AI and the Heritage Language Classroom: Drawing Insights from Psycholinguistic Research
- Tania Lorena Leal, Ph.D., The University of Arizona
Recent decades of psycholinguistic inquiry into heritage languages have led researchers to an increased understanding of how speakers use their languages dynamically and how their capabilities can differ across linguistic domains. This talk will examine two specific areas: the lexicon (encompassing lexical access and processing) and morphosyntax (focusing on morphosyntactic processing). I will propose pathways through which generative AI might enhance heritage language instruction.
Psycholinguistic research has revealed that heritage speakers typically possess more restricted lexicons in their heritage languages and access these lexicons less quickly than other native speakers for whom said language is the dominant one (Hulsen, 2000). Crucially, however, Fairclough (2010) showed that instructional interventions can bolster vocabulary growth, while Garza (2013) documented that specialized vocabulary use can be expanded with instruction. Similarly, regarding morphosyntactic processing, Fuchs et al. (2015) found that heritage speakers show reduced accuracy in gender as compared to number agreement, while Montrul et al. (2014) indicated that the explicitness of a task can aid heritage speakers in recognizing discrepancies.
Building on these findings, I suggest ways in which instructors might fruitfully employ insights from psycholinguistic studies to create generative AI-driven interventions and activities that build on the unique contributions that heritage speakers bring to the classroom environment. Considering the swift advancements in generative AI, I underscore the urgent need for controlled classroom studies to assess the impact of AI-driven interventions both on linguistic development and on the development of self-sustaining strategies for heritage language educators and learners.
ChatGPt meets the heritage speaker: A classroom-based study on writing feedback
- Julio Torres, Ph.D., University of California, Irvine
- María Carina Saiidi Padilla, University of California, Irvine
Heritage speakers report writing as the most challenging skill for them to develop in their heritage language, often leading to feelings of anxiety (e.g., Callahan, 2010; Carreira & Kagan, 2011; Hedgcock & Lefkowitz, 2011). Studies have examined the impact of classroom-based instruction on writing development (e.g., Bowles, & Bello-Uriarte, 2019), as well as the role of peer review feedback (e.g., Jegerski & Ponti, 2014). However, very little is still known about the effects of corrective feedback on heritage speakers’ writing, and to date, no study has explored the use of generative AI as a source of writing feedback. In this classroom-based study, Spanish-English heritage speakers enrolled in a tailored writing course, Heritage Spanish: Latinidades en California, used ChatGPT to receive feedback on their first writing drafts for three argumentative essays. To assess the impact of such feedback on students’ writing experience and written language use, we collected data from four distinct sources: a questionnaire regarding students’ prior experience with ChatGPT, transcripts of ChatGPT’s corrective feedback, revisions made to their essays in response to ChatGPT’s feedback, and their reflections. Drawing from these sources, we conducted a preliminary analysis of the data using qualitative methodology to identify major themes. These findings will inform future empirical studies on writing feedback with generative AI in heritage language classrooms.
Explorations in the use of GenAI and other digital resources in a Russian heritage classroom
- Irina Dubinina, Ph.D., Brandeis University
This presentation invites attendees to consider the strengths and weaknesses of generative AI tools for teaching heritage languages, using Russian HL as an example. One of the most effective pedagogical uses of genAI in writing seems to be the detailed and varied feedback the bots can provide. The question this presentation will consider relates to the nature of this feedback. While the bots can give fairly good feedback on the tone and style of writing in English or Russian, they seem to know very little about Russian inflectional morphology and for the most part cannot provide accurate information to help Russian HL learners understand and practice grammatical endings. Yet, for most of our students, the greatest area of need is precisely the mechanics of writing, i.e., spelling and orthography which are based on morphological and phonological awareness. Since this knowledge needs to be developed through explicit form-focused instruction, genAI tools could also be asked to provide a different type of feedback: i.e., explanations of grammatical rules. Here as well the bots fall short on providing accurate and meaningful information, their output must be vetted by a full speaker of the language, at least for the time being, before it can be used in teaching and learning. The presentation will show and analyze examples of different types of feedback provided by two genAI tools, ChatGPT and Claude, and highlight potentially useful and harmful effects for HL learners. The presentation will also consider a different digital resource, not based on genAI, that may be more helpful in developing morphological awareness in Russian HL learners.
Generative AI, the Digital Divide, and Academic Literacy Development for Heritage Speakers
- Qian Du, Ph.D., University of California, Irvine
The advent of generative AI tools has sparked extensive interest and debate among educational stakeholders, due to its advanced capabilities in generating human-like text and the continuous enhancements seen in recent updates. While few would now deny the presence of generative AI in educational landscapes (whether we like it or not), concerns persist among many educators about the potentially negative impact of the technologies on the development of students’ critical thinking skills and overall learning.
On the other side of the debate are those who embrace these emerging technological tools and advocate for their full integration into pedagogical practices. Educators in this camp argue that banning generative AI tools for classroom settings not only is impractical, but also deprives students of unique learning opportunities that are only possible through active engagement with such technologies (e.g., critically analyzing features of human versus machine writing, identifying deception and bias within machine outputs, etc.).
What has been largely missing from the discussion is the issue of accessibility and equity concerning technology and learning, especially in relation to students from underprivileged backgrounds. It is commonly assumed that Generation Z students are all technologically proficient because they are born and raised in an era of rapid technological advancements. However, students from underprivileged backgrounds frequently face barriers to accessing the latest technologies, which in turn restricts their access to essential resources and support vital for enhancing learning.
In this presentation, I will share findings from three English academic writing classes, which enroll a substantial number of heritage speakers of Spanish and Chinese, about their interactions and experiences with ChatGPT for academic writing tasks. I hope to highlight the crucial role that educational institutions play in creating conducive and inclusive learning environments for students to explore emerging technological tools for enhancing academic literacy skills. Through such efforts, institutions can contribute towards narrowing the digital divide and empowering students to harness the latest technologies for academic success.