Imagined Robot Intelligence Linked to Teen Engagement Through Academic Self-Efficacy

Adolescents who read a written description of a hypothetical educational social robot rated its perceived intelligence, and those ratings were indirectly associated with their later learning engagement through academic self-efficacy, according to a three-wave longitudinal study published in BMC Psychology.

The study, led by Lun Xie and colleagues, collected data at three time points: November 2025, February 2026, and May 2026. Participants attended six collaborating secondary schools in Central, South, and North China, with convenience cluster sampling conducted at the class level. After data cleaning and longitudinal matching, 1,340 adolescents were retained for analysis, with a mean age of 16.27 years (SD = 1.34), comprising 620 boys and 720 girls.

A crucial design feature sets this study apart from typical human–robot interaction research. At the first time point, participants read a standardized written scenario describing a hypothetical educational social robot; no actual robot interaction occurred. The researchers explicitly caution that judgments formed from a written description should not be equated with perceptions arising during real human–robot interaction.

The measurement schedule was structured so that each construct was assessed at a specific wave. The first time point captured scenario-based perceived intelligence, baseline academic self-efficacy, baseline learning engagement, and demographic variables. The second wave assessed academic self-efficacy and physical activity, while the third wave measured learning engagement. Analyses relied on observed composite scores using confirmatory factor analysis, correlation analysis, and conditional process analyses based on PROCESS Model 14.

The five-factor measurement model showed good fit, with χ²(692) = 1232.953, χ²/df = 1.782, CFI = 0.968, TLI = 0.966, RMSEA = 0.024, and SRMR = 0.027. Before mediator and moderator terms were entered, scenario-based perceived intelligence at the first wave was positively associated with learning engagement at the third wave after covariate adjustment (β = 0.137, p < 0.001), a total effect that supported the study’s first hypothesis.

The pathway through self-efficacy emerged clearly in the component analyses. Perceived intelligence of the described robot was positively associated with academic self-efficacy at the second wave (β = 0.174, p < 0.001), and academic self-efficacy was strongly and positively associated with later learning engagement (β = 0.575, p < 0.001).

In the full conditional process model, however, the picture became more nuanced. The adjusted direct association between perceived intelligence and later engagement was not statistically significant (β = 0.038, p = 0.085), whereas the standardized indirect effect through academic self-efficacy was significant, with an effect of 0.100 and a 95 percent confidence interval of 0.074 to 0.128. In other words, the association with engagement operated through adolescents’ academic self-efficacy rather than as a direct link.

Physical activity, measured at the second wave, moderated the association between academic self-efficacy and learning engagement (β = 0.075, p < 0.001), indicating that higher physical activity was associated with a somewhat stronger self-efficacy–engagement relationship. The researchers note that this interaction explained only a small additional proportion of variance (ΔR² = 0.005). The index of moderated mediation was 0.013, with a 95 percent confidence interval of 0.005 to 0.022.

The authors emphasize the limits of what these findings can support. Among adolescents who evaluated a written hypothetical scenario rather than interacting with a robot, higher scenario-based perceived intelligence was positively associated with later learning engagement at the total-effect level and indirectly associated with engagement through academic self-efficacy, while the adjusted direct association in the full model was not significant. These findings concern anticipatory evaluations of a described educational robot and, as the study concludes, should not be interpreted as evidence of a causal psychological mechanism operating during actual educational human–robot interaction.

Subject of Research: Longitudinal associations among scenario-based perceived intelligence of an educational social robot, academic self-efficacy, physical activity, and learning engagement in adolescents.

Article Title: Scenario-based perceived intelligence of an educational social robot, academic self-efficacy, and learning engagement in adolescents: a three-wave longitudinal study

Article References: Xie, L., Tan, X., Ma, Y., Wang, J., & Pu, T. (2026). Scenario-based perceived intelligence of an educational social robot, academic self-efficacy, and learning engagement in adolescents: a three-wave longitudinal study. BMC Psychology. https://doi.org/10.1186/s40359-026-05466-6

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05466-6

Keywords: educational social robots, perceived intelligence, academic self-efficacy, learning engagement, adolescents, longitudinal study, physical activity, moderated mediation, China, BMC Psychology, Scenario-based, perceived