Qualify the collective boundary
Establish why the observed participants constitute a theoretically defensible collective, including membership, shared environment, context, and observation window.
Conceptual model
How Collective-State Inference is structured.
The CSI framework connects observable interaction evidence to a bounded, probabilistic representation of an emergent collective condition.
Core framework
From bounded evidence to responsible use.
CSI is a staged inferential process rather than a single classification step.
Specify the state
Define the collective-level construct, temporal meaning, and theoretically justified composition model.
Integrate evidence
Use participant-level, relational, temporal, and contextual signals justified by theory.
Estimate the collective state
Estimate the most plausible collective condition while representing alternatives, uncertainty, and temporal change.
Compare and validate
Test against collective-level criteria and simpler baseline models.
Support awareness
Use validated collective-state estimates to support interpretation, explanation, projection, and reasoning under accountable human governance.
Measurement architecture
From observable interaction to validated collective-state inference.
This responsive diagram expresses the methodological pipeline as structured content rather than a static image. Each stage remains readable, selectable, searchable, and available to assistive technologies.
Boundary-qualified interaction evidence
Collect behavioral, relational, temporal, and contextual traces only after the observed population has a defensible collective boundary.
- Messages
- Replies
- Reactions
- Participation
- Timing
- Network structure
Evidence extraction
Transform raw traces into measurable features while preserving their analytical level.
- Behavioral
- Relational
- Temporal
- Contextual
Theory-aligned evidence integration
Combine the evidence categories required by the construct and composition model. One source may be sufficient only when theory and validation justify it.
Collective-state estimation
Infer a probabilistic latent collective condition at the appropriate collective level.
- Engagement
- Cohesion
- Conflict
- Alignment
- Collective emotional tone
Validation framework
Evaluate whether the estimate corresponds to the intended collective construct rather than a convenient proxy.
- Construct
- Cross-level
- Discriminant
- Temporal
- Predictive
- Incremental
Confidence updating and monitoring
Revise estimates and confidence as new evidence, validation results, and changing group dynamics become available.
Core principle: CSI estimates latent, emergent conditions defined at the collective level. Whether aggregation is adequate depends on the construct, its justified composition model, and collective-level validation.
Comparative specification
How CSI distinguishes the models it asks researchers to compare.
CSI does not prescribe one algorithm. It specifies representational commitments and asks whether each added commitment improves inference for the construct being studied.
Compact representation: [Ŝ(g,t), U(g,t)] = fθ[I(g,t), R(g,t), H(g,1:t), C(g,t)]
The estimated state Ŝ and its uncertainty U are functions of participant evidence I, relational evidence R, history H, and context C for collective g at time t. The notation is implementation-neutral; it is not a claim to a new estimation algorithm.
| Comparison | Representation | Question tested |
|---|---|---|
| Aggregate baseline | Individual summaries | Is an additive or consensus model sufficient for this target? |
| Relational model | Individual summaries plus ties, reciprocity, influence, or subgroups | Does relational structure add construct-relevant value? |
| Fuller CSI model | Construct-justified individual, relational, historical, contextual, and uncertainty terms | Do the added commitments improve validity or calibration enough to warrant the complexity? |
Evidence architecture
Four evidence categories may inform an estimate.
Participant-level
Individual traces
Participation, contribution, responsiveness, language, affective expression, and task behavior.
Relational
Patterns among participants
Reciprocity, connectivity, subgroup formation, centralization, brokerage, and influence.
Temporal
Change and persistence
Volatility, acceleration, deterioration, recovery, sequence, and path dependence.
Contextual
Situational context
Task demands, culture, hierarchy, constraints, incentives, and external events.
Qualification requirements
Not every group assessment qualifies as CSI.
A qualifying CSI inference requires an explicit and theoretically defensible collective boundary, a genuinely collective-level construct, justified composition logic, and collective-level validation. A population grouped only because its members were observed in the same environment does not automatically satisfy this requirement.
Detailed requirements, boundaries, and scholarly foundations are maintained in the canonical CSI Reference article.
Research instrumentation
Inspect candidate evidence without overstating inference.
The CSI Research Observatory applies privacy-preserving aggregation and temporal analysis to website interaction observations so the program can inspect candidate evidence, evidence windows, heuristic indicators, and publication safeguards.
Because co-observed site visitors have not been established as a CSI-qualified bounded collective, the Observatory is not presented as a qualifying collective-state estimate or a demonstration that CSI inference has been achieved.
Continue exploring
Use the Framework as the visual overview—not the canonical source.
The CSI Reference contains the governed definitions and methods. The Research Program explains how the framework will be tested and extended.
Open the CSI ReferenceView the Research Program