Collective Goal Progress

The primary measurement of progress is the percentage of sub-tasks completed. A shared objective is first decomposed into a directed acyclic graph (DAG) of atomic actions. Progress is tracked by monitoring the state transitions of these nodes (e.g., pending, in_progress, completed, failed).

• Example: For the goal 'Generate a quarterly report,' sub-tasks include fetch_sales_data, analyze_trends, write_summary, and format_document. Progress is the ratio of completed nodes to the total.
• Key Metric: (completed_sub_tasks / total_sub_tasks) * 100.

Progress is measured as the reduction in distance between the system's current collective state and a defined target state. This is crucial for goals where completion is not a simple binary but a continuous optimization.

• Implementation: The target state is defined as a vector in a high-dimensional space (e.g., specific data conditions, environmental parameters). The system's current aggregated state is compared using a distance metric like Euclidean or cosine distance.
• Example: In a swarm of cleaning robots, the target state is 'all areas have debris level < 5%.' Progress is measured by the average reduction in debris levels across all zones over time.

Effective progress must be evaluated against time constraints. This characteristic measures the rate of advancement relative to a temporal budget or deadline for the overall goal.

• Key Metrics: Planned vs. Actual Completion Time, Burn-down Rate of remaining work.
• Observability Signal: A lagging progress rate triggers alerts for potential coordination overhead or bottlenecks. It answers the question: "At the current velocity, will the collective achieve the goal within the required timeframe?"

Progress is not merely about completion but about the cost of achievement. This tracks the aggregate computational and financial resources consumed by the agent collective per unit of progress made.

• Monitored Resources: Aggregate token usage (LLM calls), API call costs, CPU/GPU time, and network bandwidth.
• Metric: Progress Units / Total Cost. A declining efficiency ratio can indicate resource contention, inefficient task delegation, or agents stuck in loops. This is a core component of Agent Cost Telemetry.

Progress is gated by successful inter-agent coordination. This characteristic monitors the health of dependencies and handoffs between agents, as failures here halt forward momentum.

• Observability Focus: Task Delegation Traces, message acknowledgment rates, and the status of shared resources or locks.
• Failure Modes: Blocked progress due to deadlocks, unmet preconditions from a peer agent, or messages lost in publish-subscribe topic flows. Monitoring these interactions is essential for Collaborative Plan Execution.

True progress measurement must account for a dynamic environment where the goalposts or available paths may shift. This characteristic evaluates the system's ability to re-plan and maintain progress velocity despite changes.

• Trigger Events: New information invalidates a sub-task, a critical agent fails (Byzantine fault), or external API constraints change.
• Metric: Progress Recovery Time – the latency between a disruptive event and when the collective's progress rate returns to its previous baseline. Low recovery time indicates resilient collective intelligence.

A data structure that models and visualizes the network of communication pathways and message flows between autonomous agents in a multi-agent system. It is foundational for understanding collaboration topology and identifying bottlenecks or single points of failure in agent communication.

• Nodes represent individual agents.
• Edges represent communication channels or message types.
• Used to analyze information flow and dependency chains.

A composite data snapshot that aggregates the internal states—such as beliefs, goals, memory contents, and action intentions—of all agents within a multi-agent system at a specific point in time. This provides a holistic view of the system's operational context.

• Enables system-wide rollback and debugging.
• Serves as input for global decision-making algorithms.
• Critical for calculating Collective Goal Progress by comparing state vectors over time.

The aggregate computational cost, latency, and resource consumption incurred by agents to communicate, negotiate, and synchronize their actions, as opposed to performing the primary task work. It is a key efficiency metric.

• Includes costs of message passing, consensus protocols, and conflict resolution.
• High overhead can negate the benefits of multi-agent parallelism.
• Measured in CPU time, network bandwidth, and increased latency.

Quantitative indicators that measure the effectiveness and efficiency of agent teamwork. These metrics provide a granular view of how well agents are working together to advance shared objectives.

• Task Completion Rate: Percentage of subtasks successfully finished by the team.
• Shared Knowledge Utilization: Frequency of agents accessing and building upon common data.
• Conflict Resolution Speed: Average time to resolve disagreements or resource contention.
• Directly feeds into the higher-level Collective Goal Progress calculation.

A Service Level Objective defined for the reliability or performance of a system composed of multiple coordinating agents. Unlike single-service SLOs, these account for the emergent behavior of the collective.

• Examples: "99.9% of collaborative workflows complete within 2 seconds" or "Agent consensus achieved in < 100ms for 95% of decisions."
• Requires monitoring Collective Goal Progress as a primary health indicator.
• Informs orchestration autoscaling and failure recovery policies.

An alert or metric indicating that a fault or performance degradation in one agent is propagating through dependencies and causing failures in other agents within the multi-agent system. This is a critical risk in tightly coupled agent networks.

• Triggered by monitoring inter-agent latency spikes, error rate correlation, and Collective Goal Progress stagnation.
• Requires dependency mapping (via Agent Interaction Graphs) for effective root cause analysis.
• Mitigation involves circuit breakers and task reallocation protocols.