Cocoa: Co-Planning and Co-Execution with AI Agents
#AI agents #co‑planning #co‑execution #human‑agent collaboration #long‑running tasks #interaction design #formative study #autonomous workflows #planning vs execution
📌 Key Takeaways
- AI agents are increasingly handling complex, long‑running tasks requiring sophisticated planning and execution.
- Existing interaction designs often treat planning and execution as separate, or use human input only to correct rigid, autonomous workflows.
- The paper introduces a novel design framework called ‘Cocoa’ for co‑planning and co‑execution between humans and AI agents.
- A formative study with nine researchers was conducted to understand current collaboration practices and identify gaps.
- The authors argue for interaction designs that support continuous, iterative collaboration rather than one‑off human interventions.
📖 Full Retelling
WHO: A group of researchers (nine participants) studied how humans collaborate
WHAT: They investigated the design of new interaction models for co‑planning and co‑execution
WITH AI agents, aiming to improve collaboration between humans and long‑running AI systems.
WHERE: The study was conducted in an academic research setting (specific location not disclosed).
WHEN: Results were published on arXiv in December 2024 (arXiv:2412.10999v4).
WHY: The authors identified that existing approaches either treat planning and execution as
separate stages or rely on post‑hoc human fixes for “autonomous” workflows that are not yet fully
autonomous, and sought to develop a framework that facilitates deeper, more integrated human‑agent collaboration.
🏷️ Themes
Human‑AI collaboration, AI planning and execution, Interaction design, Long‑running AI tasks, Co‑planning frameworks
Entity Intersection Graph
No entity connections available yet for this article.
Original Source
arXiv:2412.10999v4 Announce Type: replace-cross
Abstract: As AI agents take on increasingly long-running tasks involving sophisticated planning and execution, there is a corresponding need for novel interaction designs that enable deeper human-agent collaboration. However, most prior works leverage human interaction to fix "autonomous" workflows that have yet to become fully autonomous or rigidly treat planning and execution as separate stages. Based on a formative study with 9 researchers usin
Read full article at source