Teachers' basic psychological need satisfactions and their intentions to use generative artificial intelligence: the mediating role of engagement in [[educational-development|professional development]]

Synthesis: Yuan, Yang, Zhao and Li surveyed 146 teachers who had just completed a generative AI professional-development program and tested a self-determination theory model of their intention to keep using it. Only perceived competence in using GenAI predicted intrinsic motivation (β = 0.308); perceived autonomy (β = 0.177, p = 0.544) and relatedness (β = 0.392, p = 0.256) did not. Intrinsic motivation then predicted intention to use GenAI (β = 0.444) partly through engagement in professional development (indirect β = 0.233, 95% CI [0.025, 0.591]), which is the study's practical claim: feeling capable is what moves teachers, and what they do with that feeling in ongoing professional learning is what carries it into intention.

Key Findings

1. Competence, not autonomy or relatedness, moved intrinsic motivation. Perceived competence using GenAI predicted intrinsic motivation at β = 0.308 (p < 0.05), while autonomy (β = 0.177, p = 0.544) and relatedness (β = 0.392, p = 0.256) were not significant.
2. Motivation predicted intention directly and strongly. Intrinsic motivation had a direct effect of β = 0.444 (p < 0.001) on intention to use GenAI, with a total effect of β = 0.677 once the mediated route is included.
3. Engagement in professional development carried part of it. Motivation predicted engagement in professional development (β = 0.720) and engagement predicted intention (β = 0.324), giving a significant indirect effect of β = 0.233 (95% CI [0.025, 0.591]) — partial mediation, not full.
4. The model explained half the variance in intention. R² was 68.5% for intrinsic motivation, 51.9% for engagement in professional development and 50.9% for intention to use GenAI; overall fit was acceptable (CFI = 0.943, TLI = 0.930, RMSEA = 0.078, SRMR = 0.059).
5. Autonomy and relatedness were measured and came out null. The three need satisfactions were strongly correlated with each other (GAA–GAR HTMT = 0.905), so their separation was only qualified, which the authors acknowledge as a limit on the autonomy and relatedness nulls.
6. The sample is experienced, mainstream and small. 146 teachers, 63.7% with more than 10 years of experience and 90.4% in public schools, drawn from one vendor-run training program with no missing data.

Competence is the need that travels

Applying SDT to teachers' technology acceptance produces an unusually specific result: of the three basic needs, only competence predicted intrinsic motivation toward GenAI. The authors read that as GenAI being a skill-gap technology — teachers can imagine wanting to use it, but only feel motivated once they believe they can operate it — and they note the contrast with earlier technology-acceptance work where autonomy or relatedness sometimes carried the effect.

The autonomy null deserves its own reading. Teacher autonomy over AI use is a governance question as much as a psychological one, and a survey of teachers inside one program may not capture the institutional constraints that make autonomy salient. The authors treat the null as a finding about perceived autonomy within this program rather than about autonomy in general.

What engagement in professional development adds

The mediation matters for how professional development is designed. Motivation's effect on intention is not purely internal: 0.233 of the 0.677 total runs through engagement in further professional learning, meaning the motivated teacher who keeps participating in training is the one whose intention firms up. That makes sustained, differentiated professional development a mechanism rather than a support service.

The authors' recommendations follow the path: align training with teachers' own instructional needs, differentiate by technological literacy, and use collaborative practice so difficulty falls and the competence perception rises. The learning platform in the study, MindMate within MindGraph (https://mg.mindspringedu.com/), was built and operated by the research team, which is a design fact readers should weigh when interpreting a program-specific result.

What this design can establish

The study is a single post-program survey with no pre-measure and no comparison condition, so all paths are cross-sectional associations among self-reported perceptions. Fit indices are good (χ2(126) = 236.808, CFI = 0.943), and the measurement work is reported in detail — four items were dropped after CFA loadings fell below 0.40, leaving an 18-item instrument — but a well-fitting model of one time point cannot establish that competence causes motivation rather than the reverse.

Sample composition further narrows the reading. 102 of 146 participants were men and 52.1% were primary school teachers, in a convenience sample from one program in China. The authors' own proposals for follow-up focus on testing personal factors such as gender and digital literacy as moderators of the relatedness path.

What this means for practice

• Build perceived competence first. Demonstration, guided practice and success on real instructional tasks are the levers this model supports; messaging about GenAI's value targets motivation without the competence belief that actually predicted it.
• Differentiate training by technological confidence. The authors' second recommendation is explicitly tiered, because a teacher who cannot yet operate the tool is not in a position to develop autonomy with it.
• Treat continued participation as part of the effect. Since engagement in professional development carries a significant share of motivation's total effect on intention (0.233 of 0.677), one-off workshops leave the mechanism unfinished.
• Do not promise that autonomy support alone will raise uptake. In this sample autonomy and relatedness did not predict intrinsic motivation, so institutional autonomy without a competence route has no support here.

Limitations

• One post-program survey, no pre-measure or control group. Causal direction between need satisfaction, motivation, engagement and intention is not established; the authors call for longitudinal designs.
• Small convenience sample from a single program. 146 teachers from one vendor-run GenAI learning program in China, 69.9% men, with the platform built by the research team.
• Autonomy and relatedness were not cleanly separable. The GAA–GAR heterotrait–monotrait value of 0.905 exceeded the 0.90 threshold, so the null findings for those two needs rest on only qualified discriminant validity.
• All constructs self-reported on five-point scales. Intention to use GenAI is a stated intention, not observed classroom use.

Citation

Yuan, S., Yang, Y., Zhao, G., & Li, S. (2026). Teachers' basic psychological need satisfactions and their intentions to use generative artificial intelligence: the mediating role of engagement in professional development. Frontiers in Psychology, 17.