Deconstructing Human-AI Collaboration: Agency, Interaction, and Adaptation

TL;DR

This work addresses the lack of a unified framework for describing human-AI collaboration by introducing a design space built around three core dimensions: agency, interaction, and adaptation. It combines a systematic literature review with an interview study of nine researchers to iteratively refine 17 dimensions organized into five categories, and validates the space through three case studies (Co-Adaptive Topic Model Refinement, Podium, ProtoSteer). The resulting framework provides a concrete, extensible tool for describing, comparing, and guiding the design of mixed-initiative systems and their learning dynamics. Its practical impact lies in enabling researchers and practitioners to reason about where collaboration tasks are automated, how agents communicate, and how both humans and AIs adapt over time to improve analytical performance and understanding.

Abstract

As full AI-based automation remains out of reach in most real-world applications, the focus has instead shifted to leveraging the strengths of both human and AI agents, creating effective collaborative systems. The rapid advances in this area have yielded increasingly more complex systems and frameworks, while the nuance of their characterization has gotten more vague. Similarly, the existing conceptual models no longer capture the elaborate processes of these systems nor describe the entire scope of their collaboration paradigms. In this paper, we propose a new unified set of dimensions through which to analyze and describe human-AI systems. Our conceptual model is centered around three high-level aspects - agency, interaction, and adaptation - and is developed through a multi-step process. Firstly, an initial design space is proposed by surveying the literature and consolidating existing definitions and conceptual frameworks. Secondly, this model is iteratively refined and validated by conducting semi-structured interviews with nine researchers in this field. Lastly, to illustrate the applicability of our design space, we utilize it to provide a structured description of selected human-AI systems.

Figures (4)

• Figure 1: Methodology Overview: A two-phase process to formulate a design space for human-AI collaboration. The Literature Review Phase consists of assembling a corpus of relevant papers and extracting an initial set of dimensions. The Interview Phase is used to iteratively refine our design space through collaboration. The interviews were conducted in two participant groups: (a) where modifications and improvements were applied after each iteration and (b) where feedback was aggregated over the course of multiple interviews. The entire process converged at a compact final set of dimensions that make up our design space.
• Figure 2: Complete overview of the three use case systems positioned in our proposed design space. The interfaces of the three systems are depicted in \ref{['fig:usecases']}. Each of the three systems exhibits some form of human-AI collaboration; however, these systems cover different areas of the proposed design space. Note that this overview analysis does not consider temporally changing dynamics, as pointed out in \ref{['challenge:RO2']}.
• Figure 3: Overview of the visual interfaces of the three selected human-AI systems for the case studies described in \ref{['sec:usecases']}.
• Figure :