Jurnal Pendidikan Karakter
Vol. 17 No. 1 (2026)

Digital affective-ethical internalization media in the Curriculum of Love: Formative PLS-SEM evidence from teachers and parents

Hermanto (Universitas Muhammadiyah Pontianak)
Rini Elvri (Universitas Muhammadiyah Pontianak)
Dewi Yanti (Universitas Muhammadiyah Pontianak)



Article Info

Publish Date
19 Oct 2026

Abstract

Digital transformation expands school–family communication but does not automatically strengthen value internalization. This study tested the relationships among digital affective-ethical internalization media (IAEM Digital), the Curriculum of Love, and school–family character partnership, and examined whether the relationship pattern could be compared across teachers and parents. A cross-sectional survey involved 91 teachers and 150 parents from three senior secondary schools in Pontianak and Kubu Raya, Indonesia. The three constructs were specified formatively and estimated using Mode B PLS-SEM. The teacher model was assessed with 5,000 bootstrap subsamples, redundancy validity, f² effect sizes, and Q²predict/PLSpredict. The parent model was estimated separately, followed by MICOM and multigroup analysis with 1,000 permutations. In the teacher model, IAEM Digital positively predicted the Curriculum of Love (β=.506, p<.001) and character partnership (β=.272, p=.017), while the Curriculum of Love positively predicted partnership (β=.420, p<.001). The indirect effect was significant (β=.213, p=.001), with R² values of .256 and .366. Redundancy coefficients of .720–.779 and positive Q²predict values supported the teacher composites and their predictive relevance. The parent model produced high R² values but also severe inner collinearity (VIF=29.946), unstable coefficients, and weaker predictive error than the linear benchmark. MICOM did not establish compositional invariance for IAEM Digital; therefore, MGA results were treated as diagnostic rather than inferential. The findings support partial mediation by the Curriculum of Love from the teacher perspective and demonstrate that cross-informant comparisons require stronger measurement equivalence.

Copyrights © 2026