Designing AI Writing Workflow UX for Reduced Cognitive Loads

Abstract

This paper explores how Large-Language Model Artificial Intelligences (LLM-AIs) can be used to support people with Attention Deficit Hyperactivity Disorder (ADHD), Autism Spectrum Disorder (ASD), and other learning differences which effect cognition and self-regulation. It examines the cognitive load associated with complex writing tasks and how it affects users who have trouble with high-order thinking and planning. OpenAI’s GPT-3 API is used to analyze how AI can help with complex language-based tasks. The paper first reflects on how GPT-3 uses natural language processing to generate text, translate, summarize, answer questions, and caption images, as well as how it adapts to respond to different situations and tasks to accurately classify them. Bloom’s Taxonomy and SOLO Taxonomy are highlighted as language-driven methods of assessing learner understanding and to design activities and assessments that encourage high-order thinking. Literature is reviewed which suggests that students with disorders which effect executive functions need extra help with their writing skills to do well in school, and that early and accessible interventions such as digital self-management tools already help these learners. A model of executive-cognitive capacity is proposed to assess how best to manage the cognition of tasks and workloads, and to support a design matrix for assistive tools and processes. Finally, the Social Cognitive Theory (SCT) model for writing is evaluated for use as a procedural high-order writing process by which the tools can be designed and against which their efficacy can be validated. This review illustrates a universal design method for the development and evaluation of future AI writing tools for all users, with specific consideration towards users with atypical cognitive and sensory processing needs.