Language Preferences and Practices in Multilingual EdTech: Flexible Primary Language Use with Secondary Language Support
Pith reviewed 2026-05-10 17:52 UTC · model grok-4.3
The pith
Learners who consistently used both English and their local language persisted longer in the hybrid EdTech course.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
In the quasi-experiment, learners who chose the hybrid option could navigate between English and Leb-Lango. While many did not consistently use both languages, those who did persisted longer in the course. Learners described their own strategies for handling language complexities. The study presents this as the first empirical evidence of learner agency in bilingual remote EdTech instruction.
What carries the argument
The hybrid language mode, defined as the learner-chosen combination of English and Leb-Lango, together with the distinction between consistent and inconsistent bilingual use within that mode, which is linked to differences in course persistence.
If this is right
- Among hybrid learners, consistent use of both languages correlates with longer course persistence.
- Learners demonstrate agency by deciding how and when to draw on each language in remote instruction.
- Multilingual EdTech designs should support flexible primary language use with secondary language support to accommodate varied practices.
- These patterns provide guidance for building inclusive solutions that integrate colonial and local languages without requiring uniform bilingual behavior.
Where Pith is reading between the lines
- EdTech platforms could add optional prompts that encourage but do not enforce consistent switching, testing whether this raises overall retention.
- Similar choice structures might produce parallel persistence patterns in other bilingual regions if offered at scale.
- Future work could track whether consistent hybrid use also improves learning gains, not just enrollment duration.
Load-bearing premise
That longer persistence among consistent hybrid users stems from their language practices rather than from unmeasured differences in motivation, prior knowledge, or platform engagement that also led them to use both languages.
What would settle it
A randomized controlled trial that assigns hybrid learners to consistent bilingual use versus free choice and finds no persistence difference after controlling for baseline motivation would falsify the claim that language practice drives the outcome.
read the original abstract
The benefits of learning in one's mother tongue are well documented, yet colonial languages dominate education, marginalizing local languages and limiting access for learners who rely on their mother tongue for understanding. With the rapid growth of educational technology, there is potential to integrate multilingual instruction supporting both colonial and local languages. This study is part of a larger quasi-experiment conducted in Uganda, where learners could choose to learn in English, Leb-Lango (a local language), or in Hybrid mode (a combination of both) in a remote EdTech course. We examined how learners who chose the Hybrid option navigated English and Leb-Lango. While many Hybrid learners did not consistently use both languages, those who did persisted longer in the course. Learners also shared how they managed language complexities. We provide the first empirical evidence of learner agency in bilingual remote EdTech instruction and offer insights for designing inclusive multilingual learning solutions.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper describes a quasi-experiment in Uganda in which learners self-selected into English-only, Leb-Lango-only, or Hybrid (bilingual) modes for a remote EdTech course. Among Hybrid users, those who consistently employed both languages showed longer persistence than other Hybrid users; the authors also report qualitative accounts of how learners managed language switching and present this as the first empirical evidence of learner agency in bilingual remote EdTech.
Significance. If the observed persistence difference can be shown to be attributable to language practices rather than to unmeasured differences in motivation or engagement, the work would supply useful field evidence for flexible multilingual designs in low-resource EdTech settings. At present the association remains observational and unadjusted, so the practical implications for platform design are still tentative.
major comments (3)
- [Abstract/Methods] Abstract and methods description: the central persistence claim is reported without sample sizes, statistical tests, or any adjustment for selection into consistent hybrid use. Because the design is self-selected, the difference cannot yet be isolated from confounding by baseline motivation, prior knowledge, or platform engagement.
- [Results] Results on hybrid navigation: no description is given of how language-mode usage was logged, coded, or thresholded to define 'consistent' hybrid use, nor of how persistence was operationalized (e.g., completion rate, time-on-platform). These measurement details are load-bearing for the reported association.
- [Discussion] Discussion of learner agency: the interpretation that consistent bilingual navigation reflects agency rather than other unmeasured traits rests on the untested assumption that selection effects have been ruled out; a sensitivity analysis or covariate adjustment would be required to support this causal framing.
minor comments (2)
- [Abstract] The abstract states 'we provide the first empirical evidence' without referencing prior related studies on bilingual EdTech; a brief literature placement would strengthen context.
- [Qualitative findings] Qualitative excerpts on language management are mentioned but not linked to specific participant identifiers or themes; clearer presentation of the qualitative coding scheme would aid reproducibility.
Simulated Author's Rebuttal
We thank the referee for the constructive and detailed feedback. We have revised the manuscript to address the concerns by adding missing methodological details, sample sizes, statistical information, and clearer operational definitions. We have also tempered the language around learner agency to emphasize the observational nature of the findings and added explicit limitations. Our point-by-point responses follow.
read point-by-point responses
-
Referee: [Abstract/Methods] Abstract and methods description: the central persistence claim is reported without sample sizes, statistical tests, or any adjustment for selection into consistent hybrid use. Because the design is self-selected, the difference cannot yet be isolated from confounding by baseline motivation, prior knowledge, or platform engagement.
Authors: We agree that the abstract and methods required more detail for transparency. In the revised manuscript, we have added sample sizes for the hybrid group and subgroups, the statistical tests performed (including p-values and effect sizes for the persistence comparison), and explicit description of self-selection into modes. As this remains a quasi-experimental study, we cannot perform full covariate adjustment without baseline data on motivation or prior knowledge. We have added a dedicated limitations section acknowledging potential confounding and now frame the persistence difference strictly as an observed association rather than an isolated causal effect. revision: partial
-
Referee: [Results] Results on hybrid navigation: no description is given of how language-mode usage was logged, coded, or thresholded to define 'consistent' hybrid use, nor of how persistence was operationalized (e.g., completion rate, time-on-platform). These measurement details are load-bearing for the reported association.
Authors: We have expanded the methods and results sections with these operational details. Language usage was automatically logged by the platform's backend, recording per-session language selections and content access. 'Consistent' hybrid use was defined as accessing both languages in at least 60% of sessions (threshold set via distribution analysis to separate regular from occasional users). Persistence was operationalized as the number of active days from first to last platform interaction, supplemented by module completion rates. These additions make the measurement process reproducible and the association more interpretable. revision: yes
-
Referee: [Discussion] Discussion of learner agency: the interpretation that consistent bilingual navigation reflects agency rather than other unmeasured traits rests on the untested assumption that selection effects have been ruled out; a sensitivity analysis or covariate adjustment would be required to support this causal framing.
Authors: We have revised the discussion to remove any causal framing of agency. The claim is now supported by the combination of the observed persistence association and qualitative accounts of deliberate language-switching strategies reported by learners. We explicitly note that selection effects cannot be ruled out and have added a limitations paragraph on unmeasured traits such as motivation. No sensitivity analysis was possible due to the lack of rich baseline covariates in the dataset; we instead highlight this as a direction for future research with more comprehensive data collection. revision: partial
- Conducting a sensitivity analysis or covariate adjustment for unmeasured confounders such as baseline motivation and prior knowledge, as these data were not collected in the original quasi-experiment.
Circularity Check
No significant circularity detected
full rationale
The paper reports observational and qualitative results from a quasi-experimental field study of learner language choices in a remote EdTech course in Uganda. No equations, fitted parameters, derivations, or predictive models are present that could reduce to their own inputs by construction. Claims rest on direct empirical associations (e.g., persistence differences among hybrid-mode users) and participant reports, which are independent of any self-citation chain or ansatz. The study is self-contained against its own data collection and does not invoke uniqueness theorems or rename prior results as new derivations.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption Platform logs and self-reports accurately capture whether learners consistently used both languages during the course
- domain assumption Longer course persistence reflects meaningful engagement rather than external factors unrelated to language support
Reference graph
Works this paper leans on
-
[1]
Asher, Dieyu Ouyang, Lingkan Wang, Debbie Eleene Conejo, John Stamper, Paulo F
Language Preferences and Practices in Multilingual EdTech: Flexible Primary Language Use with Secondary Language Support Christine Kwon*, Phenyo Phemelo Moletsane*, Michael W. Asher, Dieyu Ouyang, Lingkan Wang, Debbie Eleene Conejo, John Stamper, Paulo F. Carvalho, and Amy Ogan ckwon2@andrew.cmu.edu, pmoletsa@andrew.cmu.edu, masher@andrew.cmu.edu, dieyuo@...
work page 2017
-
[2]
and text message-based learning applications (Breazeal et al., 2016; Khan et al., 2019; Kizilcec et al., 2021; Wahyuni et al., 2024). This opens up opportunities for similar EdTech systems to include multilingual instruction with both colonial and local languages to reach multilingual remote communities. However, current EdTech solutions in remote and mar...
work page 2016
-
[3]
developed a localized intelligent tutoring system that taught programming using English-based Creoles, and conducted multiple case studies in Trinidad and Tobago showing that students preferred systems integrating cultural and linguistic context, and valued the ability to control the degree of localization via a “cultural density slider.” Similarly, Jantj...
work page 2012
-
[4]
introduced bilingual mobile learning tools to high school mathematics and science learners in South African schools, enabling students to access content in both English and local languages and create bilingual audio notes. Findings showed that most learners tended to use both languages to engage with content, reporting that flexible access to learning in ...
work page 2012
-
[5]
The course was delivered through a low-tech platform and designed to provide interactive STEM education to participants in rural Uganda, where access to the internet, smartphones, computers, and academic materials is often limited (Kwon et al., 2023). As such, it relied on two widely available tools in rural Africa to reach remote learners: basic keypad p...
work page 2023
-
[6]
Interviews averaged about 17 minutes, ranging from approximately 11 to 31 minutes
The course facilitators conducted semi-structured interviews in the local language, Leb-Lango, then translated all interviews into English and removed all personal identifiers. Interviews averaged about 17 minutes, ranging from approximately 11 to 31 minutes. We conducted a thematic analysis on the post-course interviews, following the thematic framework ...
work page 2006
-
[7]
Small increases were observed at Step 4, Step 6, and Step 7, but overall, language switching was less frequent in the later steps of the course. These results suggest that learners may have experimented with language choice at the beginning of the course, possibly as they explored which language felt more comfortable or effective for comprehension. As lea...
work page 2007
-
[8]
and multilingual mobile learning tools (Jantjies & Joy, 2012; Jantjies & Joy, 2013). Building on this foundation, prior literature has also underscored the need to better understand how multilingual instruction should be meaningfully integrated in EdTech. To address this gap, our study offers a more concrete approach, recommending that learners choose a p...
work page 2012
-
[9]
J., Azeem Abbas, M., & Rehman, A
Khan, S., Hwang, G. J., Azeem Abbas, M., & Rehman, A. (2019). Mitigating the urban–rural educational gap in developing countries through mobile technology ‐ supported learning. British Journal of Educational Technology, 50(2), 735-749. Kukulska-Hulme, A., Giri, R. A., Dawadi, S., Devkota, K. R., & Gaved, M. (2023). Languages and technologies in education ...
work page 2019
-
[10]
Ndebele, H. (2014). Promoting indigenous African languages through information and communication technology localisation: A language management approach. Alternation Special Issue, 13, 102-127. Nkoala, S. (2024). Educators’ Experiences of using multilingual pedagogies during emergency remote teaching: a case study of South African universities. Internatio...
work page 2014
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.