Beyond Process: Building Collective Intelligence in Dynamic Virtual Enterprises (DVEs)

Executive Summary

In the modern global economy, success is often determined by how quickly a firm can assemble a Dynamic Virtual Enterprise (DVE)—a temporary alliance of independent partners pooling core competencies to seize a market opportunity. However, these temporary structures face a massive bottleneck: Knowledge Silos.

This paper proposes a socio-technological framework that moves away from rigid, process-centered systems toward a human-centered agent society. By combining Semantic Ontologies with intelligent software agents, the authors have developed a method to bridge heterogeneous knowledge types (explicit vs. tacit) and ensure that the right expertise finds the right person at the right time.

The Core Challenge: The Flaws of Process-Centricity

Most Knowledge Management Systems (KMS) are built into the "business process." While this works for static environments, DVEs are inherently volatile. The paper identifies four critical pain points:

• Heterogeneity: Different partners use different data formats and terminologies.
• Impermanence: Members join and leave, making stable infrastructure nearly impossible.
• Security vs. Openness: The risk of losing core competencies to partners-turned-competitors.
• Tacit Knowledge: Most valuable experience resides in human brains, not manuals.

Methodology: The Socio-Technological Layering

The authors propose a three-layered knowledge sharing model that mimics human social behavior:

1. The Knowledge Resource Space (KRS)

Instead of a flat folder structure, knowledge is mapped into a 3D space: KS (Owner, Category, Location).

• Semantic Links: Items aren't isolated; they are connected via links like Subclass, Part-Of, Sequence, and Reference. This allows for a "Semantic Link Network" that can infer hidden relationships through predefined reasoning rules.

2. The Agent Society Architecture

Every node in the enterprise network is represented by a software agent. These agents don't just "search"—they infer.

The agent architecture consists of:

• Semantic Explorer: A visual interface for humans to navigate the knowledge graph.
• Knowledge Recommendation: A mechanism that uses "Pulling" (on-request) and "Pushing" (proactive interest matching).
• Contribution Evaluation: A feedback loop where users score the quality of knowledge, helping the agent calculate the Agent Contribution Degree (ACD).

3. Social Discovery Mechanism

How does one agent find another in a sea of data? The authors use a specialized formula to calculate Agent Preference Similarity: $\mathrm{sim}(a_i,a_j)=\mathrm{sim}(a_i,a_j)/\min (M,N)$ This ensures that search queries are forwarded to agents with "similar taste," significantly reducing network noise.

Case Study: Cementing Collaboration

The system was validated at CNBMEEC, a company managing massive cement-plant construction projects.

• The Setup: High-stakes projects requiring collaboration between designers, equipment suppliers, and installation subcontractors.
• The Outcome: When an installer struggled with a "Plate Feeder" setup, the agent-based system successfully bypassed organizational silos to find relevant documentation and human expertise hosted by a different subcontractor, brokered through the core enterprise’s agent.

Figure: The domain ontology utilized to map the relationships between "Major," "Task," and "Mechanical Equipment."

Critical Insights & Limitations

The Collective Intelligence Evolution The most striking argument of this paper is the accumulation effect. Even when a project ends and the DVE dissolves, the "agent-to-agent" ties endure. This means the next time a DVE is formed, the agents are already smarter, having mapped out the global landscape of expertise.

Limitations: The authors admit that pure topic-based searching can still lead to "information overload" if too many items fall under one category (e.g., thousands of instances of "Hammer Crusher"). They suggest moving toward Attribute-based matching to further refine results.

Conclusion

This work serves as a blueprint for the future of decentralized collaboration. By treating knowledge as a dynamic social network rather than a static database, enterprises can achieve true agility. For managers and architects, the takeaway is clear: Stop building silos; start building networks.