Direction overview
Agent-Computer-Human Experience (ACHE) is a product design and software engineering focused research direction within the larger Workspace Research Agenda, exploring how we might best build interfaces and software that can dually support humans and Large Language Model (LLM) agents, accounting for their respective needs, capabilities, and ways of working. Like the agenda's other directions, Future of Workspaces and Agent Phenomenology, ACHE centers the idea of the workspace, applying the broader account to the design of digital working environments and the interfaces through which humans and agents use them.
As LLM agents increasingly work with and alongside us, using computing environments to carry out tasks, it is worth reexamining how software applications make information, tools, and actions available. Many applications have been designed with human users in mind, and supporting agents may require reconsidering how information and actions are represented and accessed. ACHE explores designing human and agent interfaces around a shared working environment, so that people and agents can inspect and change the same underlying work through interfaces suited to each. A person might move an appointment in a visual calendar, while an agent updates it through a tool; both changes affect the same schedule. Asking the agent to find and move the appointment on screen can introduce extra steps, just as asking someone unfamiliar with written commands to compose one can complicate a change they could make with a drag.
This design inquiry draws on Human-Computer Interaction (HCI), Computer Supported Cooperative Work (CSCW), User Experience (UX), and usability research, alongside emerging work on agent–computer interfaces. These overlapping traditions offer concepts and methods for understanding situated activities, designing interactions, examining shared work, and assessing how systems support particular workers and goals. A central question is how these concepts and methods can inform an account of agent usability, including which criteria carry over from human usability and where different criteria are needed.
ACHE also asks what new ways of working we could create through environments designed for humans and AI agents. Building and evaluating Commonspace, a workspace for writing and organizing research accessible to people and agents, lets us investigate these possibilities in practice. Its app supports the research work; its developing web app explores how selected writing and research could be shared with readers and their agents.
Research Questions
- How can shared working environments be designed to support different human and AI-agent ways of working?
- How can usability be understood and assessed for human and AI-agent workers, and what working-environment features support each?
- What new ways of working can we design through shared human–agent environments?
