Developing Design Features to Facilitate AI-Assisted User Interactions

Interactive software tools employing generative artificial intelligence (AI) that help users formulate custom system queries are increasingly needed with growth in data quantities, relationships, and complexity. The need to afford such interactions is not new. Indeed, chatbots have long sought to bridge gaps between an individual’s intent and the system’s response. However, generative AI chatbots – in contrast to traditional chatbots that navigate pre-defined, rules-based decision trees – are unique in their promise to accept and respond to highly customized queries. At present though, most still rely upon the precise articulation of a structured prompt. The work herein develops and evaluates design features to facilitate AI-assistive user interactions in query formulation. The design features attempt to balance functional needs of users to make specific, goal-oriented, customized queries, with minimal constraints on exactly articulating pre-defined prompts. In a case study, we wireframe user interface prototypes in the domain of data log management, for evaluation with expert and novice users. Key elements of the design features revolve around the 1) refinement of search categories, 2) context-aware prompt recommendations, and 3) customization of query input per a user’s technical ability.