Fridolin Wild, Mikhail Fominykh, Andreas Kißmehl, Tormod Aagaard, Antigoni Parmaxi, Eirini Christou, Anke Berns, Lina Adinolfi, Skevi Vassiliou, and Angeliki Voreopoulou (2025) Technical recommendations for Artificial Intelligence and Augmented Reality tools to support language learning: A guide for developers, instructional designers and language teachers. ARIDLL Consortium. https://aridll.eu/

Executive summary

This guide provides technical recommendations for using Artificial Intelligence (AI) and Augmented Reality (AR) tools to support instructed language learning. It is primarily aimed at application developers, instructional designers and language teachers who utilise AI technologies and designed experiences, though parts may also benefit learners interested in enhancing their media competence.

With this guide we reflect on a series of experiments, pilot studies and expert workshops, to gather the knowledge we have built in using and teaching these technologies. We have explored the longitudinal use of AI+AR in our Erasmus+ funded project for almost three years (at the time of writing), and this guide provides a summary of our insights on both what works and what to pay attention to.

The guide is based on principles grounded in psychological research, user experience, and pedagogical good practice. It prioritises user experience, drawing on frameworks such as the Motivation, Engagement and Thriving in User Experience (METUX) model, which focuses on psychological needs such as Autonomy, Competence, and Relatedness. The CLEAR Framework (Concise, Logical, Explicit, Adaptive, Reflective) is also adopted to improve interactions with AI language models.

The guide is divided into two key parts:

Part A provides an overview of the role and benefits of AI-powered AR. AR overlays computer-generated images onto the real world, while AI enables computer systems to perform tasks requiring human intelligence, such as reasoning and problem-solving. The fusion of AI into AR applications creates a powerful and personalised educational tool, capable of analysing a learner’s proficiency and tailoring learning experiences to their needs. This can include AI-powered virtual humans providing personalised feedback within immersive, real-world scenarios for conversational practice.

Part B in turn focuses on prompt engineering. It introduces prompt patterns and a pattern language for designing AI+AR prompts, modifying traditional design pattern formats to include key elements such as context, knowledgebase, and user prompt. These are crucial for configuring AI+AR-enabled virtual humans with specific conversational traits and behaviours. This part covers good practice in writing prompts, technical considerations like Chain of Thought (CoT), Few-Shot learning (FS), Retrieval-Augmented Generation (RAG), and Memory. It also details language-related patterns and pedagogical patterns for language learning, including feedback strategies, roles/personas, and conversation management. It concludes with patterns for situated task practice in various real-life scenarios.