Toward Artificial Intelligence-Supported Development of Image-Based Tasks in Aphasia

Images play a central role in language assessment and therapy for people with aphasia, yet the development of psychometrically robust image stimuli is often time consuming, costly, and difficult to tailor to specific research questions or individual patient needs. Recent advances in artificial intelligence (AI), particularly text-to-image generation models, offer new opportunities to rapidly produce high-quality, customizable visual stimuli. Despite this potential, methodological guidance for integrating AI-generated images into aphasia research and clinical practice remains limited. This tutorial aimed to provide practical recommendations for the use of AI text-to-image tools in developing image-based tasks for people with aphasia.

Using a case study framework, we demonstrate the development of a semantic judgment task with images generated with the help of the DALL-E 3 model. We outline key decision points that went into this process and provide recommendations across the major stages of task development: task conceptualization, stimulus design, image ideation, prompt engineering, and image quality evaluation by AI models.

We demonstrate that structured prompt engineering combined with systematic validation can yield high-quality, interpretable image stimuli suitable for aphasia tasks. The approach allows for flexible control over stimulus characteristics while maintaining consistency across experimental conditions.

AI text-to-image generation offers a promising, scalable approach to developing customized image stimuli for aphasia research and clinical practice. We discuss methodological considerations, limitations, and future directions for the responsible integration of generative AI technologies into aphasia-related research and clinical applications.