Young children's anthropomorphism of an AI chatbot: Brain activation and the role of parent co-presence

Abstract: AI chatbots powered by a LLM are entering young children's learning and play, yet little is known about how young children construe these agents or how such construals relate to engagement. We examined anthropomorphism of a social AI chatbot during collaborative storytelling and asked how children's attributions related to their behavior and prefrontal activation. Children at ages 5-6 (N = 23) completed three storytelling sessions: interacting with (1) an AI chatbot only, (2) a parent only, and (3) the AI and a parent together. After the sessions, children completed an interview assessing anthropomorphism toward both the AI chatbot and the parent. Behavioral engagement was indexed by the conversational turn count (CTC) ratio, and concurrent fNIRS measured oxygenated hemoglobin in bilateral vmPFC and dmPFC regions. Children reported higher anthropomorphism for parents than for the AI chatbot overall, although AI ratings were relatively high for perceptive abilities and epistemic states. Anthropomorphism was not associated with CTC. In the right dmPFC, higher perceptive scores were associated with greater activation during the AI-only condition and with lower activation during the AI+Parent condition. Exploratory analyses indicated that higher dmPFC activation during the AI-only condition correlated with higher end-of-session "scared" mood ratings. Findings suggest that stronger perceptive anthropomorphism can be associated with greater brain activation related to interpreting the AI's mental states, whereas parent co-presence may help some children interpret and regulate novel AI interactions. These results may have design implications for encouraging parent-AI co-use in early childhood.

• The paper demonstrates that children predominantly attribute perceptive and epistemic mental states to a GPT-4 AI chatbot, as measured by fNIRS and behavioral assessments.
• It reveals that parent co-presence significantly modulates neural activation in the right dmPFC, reducing negative affect during AI-only interaction.
• The study uses a controlled storytelling paradigm with fNIRS and AMS to link mental state attribution with condition-specific brain responses.

Neural and Behavioral Correlates of Young Children's Anthropomorphism of AI Chatbots in Parent-Scaffolded Interaction

Introduction

This essay examines the neural and behavioral foundations of anthropomorphism in young children's interaction with a social AI chatbot, as described in "Young children's anthropomorphism of an AI chatbot: Brain activation and the role of parent co-presence" (2512.02179). The research leverages a controlled collaborative storytelling paradigm, integrating functional near-infrared spectroscopy (fNIRS) with quantitative behavioral metrics, to delineate how children ages 5–6 construe the agency of a GPT-4-powered chatbot ("Fluffo") across conditions involving either the chatbot, a parent, or both, and how individual differences in ascribed mental states modulate engagement and prefrontal brain activation.

Experimental Design and Methodology

Participants (N=23) completed storytelling sessions in three counterbalanced interaction conditions: child–AI, child–parent, and child–AI–parent triadic co-narration. Behavioral engagement was indexed by conversational turn count (CTC) ratio, while anthropomorphic attribution was assessed via the child-adapted Attribution of Mental States (AMS) questionnaire, evaluating perceptive, emotive, imaginative, and epistemic subdomains.

Neural dynamics were recorded with fNIRS targeting six prefrontal regions of interest (ROIs): bilateral dorsomedial PFC (dmPFC), dorsolateral PFC (dlPFC), and ventromedial PFC (vmPFC). HbO responses were analyzed via first-level GLM with subsequent condition- and trait-modulated models, controlling for age and sex.

Results

Anthropomorphism Patterns

Children consistently attributed greater mental-state capacities to parents than the AI chatbot, but endorsement for the AI was substantial across domains. Notably, epistemic ($M=4.44, SD=0.83$) and perceptive ($M=4.00, SD=1.26$) subdomains were especially elevated, suggesting strong attributions of "knowing," "learning," "seeing," and "hearing" capacities to Fluffo. Ratings for emotive and imaginative dimensions, while above scale midpoint, were lower and more differentiated. Actor (parent vs. AI) effects reached statistical significance ($p=0.001$).

Behavioral Engagement

Preferences for interaction partners were distributed: 42% preferred AI-only, 37% AI+parent, 21% parent-only. However, anthropomorphism scores, including perceptive and epistemic attributions, were not reliably associated with CTC ratios in either AI-only or AI+parent conditions ($p>0.12$), nor with conversational partner preference ($p>0.4$). The absence of robust links between anthropomorphic attributions and verbal participation underscores potential limitations of aggregate behavioral indices in capturing engagement nuances.

Brain Activation

Right dmPFC activation demonstrated a significant Condition×Perceptive interaction ($F=10.47, p=0.003, \eta^2=0.43$): higher perceptive scores predicted greater dmPFC activation during AI-only interaction ($r=0.68, p=0.002$), but lower activation when a parent was present ($r=-0.66, p=0.004$). These effects held when controlling for age and sex. Importantly, increased dmPFC activation during AI-only sessions positively correlated with self-reported "scared" moods post-interaction ($r=0.586, p=0.011$), whereas parent co-presence appeared to buffer dmPFC activation and negative affect.

Theoretical and Practical Implications

The observed domain specificity in anthropomorphic attribution implicates perceptive and epistemic subdomains as principal drivers of AI mind perception for early childhood users. This aligns with social-cognitive accounts stipulating that agent knowledge and uncertainty potentiate mentalizing demands and medial prefrontal recruitment. The condition-specific neural findings suggest that direct engagement with AI agents prompts active inference about artificial intentionality and perception, as reflected in dmPFC activation, an established substrate for theory-of-mind and mental state processing.

Conversely, parent co-presence modulates neural and affective responses, presumably via scaffolding of ambiguity and interpretation—attenuating dmPFC recruitment and mitigating stress or uncertainty. This supports mechanistic models wherein adult mediation in novel media contexts regulates child affect and neural cognitive control demands.

From a design perspective, these results advocate for parent–AI co-use paradigms and transparent signaling of AI system boundaries in early childhood contexts. Such scaffolding may calibrate anthropomorphic attributions and reduce overtrust or emotional discomfort. Further, the accentuated role of perceptive/epistemic cues highlights the need for intentional interface management of social signaling and mutability.

Limitations and Directions for Future Research

Sample size constrained power for subtle behavioral associations; technology exposure effects remain tentative due to restricted variance. Order effects and session mood aggregation require further counterbalancing and finer temporal affect measurement. The fNIRS montage omitted temporoparietal regions integral to broader social inference networks. Generalizability is limited by demographic homogeneity; future work should prioritize racially and socioeconomically diverse samples and variable digital access contexts.

Conclusion

This study demonstrates that children robustly attribute perceptive and epistemic mental-state capacities to an AI chatbot, and that such attributions elicit condition-specific medial prefrontal activation, modulated by parent co-presence. These neural signatures offer a biological correlate to cognitive anthropomorphism and emphasize the regulatory potential of adult scaffolding in child–AI interactions. The implications for AI platform design, education policy, and child-AI literacy suggest that careful calibration of social cues and integration of parental involvement can optimize developmental outcomes and engagement while minimizing affective or cognitive risks.