← Research

Position Paper: Towards Open Complex Human-AI Agents Collaboration Systems for Problem Solving and Knowledge Management

Ju Wu, Calvin K. L. Or

We propose a technology-agnostic, collaboration-ready stance for Human-AI Agents Collaboration Systems (HAACS) that closes long-standing gaps in prior stages (automation; flexible autonomy; agentic multi-agent collectives). Reading empirical patterns through a seven-dimension collaboration spine and human-agent contrasts, we identify missing pieces: principled budgeting of initiative, instantaneous and auditable reconfiguration, a system-wide knowledge backbone with an epistemic promotion gate, capacity-aware human interfaces; and, as a prerequisite to all of the above, unified definitions of agent and formal collaborative dynamics. We respond with (i) a boundary-centric ontology of agenthood synthesized with cybernetics; (ii) a Petri net family (colored and interpreted) that models ownership, cross-boundary interaction, concurrency, guards, and rates with collaboration transitions; and (iii) a three-level orchestration (meta, agent, execution) that governs behavior families via guard flips. On the knowledge side, we ground collaborative learning in Conversation Theory and SECI with teach-back gates and an evolving backbone; on the problem-solving side, we coordinate routine MEA-style control with practice-guided open-ended discovery. The result is the Hierarchical Exploration-Exploitation Net (HE2-Net): a policy-controlled stance that splits provisional from validated assets, promotes only after tests and peer checks, and budgets concurrent probing while keeping reuse fast and safe. We show interoperability with emerging agent protocols without ad hoc glue and sketch bio-cybernetic extensions (autopoiesis, autogenesis, evolving boundaries, synergetics, etc). Altogether, the framework keeps humans central to setting aims, justifying knowledge, and steering theory-practice dynamics, while scaling agents as reliable collaborators within audited governance.

PDF ↗

Models

Loading
EVIDENCE / COMPARISON

Benchmarks

Explore every benchmark

Loading

These charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.

Install TensorTrace