Abstract
The purpose of the two studies was to investigate self-efficacy development in a generative AI-based language learning experience.
Research Questions
RQ1
Are there significant increases in learners’ self-efficacy for language skills after 1 month of using AI-based features?
RQ2
What percentage of learners believe that AI-based features effectively support language learning?
RQ3
Do learners use what they learned from AI-based features outside the app, and if so, how?
Methodology
The study analyzed data from 385 learners with French or Spanish as their mother tongue.
- 1 Study 1: Existing premium subscribers already using AI features.
- 2 Study 2: Free users granted premium access (Novelty group).
Data sources: Background questionnaires, Pre/Post surveys (Likert scale), and 30-day usage tracking.
Participant Profile
Key Findings
The core finding is unambiguous: AI features significantly increased learner self-efficacy.
Positive Shift
Regression analysis revealed a significant increase in agreement with self-efficacy statements.
Real-World Application
For RQ3, the majority of learners reported using what they learned via AI features outside of the app environment, bridging the gap between digital practice and real-world usage.
"Findings are the first evidence that language learning self-efficacy increases in learning experiences that use generative AI."
The Novelty Effect
Study 2 participants showed significant increases in several types of self-efficacy, compared to Study 1's single type. This suggests a strong "Novelty Effect" for new users of the technology.
Reflection: Duocon 2025
I chose this article because of a recent event I attended: Duocon 2025. I have always been interested in the Duolingo app as a learning environment, especially in its recently introduced AI-enhancements such as generative features and personalised feedback. The findings are significant, providing the first evidence that language learning self-efficacy increases in learning experiences that use generative AI.
Research in SLA is vast. This study could have varied by language, app, or focus, yet its specific focus on Generative AI marks a crucial step in modern educational technology research.
The Nuance of Novelty
Clark (1983) & Cheung/Slavin (2013) Suggest that motivational boosts from novelty tend to diminish over time once the technology becomes familiar.
Radu (2014) Counters that if a tool is pedagogically sound, the effect isn't just novelty—it's about creating memorable, multisensory learning experiences that endure.