https://ift.tt/JgTnILO
Exploring Materials Creation with ChatGPT for a Multilingual Professional Development Workshop
By Betsy Kells, Technology Director, Penn Language Center; Ed.D. student in Educational Linguistics, Graduate School of Education and Siyi Yan, Graduate Assistant, Penn Language Center; M.S.Ed. student in Learning Sciences & Technologies, Graduate School of Education, University of Pennsylvania
.
DOI: https://www.doi.org/10.69732/NPBU8026
In the professional development context for world language instructors, addressing AI and its myriad applications in language education is growing in importance for many reasons, among which is its ubiquitous presence in educational technologies and varied performance in languages other than English. Working with a group primarily composed of Less Commonly Taught Language (LCTL) teachers at the Penn Language Center in Fall 2024, we explored how AI technologies could be leveraged to enrich our professional development workshops series for instructors of world languages. This piece reflects on our iterative prompting process for using ChatGPT to create engaging, multilingual lesson plans that resonate with teachers, connect to their pedagogical practices, and enrich professional development opportunities.
Moving Beyond English as the Lingua Franca
Our journey started in the fall 2024 when designing and delivering a two-part workshop series on Task-Based Language Teaching for LCTL education. After the first workshop in September, while successful, we noticed a need for more linguistically and culturally contextualized examples to help teachers apply the TBLT concepts to their specific languages, learners, and contexts. In particular, one of the workshop participants observed that while it is helpful to know what such activities look like in English, the common language among all of the 15 participants, it would be even more beneficial to see TBLT lesson plans in the target language. As a result, in the subsequent workshop that focused on TBLT for novice-level language learners, we gave ourselves the goal to provide any lesson plan or activity examples in the target language of the teacher-participants. Doing so allowed us to foster a collaborative, multilingual experience that would deepen participants’ understanding of TBLT concepts and its possibilities. We wanted to use a tool that everyone could access, so we selected ChatGPT as it was well-suited to make this experience more engaging and dynamic by embracing its capacity in many languages. Lastly, we chose a lesson topic that is likely found in most curricula, the weather, and prompted ChatGPT to create lesson plans in each of the languages of the anticipated attendees. (Lee & Ta, 2023)
Iterative Prompting for Best Results
ChatGPT was used most extensively in helping us generate multilingual lesson plans that participants used as a starting point to think of how TBLT can be incorporated in the specific language that they teach. The result was novice-level lesson plans across 23 languages, guided by ACTFL descriptors. In this section, we will go into detail about the iterative prompting process that gave us the final lesson plans that we shared with the workshop participants. Before moving ahead, we would like to acknowledge that there is not parity among the languages for which ChatGPT generates output, and for low-resource languages, the process typically utilizes English-language content to provide a response in the target language (Lee & Ta, 2023). There may be technologies which are better-suited for particular languages, but as mentioned above, ChatGPT was the best option for our goals for this workshop due to its low barriers to entry.
To start, we probed its familiarity with ACTFL standards through a broad prompt. The initial output provided a generic framework centered on basic topics like greetings, which included a lesson plan featuring objectives, required materials, and suggested activities. An example Italian lesson plan was provided, followed by a word-by-word substitution guide for adapting the framework to other languages. In reviewing this output, we believed that a theme of “greetings”, while fundamental, is often too simple and basic for novice learners in real classroom settings and may not be suitable for the TBLT activity planned for the end of the workshop. Also, such a straightforward framework might not sufficiently challenge learners or promote meaningful engagement or cultural connections. This reflection prompted us to refine our prompt, shifting towards creative, theme-based activities that offered a balance of engagement and slight challenge. This resulted in varied lesson plans on topics such as shopping and weather, with distinct cultural elements and skill focuses tailored to each language.
Prompt 1:
I’m working on a workshop with my less uncommonly taught languages teachers, I’d like you to come up with exercises and lesson plans for ACTFL descriptors (novice-level learners) for each of those languages I’m targeting at: 1. Italian 2. Spanish 3. German 4. Kazakh 5. Hindi 6. Arabic 7. Bengali 8. Mandarin 9. Brazilian 10. Portuguese 11. Thai 12. Vietnamese 13. French 14. Indonesian 15. Swahili 16. Malagasy 17. Zulu 18. Igbo 19. Turkish 20. Persian 21. Kurdish 22. Russian 23. Amharic |
Output 1:
To design exercises and lesson plans based on the ACTFL (American Council on the Teaching of Foreign Languages) novice-level proficiency descriptors, it’s essential to focus on the key areas of listening, speaking, reading, and writing at the novice level. These learners are typically able to communicate using short, simple phrases and can understand and produce memorized phrases, familiar vocabulary, and basic grammatical structures. Here’s a framework for each language targeting novice learners, along with example activities that can be adapted for each of the 23 languages. I’ll provide one example and then similar plans can be created for each language. Lesson Plan Framework: Novice Level (ACTFL Descriptors)Lesson Objective:By the end of this lesson, students will be able to:
Example: ItalianTopic: Introductions and Basic Greetings Materials: Flashcards, worksheets with greetings, simple dialogues, and vocabulary Warm-Up (10 minutes):
Presentation (15 minutes):
Guided Practice (20 minutes):
Application (10 minutes):
Wrap-Up (5 minutes):
Adapting for Other Languages:
These lesson plans can be tailored to the cultural contexts of each language, incorporating relevant names, places, and greetings. The idea is to build a foundation of simple conversational skills that align with the ACTFL novice-level descriptors. |
Prompt 2:
too lazy!! Let’s think about more different activities for different languages |
Output 2:
Let’s get creative with a variety of activities for each language! I’ll mix up the formats to make sure the lessons are diverse and engaging, considering cultural elements, communicative goals, and the ACTFL novice-level focus on basic phrases, vocabulary, and interaction. 1. Italian: Visual Vocabulary Match
2. Spanish: Market Role-Play
3. German: Scavenger Hunt
4. Kazakh: Simple Song and Movement
5. Hindi: Colors and Shapes Bingo
6. Arabic: Calendar Time
7. Bengali: Greeting Chain
8. Mandarin: Number Relay
9. Brazilian Portuguese: Memory Card Game
10. Portuguese (Portugal): Street Signs
11. Thai: Emotion Flashcards
12. Vietnamese: Daily Routine Timeline
13. French: Simon Says
14. Indonesian: Family Tree
15. Swahili: Weather Report
16. Malagasy: Counting with Seashells
17. Zulu: Traditional Clothing Vocabulary
18. Igbo: Classroom Object Identification
19. Turkish: Food Menu Role-Play
20. Persian: Object Description
21. Kurdish: Directions Game
22. Russian: Grocery List
23. Amharic: Greeting Relay
These activities vary by theme (e.g., shopping, weather, introductions) and skill (speaking, listening, cultural awareness), creating a rich array of experiences for novice learners across the 23 languages. |
Recognizing the need for more engaging and diverse content, we refined the prompt to request creative, theme-based activities. This resulted in varied lesson plans on topics such as shopping and weather, with distinct cultural elements and skill focuses.
In order to address the issue of real-time collaboration across languages during the workshop, we decided to center the lesson plans on the shared theme of “weather.” This way, even though teachers would be reviewing different activities in their respective languages, they could still engage in meaningful conversations about a common topic. Prompts were then refined to include clear expectations: variety in the activities, use of a communicative approach, and a central focus on the theme of weather. These steps support the culminating workshop activity in which participants were to transform a lesson plan from standard Communicative Language Teaching (CLT) to TBLT.
Further Refining the Prompts for Lesson Plans Across Languages
At this stage, ChatGPT was asked to generate several lesson plans in English, which we could easily review for clarity and structure before moving on to translations in the target languages.
Prompt 3: Good! Now here’s what I’m thinking: I want the novice plans to be based on the same theme: weather, but with a variety of activities; Instead of a TBLT lesson plan, make it similar to a 50-min CLT lesson plan (so teachers can practice how to transform a CLT lesson plan to a TBLT one); The most important thing is the lesson plans must be communicative! Once you give me the English version, I’ll later have you translate them into their target language version so I can have the teachers evaluate how ChatGPT does in helping them generate the lesson plan. I know you have a word count limit, so don’t worry if you can’t get me everything at one time |
Output 3:
Got it! I’ll generate individual, communicative weather-themed lesson plans for each target language, keeping each plan distinct and tailored to the language and cultural context. I’ll start with a couple, and we can go from there. 1. Italian: Weather Report Role-PlayTheme: Weather
Materials Needed:
Lesson Breakdown:1. Warm-up (5 minutes):
2. Vocabulary Introduction (10 minutes):
3. Guided Practice (15 minutes):
4. Role-Play (15 minutes):
5. Wrap-Up and Review (5 minutes):
2. Spanish: Weather DialoguesTheme: Weather
Materials Needed:
Lesson Breakdown:1. Warm-up (5 minutes):
2. Vocabulary Introduction (10 minutes):
3. Guided Practice (15 minutes):
4. Dialogue Practice (15 minutes):
5. Wrap-Up and Review (5 minutes):
Let me know how these feel, and I can continue with other languages! |
When the AI-generated French lesson plan felt overly simple, for instance, we asked ChatGPT to revise it, incorporating feedback on areas such as simplicity and cultural integration. This iterative process helped refine subsequent plans. The updated French lesson plan now explores weather across different regions in Paris, while the Zulu lesson plan features a weather-informed clothing activity highlighting local geography and traditional attire. The text box below demonstrates a prompt requesting that ChatGPT enhance the activity in the French lesson plan to make it more engaging. The revised lesson plan includes a Parisian weather report activity, which we praised for successfully incorporating cultural elements such as landmarks into the design.
Prompt 4:
The French one looks a bit boring, would you mind revising it before we continue? |
Output 4:
Sure, let’s add some more engaging activities for the French lesson plan: 12. French: Exploring Parisian WeatherTheme: Weather
Materials Needed:
Lesson Breakdown:1. Warm-up (5 minutes):
2. Vocabulary Introduction (10 minutes):
3. Interactive Map Activity (15 minutes):
4. Create a Parisian Weather Report (15 minutes):
5. Wrap-Up and Review (5 minutes):
I hope this revised version adds a bit more excitement to the lesson! Let me know if you have any other ideas or if you’d like to continue with more lesson plans. |
From this point, ChatGPT initiated a check-in after every three or four lesson plans. We were initially surprised to see it asking for feedback, but took the opportunity to encourage more novelty in the lesson plans. Subsequently, since each lesson plan appeared to have clear objectives, materials needed, and clear lesson breakdown, we proceeded to generate the rest in the same manner.
After finalizing all the English lesson plans, we instructed ChatGPT to translate them into their respective target languages. When we initially attempted to translate the plans all at once, upon reviewing the translation, we noticed that words and sentences in the translated version differed from what we had anticipated. To confirm our suspicion, we backward-checked the English versions of the translated plans and found that the outputs were indeed different lessons. To address these inconsistencies, we opted to feed in each copy-pasted lesson plan one at a time and have ChatGPT translate them individually into the target language.
This iterative approach further revealed that translations were of higher quality when ChatGPT worked on them one by one rather than in batches. By refining this step, we were able to ensure the translated lesson plans aligned with the original English versions, preserving their objectives, content, and structure. This precision was crucial to avoid confusion during the workshop and highlighted the importance of monitoring AI-generated outputs, especially for multilingual content. By prioritizing accuracy and consistency, we ensured the lesson plans met instructional goals and set a standard for responsibly integrating AI into language education.
Using the Lesson Plans During the Workshop
On the day of the workshop, we let participants know that we addressed their prior request to have lesson plans in the target language, and that we did so by utilizing ChatGPT as a lesson plan creator and translator. We distributed both the English and translated versions to workshop participants, and referred to the English versions of the lesson plans due to it being our common language. The reasons for also including the English version were twofold: first, to leverage English as the group’s lingua franca, and second, to allow for collaboration across languages, with the English-language version serving as a reference in pair work discussions in which the participants did not teach the same language. Lastly, the finalized lesson plans were compiled into a single digital document for easy distribution during the workshop. Readers can find a link to the lesson plans above.
The participants’ reactions to using the ChatGPT generated lesson plans as part of the workshop activity were indeed diverse and thought-provoking. Some approached the plans critically, using them as a springboard to reflect on their own teaching practices and push back on stereotypes found in the output; for instance, a French instructor noted the need to situate activities within the broader Francophone world rather than focusing solely on Paris. Similarly, ideas like inviting students to choose a city or several cities on a map of a place where the language is spoken emerged and enriched our discussion. Others expressed curiosity about the prompts used to generate the lesson plans as they wanted to understand how specific inputs influenced the output we received. Interestingly, a few participants skipped the review process altogether and asked ChatGPT to generate new lesson plans during the workshop. It is possible that they wanted to compare the CLT and TBLT lesson plans that ChatGPT would generate.
We found that the multilingual nature of the workshop added a unique emotional and intellectual dimension to the experience, especially when participants felt a sense of validation in seeing their LCTLs represented in lesson plans created with cutting-edge AI technology. When we followed up with the workshop participants via email to get their take on the use of AI/ChatGPT to create the workshop activity materials, we identified two themes across their responses. First, instructors thought that the lesson plans were a great starting point for creating task-based lessons, which was particularly productive when colleagues teaching the same target language were collaborating during the workshop. Second, some instructors felt that the output in their target language required more effort to make it suitable for students. We will go into more detail in the following paragraphs.
Instructor Feedback on ChatGPT Lesson Plans in the Target Language
We emailed all 22 participants to ask for their feedback on the ChatGPT lesson plans, and three people responded. Two are instructors of Mandarin, and the other is a teaching assistant for Vietnamese. Since this is a small percentage of those who attended, the views expressed might not reflect everyone’s attitudes towards the ChatGPT-generated target language lesson plans. However, we found that these written testimonials were reflective of the discussions that took place during the pair and group activities in the workshop.
On the topic of effective collaboration in the target language, for example, one Mandarin instructor found the AI-generated lesson plans to be a valuable resource for understanding TBLT principles. She shared that the structured examples provided by ChatGPT offered a clear framework for tasks that supported one another, making it easier to adapt the plans to her students’ needs. “Although the tasks themselves were not overly complex,” she remarked, “they offered a solid framework that I could easily adapt and innovate upon based on the needs of my students.” This adaptability indeed proved to be particularly useful during the workshop’s collaborative activities, where instructors build upon these foundational ideas to develop personalized approaches.
Another Mandarin instructor echoed this sentiment, sharing that “the activity it (ChatGPT) suggested was engaging and inspired me to enhance the plan with our (me and my colleagues’) own experiences and ideas.” This highlights the potential of AI to create target-language specific materials that are meaningful even when there is a different shared language, and brings us closer to centering multilingual instructors’ expertise rather than relying on English as the sole medium of communication.
In terms of challenges faced in the effective integration of AI outputs into the classroom, while instructors generally appreciated the convenience and structure provided by ChatGPT, during the workshop several noted that the outputs required substantial adaptation to suit their target languages and students. This was particularly evident for languages that ChatGPT appears less equipped to handle, such as Vietnamese. A Vietnamese teaching assistant shared that while ChatGPT provided helpful ideas, its output often needed adjustments before she could use it with the students. “For TBLT, it seems that ChatGPT may have abundant resources with more popularly-taught languages like English, Spanish and Mandarin,” she explained. “Many of these could be applied to the language I am teaching (Vietnamese), however, it often needs quite a few adjustments from the teacher’s side to make it usable in class.” She further noted that while ChatGPT could effectively break down authentic materials in Vietnamese, it struggled to “reproduce graded versions of texts,” by which she means simplified versions using lower-level vocabulary and structures. This limitation further highlights an area where human expertise remains indispensable in leveraging the use of AI tools such as ChatGPT.
The feedback from these instructors illustrates how AI can be both a powerful resource and a tool that still requires human expertise and creativity. On the one hand, its structured outputs offer solid foundations that save time and foster collaboration. On the other hand, its limitations, especially for less commonly taught languages, underscore the need for ongoing teacher evaluation and adaptation. In all feedback, instructors expressed enthusiasm in the potential of incorporating AI into their future work. Paired with an understanding of its limitations, it is clear that professional development that empowers educators to engage with these tools critically and effectively is urgently needed as we co-create the future of language education with AI innovation and teacher expertise.
Conclusion
Looking ahead to more PLC professional development workshop opportunities in Spring 2025, we continue a three-part series called Mastering the Scales of AI Literacy in which we will model some of the iterative prompting that was featured in this article, both for spontaneous oral production and a multi-stage writing process. Having first-hand experience with the output in different languages will assist us in advising instructors as they develop prompts for different tasks in the target language. Ultimately, in developing the prompting skills for their own purposes, instructors are better positioned to teach their students the art of prompt engineering in the TL, thus handing them the keys to drive their own learning. After this academic year, we intend to design workshops in which teachers themselves go through the process of generating lesson plans using AI to further enhance their skills and autonomy in lesson design. There is also potential for similar workshops in other instructional areas to allow educators to leverage AI in developing a wide range of educational materials. While the inclusion of AI in professional development activities could allow for more autonomous use of AI in the future, the limitations of AI tools must be addressed so that instructors can make informed decisions about when and how to use it.
In conclusion, utilizing ChatGPT to generate lesson plans in the target language of each of the instructor-participants at our recent workshop was an effective way to incorporate AI in a professional setting. In this article we outlined the process of prompting that we undertook to align the workshop activity to the overall objectives of the workshop, and found that the final product was well-received. By leveraging AI to support collaboration and creativity, we took a significant step toward designing more inclusive and language-specific educational practices which prompts meaningful engagement with the world’s linguistic and cultural diversity.
Others who are looking to move beyond English as a common language in professional development settings for multilingual participants may find a model in our approach. As exemplified by the feedback we received from participants, they are keenly interested in the different ways that generative AI/ ChatGPT can assist them with developing lesson plans and course materials, an area ripe for future exploration.
References
Lee, N. T., & Ta, R. (2023, October 24). How language gaps constrain generative AI development. Brookings. https://www.brookings.edu/articles/how-language-gaps-constrain-generative-ai-development/
Mollick, E. (2024). Co-Intelligence: Living and working with AI. Random House.
OpenAI. (2024). ChatGPT (4o version) [Large language model]. https://chat.openai.com
February 4, 2025 at 09:38PM
https://ift.tt/wLS1XKO