Bot or not? Materials & considerations for using AI in language instruction

On Wednesday, February 28th, 2024 the Davis Language Center (DLC), several language programs, and Academic Technology Services (ATS) joined forces to host the event “AI-Powered Language Instruction: Innovative Strategies and Applications” for the UC Davis community. The event focused on two main goals: firstly, to share our community’s collective knowledge on artificial intelligence (AI-) assisted language teaching and learning; and secondly, to create a supportive environment for instructors to experiment with AI prompt design by learning through trial-and-error and discussing their experiences with peers and technology experts.

This event was inspired by the preceding DLC talk “ChatGPT or 1984?: A Primer for Language Teachers”, presented by Dr. Robert Blake on October 25th, 2023. The initial event attracted language learners and instructors curious about the functionalities, affordances, and uncertainties generated by advances in AI. The follow-up workshop on February 28th, 2024 provided a space for the audience to continue the discussion, this time hearing from several voices about practical approaches to integrating AI in language learning and offering hands-on time to explore several tools.

The February 28th event began with mini presentations delivered by instructional designers, undergraduate and graduate students, and language program instructors. The panelists presented their experiences, insights, and uses of generative artificial intelligence in higher education, with a special focus on language classrooms. Find the presenters' materials using the links below.

The second half of the event included a brief guided discussion on AI, including what it is, a few examples of AI tools, and suggestions about access, security, and best practices. The co-presenters also offered a step-by-step process for engaging with AI tools from the perspective of an instructor (i.e. for lesson planning) or a student (i.e. for homework or studying). In what followed, attendees worked on their own, practicing with different AI tools. Access the interactive handout here.

While showcasing the potential applications of AI tools, the presenters also outlined some important limitations. One such limitation is that ATS has not yet formally vetted, nor approved, the use of AI tools in our classes. The vetting process considers vital risks that technology tools may pose to student privacy. Hence, UCD instructors would benefit from discussing their intended use with ATS professionals before implementing the tools. Additionally, Dr. Margaret Merrill recommended that instructors not require students to use the tools and foster open conversations with the students about AI, while also conducting research about the long-term impacts of using these tools to inform pedagogical decisions.

Another recurrent theme centered around the utility of AI and its role vis-a-vis that of instructors. Many of the panelists demonstrated cases in which AI cannot be trusted, such as Salvador García’s example of JotBot providing an incorrect metalinguistic explanation of a non-standard feature in Spanish writing. In scenarios such as this one, AI does not measure up to a seasoned instructor who can provide appropriate metalinguistic feedback to improve students’ language accuracy. The presenters stressed that AI tools do not really “understand” language with a semantic system, but rather represent a robust application of statistical co-occurrences of syntactic structures.

However, several panelists pointed to scenarios in which AI can function as a one-on-one conversational partner, augmenting a student’s language input and output. For example, Lillian Jones demonstrated how ChatGPT can help students notice and practice producing cases of the Spanish subjunctive mood. Similarly, Dr. Yutian Tan offered examples of utilizing ChatGPT to create fill-in-the-blanks exercises for an Advanced Chinese course, while simultaneously helping learners reflect on their own output and the feedback produced by AI. The panelists underscored the importance of effective prompting to ensure that AI assists with the task at hand. Furthermore, Sophia Minnillo emphasized that AI feedback should not replace instructor or peer feedback, but rather complement the two by allowing students to reflect on their writing at the drafting stage. It is safe to say that AI will not be replacing language instructors anytime soon, but teachers can find ways to augment their instruction by effectively prompting AI systems.

Overall, the notion of keeping the human at the center of AI tool development, usage, and policies was a prominent theme. As emphasized by the panelists, students must remember that AI is not a content expert and can generate inaccurate information. Human instructors are experts in their specific domain and can easily detect when AI has presented false information. Advanced students may also be capable of making these observations, but it is especially necessary for instructors (not AI) to provide content-expert guidance to early or younger learners. The Office of Educational Technology in the US Department of Education articulates this larger concept as “Center People” (Parents, Educators, and Students) as the number one principle for improving “opportunity, equity, and outcomes for students” (Cardona et al., 2023, p. 6). Keeping the human at the core of AI development and interaction has been prioritized, both by our panelists and the broader educational community.

This event’s panelists showcased the value of AI tools for personalized learning, immediate feedback, and lowering the affective filter–that is, making students feel more comfortable during language learning. Dr. Shagufta Fatema especially emphasized the value of AI for personalized learning and tailored instruction. Likewise, Dr. Kirsten Harjes underscored AI’s potential to help students overcome writer’s block by offering them a starting point for an essay. Students therefore can develop their confidence as writers in the target language by co-authoring with AI, leading to more successful individual writing in the future. The value of AI’s availability and timeliness, as a conversation partner and feedback provider, was stressed throughout the event.

Instructor materials:
All of the panelists shared their presentations, offering snippets of diverse and actionable language learning activity ideas, as well as resources and survey data. Take a look and feel free to adopt and/or adapt these activities in your own classes!

 Title & Link

In closing, we wish to emphasize that this blog post is intended to continue a discussion that the DLC events have facilitated. The events have demonstrated that AI can assist instructors in their teaching by co-creating/brainstorming a selection of activities and materials that instructors may choose to adopt/adapt for their specific context. Some of our presenters highlighted these different techniques, along with a variety of AI platforms and prompts. AI tools can support both students and instructors alike, although they still come with limitations that we must consider. These limitations include ethical concerns such as plagiarism, reliability, accessibility, privacy, and utility. Many of the presenters have begun to answer these questions through their examples, but the conversation does not end there.

When we think of the future of AI, we want to think of it as a potential tool to enhance learning, not as a forced component in the classroom setting. As instructors, we understand that learning practices change as new teaching models emerge, pushing a new era of innovation and accessibility. We leave for future debate some questions that may help to guide your journey with AI:

• How can we navigate the challenge of providing equitable access to AI education without requiring it from those with limited access, ensuring everyone has the opportunity to learn and benefit from these tools?
• How can teachers use the skill of detailed and context-sensitive AI prompting to enhance students’ language abilities? And, what lessons can instructors draw from unsuccessful AI interactions to improve their own communication with students?

Please contact the authors for questions or comments.

Lillian Jones, PhD Candidate, Spanish Linguistics
liljones@ucdavis.edu

Sophia Minnillo, PhD Candidate, Linguistics
smminnillo@ucdavis.edu

Salvador García, 4th Year Undergraduate Spanish and Psychology Double Major
salgarcia@ucdavis.edu

References

U.S. Department of Education, Office of Educational Technology, Artificial Intelligence and Future of Teaching and Learning: Insights and Recommendations, Washington, DC, 2023. https://www2.ed.gov/documents/ai-report/ai-report.pdf