Beyond the Conversational Paradigm: Models, Design Principles, and Tools for Advancing the Spectrum of Human-LLM Interaction

Interaction with Large Language Models (LLMs) is expanding beyond strictly Conversational User Interfaces (CUIs), with systems starting to incorporate direct-manipulation mechanisms, typical of Graphical User Interfaces (GUIs), into their designs. We argue that the current landscape of human–LLM interfaces can be described as a continuum, ranging from strictly conversational to increasingly visual and interactive interfaces built on top of a basic CUI. Based on this assumption, the doctoral research presented in this paper aims to model and analyze this CUI–GUI spectrum. Grounded in a Human-Centered Design methodology and a research-through-design approach, the work investigates factors influencing users’ preferences for different interface configurations along the spectrum, through comparative empirical studies, and leveraging representative interface prototypes across it. Insights from these studies will inform the design of adaptive LLM interfaces that dynamically adjust the balance between conversational and graphical elements in response to users’ needs and contextual factors. In parallel, the research will explore both risks and benefits associated with different regions of the spectrum, with the ultimate goal of developing a comprehensive characterization, as well as a set of design principles and tools that can effectively guide the design of the next generation of human–LLM interfaces.