Scaling up SLA research in the classroom: Digitally supported randomized controlled field studies

Detmar Meures, University of Tübingen

Second Language Acquisition research has uncovered a range of factors that influence acquisition, with different strands of SLA highlighting different aspects - from the Input and Noticing Hypotheses to the Output and the Interaction Hypotheses, to name just a few of the ways research has focused on cognitive, linguistic, and social factors shaping language learning.

Arguably, any ecologically valid data on authentic language learning results from a combination of thesedifferent factors - but since different levels of granularity of data at different time scales are at stake, it is difficult for research to target the interaction of the factors shaping such authentic data. Inaddition, interaction effects require a substantial amount of data to be reliably identified, especially in noisy authentic data.  A factor making this even more challenging is the important role that individual differences play in SLA. As a result, many of the SLA hypotheses and potentially impactful research issues, such as aptitude treatment effects, remain empirically undersubstantiated and with limited impact on real-life teaching and learning.

But where could large scale longitudinal data on authentic learning processes and products come from? Over 20 million school children in Europe learn a foreign language in upper secondary schools! Yet children in the school setting are relatively underresearched in SLA,with legal and practical issues making it difficult to stage interventions and collect data at scale. Given the increasing use of digital tools in school, with a substantial boost during the Coronapandemic, a growing number of language learning contexts leave digital traces, be it when searching for and reading texts or when practicing with online activities. Adding AI methods to the mix to support the automatic analysis of language and learner modeling, it becomes possible to enhance digital practice environments so that they offer clear advantages to learners and teachers: from individually adaptive selection of motivating and input-enhanced materials to adaptive activity sequencing and automated scaffolded feedback during practice(Ruiz, Rebuschat & Meurers 2023). Complementing the practical benefits facilitating large-scale use, such digital environments can in principle be used to conduct year-long SLA experiments fully embedded in regular school contexts and provide detailed information on learners, learning processes, and products.

In this talk, I present our first steps in this direction, reporting on several randomized controlled field trials carried out in German schools using the Intelligent Tutoring System "FeedBook" to provide practice opportunities for high school students as part of the regularEnglish classes. The first field study (Meurers et al 2019) confirmed the effectiveness of specific scaffolded feedback provided during practice and illustrates some solutions to the challenges that arise when scaling up experiments to settings that entail a substantial loss of control, including the use of learning analytics to interpret learning process data (Hui, Rudzewitz, Meurers, in press). We then report on the first results from a second year-long field study in the Interact4School project exploring motivational feedback and a learner dashboard integrating practice as pre-task activities in a task context with criterial feedback. As an outlook we then sketch two current studies: The just completed DigBinDiff study illustrating the integration of ambulatory assessment of working memory to support adaptive activity sequencing based on linguistic and cognitive factors, and the AI2Teach study that will introducing a teacher dashboard and teacher training to link the individualized practice with the teacher-orchestrated classroom. Negotiations are currently under way to potentially support all schools in the state ofBaden-Württemberg. While the studies conducted so far only scratch the surface of the SLA research issues that could be investigated in such a setting, we hope it will encourage systematic collaboration onlarge-scale experimentation in authentic education settings, which we believe has the potential to empirically substantiate and advance complex SLA research hypotheses and highlight the relevance of SLA research for improving education.

References

Meurers, D., De Kuthy, K., Nuxoll, F., Rudzewitz, B., & Ziai,R. (2019). Scaling up intervention studies to investigate real-life foreign language learning in school. Annual Review of Applied Linguistics, 39, 161-188. 

Hui, B., Rudzewitz, B., & Meurers, D. (in press). Learning processesin interactive CALL systems: Linking automatic feedback, system logs, and learning outcomes. Language Learning & Technology. (preprint link)

Ruíz, S., Rebuschat, P., & Meurers, D. (2023). Supporting Individualized Practice through Intelligent CALL. In Suzuki, Y. (Ed.) Practice and Automatization in Second Language Research (pp. 119-143). Routledge (preprint link) https://doi.org/10.4324/9781003414643-7.