Canonical definition
A collective state is a temporally situated, latent property of a collective that arises, stabilizes, or changes through member interaction and has meaning or consequences at the group level.
Claim status: The canonical definition is an original CSI proposition. Its distinction between lower-level evidence and higher-level collective constructs is grounded in established multilevel theory (Morgeson & Hofmann, 1999; Kozlowski & Klein, 2000).
Why the definition matters
When a construct depends on interaction, member attributes alone may leave out information needed to represent it. This follows the established distinction between lower-level observations and the structure of collective constructs (Morgeson & Hofmann, 1999). The appropriate composition rule depends on the target: an additive measure may be adequate for one construct, while a relational or temporal representation is needed for another (Chan, 1998).
CSI's proposed use: CSI requires a collective-level construct and a composition model linking evidence to that construct. Additive or consensus models may be appropriate for some targets; relational, dispersion, or process models may be needed for others.
Reasoning and proposed evaluation
Established foundations and proposed contribution: Collective constructs and their composition are established concerns of multilevel theory (Morgeson & Hofmann, 1999; Chan, 1998). CSI’s proposed contribution is to make these commitments explicit in computational state estimation, connecting the target construct and composition rule to temporal and contextual alignment, uncertainty, and collective-level validation. Its added value must be tested against aggregate and relational alternatives under conditions where the construct is expected to require additional structure. Additive composition remains appropriate when justified by the target.
Two collectives can have the same aggregate indicator while differing in interaction structure. For example, equal average participation can coexist with different patterns of reciprocity or influence concentration. This illustrates a distinction that aggregation can miss; it is not an empirical validation result.
CSI predicts that appropriate composition will add discriminant and incremental validity beyond an individual aggregate for relational, dispersion-based, or process-based targets. CSI also predicts more accurate and better calibrated estimates of dynamic states from temporal histories and state transitions than from static observations. These are testable propositions, not established benefits.
Defining characteristics
It belongs to the bounded collective
The construct describes the condition of the collective rather than merely summarizing the people within it.
It is not directly observable
Observable interactions and outcomes provide evidence, but the state itself is an inferred theoretical object.
It arises through member interaction
Relationships, configurations, coordination patterns, and continuing interaction contribute to its formation and change.
It exists at a specified time
A collective state may arise, stabilize, intensify, weaken, or transform as the collective and its context change.
Why the state is latent
Engagement, cohesion, conflict, and alignment are inferred conditions rather than directly recorded messages, meetings, or network ties. Observable evidence must be distinguished from the latent construct: the same pattern can support different interpretations, and an interaction metric does not automatically establish a collective state.
Calling the state latent preserves the distinction between evidence and construct. Communication frequency, sentiment, reciprocity, centralization, and performance may inform an estimate, but none is automatically identical to the collective state being studied.
Why time matters
A collective state must be indexed to an observation window. A team may exhibit strong alignment during one phase of work and fragmentation during another. A short conflict episode may be consequential without defining the collective's longer-term condition, while a persistent pattern may indicate a more stable state.
CSI therefore treats timing, duration, persistence, volatility, sequence, and trajectory as part of the construct and its measurement rather than as incidental metadata. Multilevel research identifies temporal dynamics as central to the emergence of collective phenomena (Kozlowski & Klein, 2000; Kozlowski et al., 2013).
Collective state versus group statistic
| Measure | What it represents | Relationship to a collective state |
|---|---|---|
| Mean individual sentiment | The average of individual sentiment estimates. | May be an indicator under an appropriate composition model, but it does not by itself establish collective affect or cohesion. |
| Participation count | The number or volume of contributions. | Describes activity; it does not establish engagement without evidence about distribution, quality, context, and persistence. |
| Network density | The proportion of possible relationships that are present. | May inform relational structure, but high density is not automatically cohesion, trust, or alignment. |
| Performance result | An outcome produced by the collective. | May validate or correlate with a state, but performance and collective condition are conceptually distinct. |
| Collective-state estimate | A theory-aligned representation of the collective's latent condition. | Integrates justified evidence, composition logic, time, context, uncertainty, and validation. |
Composition and emergence
Some collective states may be represented by additive or consensus-based composition. Others depend on dispersion, subgroup structure, influence concentration, reciprocity, temporal coordination, or process dynamics. The appropriate composition model follows from the construct rather than from a general assumption that aggregation is always adequate or always inadequate (Chan, 1998; Kozlowski & Klein, 2000).
Emergence means that lower-level activity contributes to a meaningful collective-level property through a specified process. It does not mean that the collective state is mysterious, independent of its members, or irreducible in every possible sense (Morgeson & Hofmann, 1999; Sawyer, 2005).
Examples and nonexamples
Potential collective states
- Engagement: sustained collective investment of attention, participation, and effort in shared activity.
- Cohesion: the strength and pattern of forces connecting participants to the collective and to one another.
- Conflict: an emergent pattern of incompatibility, opposition, or interference within the collective.
- Alignment: the degree to which interpretations, priorities, and coordinated actions support a shared direction.
Not automatically collective states
- The number of people in a group
- The average personality score of its members
- A single network metric without a construct model
- A one-time narrative summary of a conversation
- An organizational label applied without a bounded collective and observation window
The examples above are CSI construct illustrations, not claims that each listed construct has already been operationalized or validated within the CSI program.
Measurement implications
Estimating a collective state requires alignment among its definition, collective boundary, observation window, evidence, composition model, and validation strategy. CSI distinguishes the estimated state from its indicators and validation outcomes, and proposes convergent, discriminant, incremental, predictive, and calibration evidence.
References and scholarly foundations
Chan (1998) — alternative composition models across levels of analysis.
Morgeson and Hofmann (1999) — collective constructs arising through interaction.
Kozlowski and Klein (2000) and Kozlowski et al. (2013) — multilevel emergence and temporal dynamics.
These measurement and validation requirements are CSI research-program proposals.
Complete bibliographic details are available in the Scholarly Sources registry.
Relationship to CSI
The collective state is the target object of Collective-State Inference. CSI is the process; the collective state is the latent property being estimated. Collective awareness is the broader functional capacity to interpret, project, explain, and reason about calibrated collective-state estimates.
Research status
RA-002 documents the current CSI research-program position. It remains a foundational draft subject to scholarly review, empirical operationalization, and validation. The proposed capabilities and expected benefits have not been empirically established.
Preferred interim citation
Clark, E. D. (2026). Collective State. CSI Reference, RA-002, Version 0.6. Collective-State Inference Research Program.
See the Reference version history.