Drawing from my daily teaching practice with learners of all ages and various formal exam preparations, I extensively leverage Artificial Intelligence Personas in Virtual Reality (VR) and Virtual World (VW) environments. This hands-on experience led me to develop AI TOTALLER, a framework designed to help language educators create prompts for the AI Companions they build tailored to their class objectives.
As language educators seek innovative methods to enhance learning, the integration of VR, VWs together with AI Companions offers exciting possibilities. AI TOTALLER, developed by Helena Galani, provides a structured approach to prompt engineering, enabling teachers to design effective AI NPC interactions in virtual environments.
This post will explore the AI TOTALLER framework’s value for English language teachers and educators of other languages, demonstrating how it can enhance learning experiences through well-crafted AI prompts.
Understanding AI TOTALLER for prompt engineering
The AI TOTALLER mnemonic encapsulates essential parameters to consider when creating prompts for AI companions in immersive environments. Each component of the framework is designed to ensure that interactions are engaging, relevant, and tailored to learners’ needs:
T - Tone and Emotion Set the desired emotional tone for your AI companion's interactions. This consistency helps create an engaging atmosphere. For example, an NPC in a VR fitness world might adopt a motivational tone while guiding users through exercises.
O - Output Formatting for Optimized Multimodal Content Design prompts that align with specific language goals, utilizing various output formats—text, audio, visuals, or interactive elements. For instance, a virtual chef NPC could provide step-by-step cooking instructions while displaying visual aids.
T - Targeted Design Align prompts with the specific activities and objectives of the VR/VW experience. This clarity ensures that the NPC’s role is relevant to learners' actions. For example, an AI language tutor might teach new vocabulary based on the user's proficiency level.
A - API for Adaptive Strategies Utilise APIs to enhance NPC responses with real-time data, allowing for dynamic interactions. An NPC could adjust its explanations based on the learner’s progress or retrieve current information relevant to a discussion topic.
L - Level-Based Differentiation Adapt NPC behavior according to individual learner levels and preferences. This internal differentiation allows learners of varying proficiencies to engage with material suited to their unique needs—beginners might focus on simple sentences while advanced learners tackle complex dialogues.
L - Logical Structure and Learner-Centered Choices Provide clear guidance through logical steps in interactions. This structure helps learners navigate tasks effectively. For instance, a virtual tour guide could lead users through a museum while explaining exhibits in order.
E - Experiential Engagement Focus Encourage user interaction by creating prompts that elicit questions or offer choices. An NPC in a VR city might ask tourists if they want to explore a new area or take a quiz after a tour.
R - Role Specification for Reflective Opportunities Clearly define the NPC’s role to facilitate understanding of interaction styles and purposes. Incorporate opportunities for self-evaluation and reflection, such as an NPC historian sharing interesting facts about historical events while inviting learners to discuss their thoughts.
AI TOTALLER Specifics for ELT
In addition to the general framework, here are specific parameters to consider when designing your AI NPC prompts for English Language Teaching (ELT):
- Task: Define the type of task and activity learners will engage in, such as quests or role-plays.
- Objective: Clarify the goal of the responses and what lesson objectives are being targeted.
- Target: Identify who the NPC communicates with and what information is crucial to clarify.
- Actor: Specify who the NPC acts as and what type of behavior or character they embody.
- Learner Level: Consider existing knowledge and proficiency levels of learners.
- Lesson Aims: Outline expected outcomes from each lesson.
- Expectations of Learner Pragmatic Knowledge: Anticipate how learners are expected to react, potential mistakes they might make, and methods for correction.
- Register and Style: Determine the politeness level and formality of the NPC's language.
The Importance of Feedback Mechanisms
Incorporating feedback mechanisms within the AI TOTALLER framework is crucial for effective learning. By providing real-time corrections, personalised suggestions, and positive reinforcement, AI companions can significantly enhance learners' language acquisition processes.
While AI Companion API keys are not feedback mechanisms themselves, they play a crucial role in enabling effective feedback mechanisms by allowing AI companions to access real-time data and personalize interactions. This capability enhances the overall learning experience by providing timely and relevant feedback to learners.
Feedback mechanisms are enhanced in the following ways: Real-Time Feedback:
By using APIs to access current data (e.g., language usage examples or pronunciation guides), AI companions can provide immediate and contextually relevant feedback during interactions.
Personalized Learning:
APIs can help AI companions track learner progress and preferences over time. This information allows for more personalized feedback that addresses specific strengths and weaknesses.
Enhanced Interaction:
APIs can enable AI companions to pull in multimedia resources (like videos or articles) that provide additional context or examples, enriching the feedback provided to learners.
Data-Driven Insights:
Using APIs to collect and analyse learner data allows educators to gain insights into student performance patterns. This information can inform the feedback given to learners and help tailor instruction accordingly.
Insight
To ensure my prompts are best designed for my ELT learners, I implement features of the AI TOTALLER Framework.
The AI TOTALLER Framework was first introduced at VWMOOC 2024 on August 14th during my session titled "AI-Powered Teaching Quest with NPCs in Second Life." You can watch it here.
As Lead Moderator, I presented practical applications of utilising AI TOTALLER and for engineering your class prompts in VR/VWs such as FrameVR, Kitely/OpenSim and Second Life, at the Electronic Village Online 2025, a TESOL International CALL IS Session, "Immersive AI Companion Design" in January - February 2025.
It illustrates the content of the AI Companions I showcased at the Immersive Quest around EduNation III, Second Life (presented at the VWBPE Conference 2024) with three main characters the Dragon, Lexis and Captain, and the AI Agents built for my learners in FrameVR, ELT Treasure Island/Kitely and WondaVR.
The AI TOTALLER framework offers language teachers a robust foundation for designing engaging and effective prompts for AI companions in VR and VWs. By focusing on tone, output formatting, targeted design, adaptive strategies, differentiation, logical structure, experiential engagement, and role specification—along with specific parameters tailored for ELT—educators can create immersive learning experiences that cater to diverse learner needs.
As we embrace these innovative teaching tools, let us harness the power of AI TOTALLER to transform language education worldwide. For more insights into implementing this framework in your teaching practice, stay tuned for further resources and examples!
For additional information on the AI TOTALLER framework and its applications in TESOL, visit Helena Galani's resource page.
For any references or citations regarding the AI TOTALLER framework and its applications in language teaching, please use this site https://www.agogepaedeia.com/ai-totaller-framework and quote me as:
Helena Galani, Developer of the AI TOTALLER Framework for prompt Engineering in Language Teaching
©Helena Galani - All rights reserved
Edited January 2025
