digital anthropomorphism is not a failure of rationality, but rather a manifestation of human social cognition (Fakhimi et al., 2023). Developmental psychology has demonstrated that even children ascribe intention and moral status to animated forms if they move in goal-oriented manners. Adults too habitually treat chatbots, GPS, and voice assistants as being quasi-social actors—to thank them, apologize, or obey their instructions. Generative AI amplifies this impact with linguistic anthropomorphism. Its natural language proficiency activates people's social brain mechanisms— soliciting empathy, engagement, and even perceived moral agency (Alabed et al., 2022; Q. Chen Troshani et al., 2020), though its generalizability to all classroom settings remains to be confirmed. Such a 'confidence heuristic' is problematic when used with AI systems trained to optimize fluency and not epistemic truth. This aligns with findings by Atf and Lewis (2025), who demonstrate that user trust in AI systems is often driven by surface fluency and not correlated with explainability, especially in educational domains(Maeda, 2025). Figure 1. Conceptual synthesis — Psychological Pathway Linking Digital Anthropomorphism to Epistemic Vulnerability in AI-Mediated Learning. This diagram illustrates how interface design features that evoke human-like qualities can lead to affective trust, which in turn may reduce learners' epistemic vigilance, resulting in over-reliance, diminished critical thinking, and role confusion. Note: This is a conceptual synthesis derived from the thematic literature review and is not an empirically estimated model. Trust, Dependency, and the Erosion of Epistemic Vigilance From a psychological perspective, trust in learning is both required and dangerous. Students need to trust instructors to direct them, but they must also cultivate epistemic vigilance—the capacity to evaluate the believability of information sources. When students anthropomorphize AI tutors, their epistemic filters could weaken. Emotional trust in AI can be expressed as: • Over-reliance on AI feedback over teacher guidance. • Inadequate effort to cross-check or challenge AI-produced responses. • Acceptance of imperfect or slanted results, particularly if they come with persuasive voice (A. Chen Ryan, 2020), rather than as universally established findings. This quiet process from doubt to submission is a pivotal moment in the psychology of trust (Ryan, 2020). This accords with Pergantis et al.'s (2025) research, which shows that extensive AI interactions have the potential to move cognitive control processes underlying autonomous learning. Even though such flaws require close analysis, no less true is the fact that anthropomorphic indicators have, in certain scenarios, the potential to render useful pedagogical roles if appropriately and responsibly conceptualized. Productive Anthropomorphism and Ethical Design Although much of the debate about anthropomorphism in AI tutors centers on its possible dangers, it is valuable to note that human-like signals can have positive teaching outcomes as well, when implemented sensitively. Anthropomorphic design features can improve students' engagement, minimize feelings of loneliness in online classrooms, and give emotional comfort to students who are anxious or self-doubting. For instance, learners who have mathematical anxiety or who have limited exposure to human tutors may respond positively to an AI tutor's persistent, nonjudgmental feedback (Polydoros et al., 2025). Others who are shy or socially fearful may be more at ease conversing with an amicable AI interface than with colleagues or teachers in live classes. Ethical calibration is the answer: balancing motivational advantages of anthropomorphism with characteristics that maintain critical thinking and epistemic vigilance. Characteristics like that may be achieved through the use of transparency prompts, source citations that are visible, and infrequent "reflection nudges" which get students to stop and double-check information. In combination with instructional guidance, design strategies like these hold promise for making anthropomorphic cues function as a learning scaffold, not a shortcut to mere passive acceptance. Toward a Psychology of Critical Trust in AI Tutors To enter this psychological territory of anthropomorphism in the digital age, teachers must encourage students to cultivate critical trust—a mindset to be open to the affordances, yet cautious about the limits, of AI. Even technical literacy won't suffice; psychological sensitivity is needed (Mulcahy et al., 2023). Educational interventions might include: ➢ AI debriefs: Short reflection exercises that get students to present an AI-generated response that was utilized and answer three guiding questions: (1) What was the chief argument of the AI? (2) What sources, if any, did it reference? (3) How did you test or refute it? This helps students be mindful of their uses of AI intentionally. ➢ Counter-anthropomorphism exercises: Students reword an AI's polite, human-sounding response in purely technical terms, removing social signals. This helps students contrast how tone and style affect their perception of authority and reliability. ➢ Trust calibration training: Checklists or short classroom protocols that encourage students to ask, before accepting an AI's response: (1) Is there a legitimate source? (2) Is my explanation consistent with my prior knowledge? (3) Have I checked it elsewhere? This training induces the habit of separating interface ease from epistemic reliability. Educators can model critical trust through transparent and explainable use of AI in class, revealing its benefits and its limitations. Guided classroom debates about issues like algorithm bias, hallucinations, and surface fluency versus deep knowledge can "immunize" students against excessive faith. Classroom activities that engage students in collaborative tasks can further erode passive dependence: for instance, group debates where students are asked to argue against an answer generated by an AI, or collaborative projects where human and AI readings of the same content are evaluated side by side for nuance, tone, and cultural reference. These exercises tie directly to earlier interventions like AI debriefs, counter-anthropomorphizing, and calibration of trust, building upon them through active exercise. In the long run, establishing critical trust may even necessitate interface redesigns—with features like visible source quotation, easy-to-understand explainability tools, and interactive prompting that invite reflection before accepting an AI's answer. Research Pathways for Calibrating Trust in Generative AI Tutors Future research should explore the psychology of anthropomorphism in AI tutors across diverse educational contexts (Létourneau et al., 2025). We propose two complementary tracks: Track A – Affective Trust Calibration ❖ Investigate how learners distinguish between the emotional tone and epistemic validity of AI responses. ❖ Test interventions such as meta-cognitive prompts, counter-anthropomorphism training, and AI explanation auditing to determine their effectiveness in sustaining critical vigilance (Chakraborty et al., 2024; Israfilzade Sarfaraj1, 2025). Such responses can enrich the learning experience when they foster motivation, confidence, and a sense of social presence (Polydoros et al., 2025). However, they also carry the risk of distorting teacher–student dynamics and encouraging uncritical trust (Vanneste Yuan Mulcahy et al., 2023). In an algorithmically mediated educational future, the goal is to develop learners who can recognize when AI offers valuable support and when its persuasive surface masks the need for independent reasoning. Ultimately, critical trust allows students to use AI as a partner in learning without surrendering their intellectual autonomy (Ryan, 2020).
Jose et al. (Thu,) studied this question.