Self-Complexity Research Network Get Involved
Methods infrastructure

Identity Strength Index

Specification (v1.0)

A six-dimensional profile of identity system strength — a companion construct to the Self-Complexity Measurement Specification. Living specification — cite the version you used.

Download Markdown Download PDF (coming soon) Cite this guide

How this relates to the Self-Complexity Specification

The Self-Complexity Measurement Specification (v2.2) asks how an identity system is structured — the number of self-aspects, their trait overlap, their geometry in trait space, and their network topology. The Identity Strength Index asks how strong or healthy that structure is, along six dimensions: Differentiation, Coherence, Balance, Integration, Commitment, and Positivity. ISI is reported as a profile, not a scalar. The dashboard exposes an _overallMean field for convenience, but the underscore prefix signals that this average is diagnostic only and should not be used as an outcome variable in inferential models.

Overview

Everythingist Research Dashboard — May 2026.

The Identity Strength Index (ISI) is a higher-order composite construct that aggregates Self-Complexity (SC) primitives with affective and conviction-based ratings into six dimensions characterizing the overall strength of a participant's identity system. ISI is distinct from Self-Complexity:

  • SC asks: how is the identity system structured? (number, overlap, geometry, network)
  • ISI asks: how strong/healthy is that structure? (along six dimensions, each calibrated against either a theoretical target or a normalization ceiling)

ISI is designed to be reported as a profile — a 6-element vector — not as a scalar. The dashboard does expose an _overallMean for convenience, but that field is prefixed with an underscore by convention to signal "diagnostic only, do not use as an outcome." Each ISI dimension reads on a 0–1 scale where 1 = stronger/healthier identity along that dimension.

ISI is computed from the same journey.computedMetrics[period] block that holds SC; it is stored under journey.computedMetrics[period].isi as a flat object of six numeric (or null) fields.

Novel composite construct developed by Sean P. Mullen, PhD, University of Illinois Urbana-Champaign.

Data Structure and Conventions

ISI is computed from a fully-populated metrics object as returned by MetricsEngine.computePeriod(). Required inputs:

  • aspectCount ($K$)
  • attributeCount ($n$)
  • avgImportance, avgCertainty — mean of the 1–11 rating scale across all aspects + attributes
  • positivityRatio — proportion of items with valence = positive
  • scMetrics.scottH, scMetrics.overlapOL, scMetrics.meanSeparation, scMetrics.withinSpread, scMetrics.balanceIndex, scMetrics.identityEntropyNorm, scMetrics.propShared, scMetrics.modularity, scMetrics.spilloverRisk

Global Constants

ConstantValuePurpose
BBAR_CEIL1.0Ceiling for meanSeparation normalization. Values above the ceiling clip to 1.
W_CEIL1.0Ceiling for withinSpread normalization.
PS_OPTIMAL0.25Target value of propShared for "appropriate connectivity" (neither over-isolated nor over-connected).
PS_WIDTH0.20Gaussian half-width around PS_OPTIMAL.
Q_OPTIMAL0.20Target value of modularity $Q$ for appropriate community structure.
Q_WIDTH0.20Gaussian half-width around Q_OPTIMAL.
SRI_CEIL5.0Ceiling for spilloverRisk normalization.
$\varepsilon$$10^{-9}$Stabilizer (unused in current dimensions; reserved).

Null Propagation

ISI returns null from compute() when aspectCount === 0. Each individual dimension may also return null when its inputs are insufficient (documented per-dimension below). The _validDimensions field reports the count of non-null dimensions; reporting guidelines below recommend interpreting profiles only when _validDimensions ≥ 4.

I. ISI Dimensions

Six composite dimensions, each on a 0–1 scale where 1 = stronger/healthier identity along that dimension. Novel composite construct developed by Sean P. Mullen, PhD, University of Illinois Urbana-Champaign.

1. Differentiation (DIF)

How distinct aspects are from one another, combining cognitive sorting independence (Scott $H$), trait-set overlap (OL), and spatial separation between aspects.

$$\mathrm{Scott\,H}_{norm} = \min\!\left(\frac{H}{\log_2(n)},\, 1\right)$$
$$\mathrm{OL}_{inv} = 1 - OL$$
$$\bar{B}_{norm} = \min\!\left(\frac{\bar{B}}{BBAR\_CEIL},\, 1\right) \quad \text{(included only when } \bar{B} \text{ is finite)}$$
$$DIF = \mathrm{mean}\!\left(\mathrm{Scott\,H}_{norm},\; \mathrm{OL}_{inv},\; \bar{B}_{norm}\right)$$

Inputs:

  • scottH
  • attributeCount ($n$)
  • overlapOL
  • meanSeparation (optional; included only when finite)

Dashboard key: isi.differentiation

Attribution: Novel — Mullen. Builds on Scott (1969) and Rafaeli-Mor et al. (1999).

Interpretation

  • High → aspects sort traits differently (high Scott $H$), share few traits (low OL), and sit far apart in trait-space (high mean separation).
  • Low → aspects are redundant, overlapping, and clustered.

Null cases

  • attributeCount === 0 → DIF = null.

Mathematical notes

  • Scott $H$ is normalized against its theoretical maximum $\log_2(n)$. When $n$ is small ($n \approx K$) the ceiling is tight; when $n$ is large, Scott $H$ rarely approaches it. The normalization is conservative.
  • meanSeparation is conditionally included (omitted from the mean when null, e.g., when attribute coordinates aren't yet populated). This makes DIF computable for journeys without the spatial substrate, at the cost of comparability across journeys with vs. without spatial data. Report _validDimensions and the per-dimension validity flags when this matters.
In plain language

Are your roles genuinely distinct, or do they collapse into one big undifferentiated self? Your professor-self and athlete-self should feel like different rooms in the same house — built from different traits, sitting in different places. Differentiation is three different windows onto that same question — sorting patterns, trait overlap, spatial separation — averaged into one score.

2. Coherence (COH)

Internal consistency within each aspect — how tightly the trait set of a single role hangs together. Reciprocal-shaped from within-aspect spread ($W$).

$$W_{norm} = \min\!\left(\frac{W}{W\_CEIL},\, 1\right)$$
$$COH = 1 - W_{norm}$$

Inputs:

  • scMetrics.withinSpread ($W$)

Dashboard key: isi.coherence

Attribution: Novel — Mullen. Derived from the within-aspect spread of Schleicher & McConnell (2005)'s associated-systems formulation.

Interpretation

  • High → tight, internally consistent aspects ($W$ is small).
  • Low → diffuse, internally inconsistent aspects.

Null cases

  • scMetrics.withinSpread === null → COH = null. This occurs when all aspects are singletons (see SC Spec v2.2 §Singleton Aspect Handling).

Limitations

  • Requires attribute coordinates (any embedding) to compute $W$. If the dashboard hasn't yet populated coordinates for this journey, $W$ = null and COH inherits null.
  • The ceiling W_CEIL = 1.0 assumes attribute coordinates are roughly in the unit cube; for embeddings on a wider scale, COH may saturate.
In plain language

When you describe your athlete-self, do the traits hang together coherently — "competitive," "driven," "energetic" — or do they sprawl across the trait space? High coherence means each role feels internally consistent. Low coherence means the traits inside a single role don't really belong with each other.

3. Balance (BAL)

Even distribution of attributes across aspects. Average of two distributional measures — Shannon-normalized entropy and the explicit balance index — which capture the same intuition with different sensitivities.

$$BAL = \frac{H_{norm} + Balance_{idx}}{2}$$

Where $H_{norm}$ is scMetrics.identityEntropyNorm (SC Spec v2.2 §21) and $Balance_{idx}$ is scMetrics.balanceIndex (SC Spec v2.2 §22).

Inputs:

  • scMetrics.identityEntropyNorm
  • scMetrics.balanceIndex

Dashboard key: isi.balance

Attribution: Novel — Mullen. Combines Shannon (1948)-derived normalized entropy with the SC v2.2 balance index.

Interpretation

  • High → roles are evenly weighted (no single role dominates the attribute portfolio).
  • Low → one or two roles consume most of the attribute budget.

Null cases

  • Either input null → BAL = null.

Mathematical notes

  • The two inputs use different functional forms (entropy is logarithmic; balanceIndex uses standard-deviation-of-proportions). Averaging them smooths over either's edge cases.
  • Both inputs are already normalized to $[0, 1]$, so BAL is in $[0, 1]$ without further scaling.
In plain language

Are your roles evenly stocked with traits, or does your work-self hog forty descriptors while your friend-self gets two? High balance means no single role dominates the portfolio. Low balance means one identity is doing all the work and the rest are skeletal.

4. Integration (INT)

Appropriate connectivity between aspects — not too isolated, not too entangled. Three sub-components, each calibrated against a theoretical target. Unlike the other ISI dimensions where "more is better," Integration is target-shaped: a participant whose aspects are completely disconnected scores low, and so does one whose aspects bleed into each other.

$$PS_{score} = \exp\!\left(-\frac{(PropShared - 0.25)^2}{2 \cdot 0.20^2}\right)$$
$$Q_{score} = \exp\!\left(-\frac{(Q - 0.20)^2}{2 \cdot 0.20^2}\right)$$
$$SRI_{norm} = \min\!\left(\frac{SRI}{SRI\_CEIL},\, 1\right), \quad SRI_{score} = 1 - SRI_{norm}$$
$$INT = \frac{PS_{score} + Q_{score} + SRI_{score}}{3}$$

Inputs:

  • scMetrics.propShared
  • scMetrics.modularity ($Q$)
  • scMetrics.spilloverRisk ($SRI$)

Dashboard key: isi.integration

Attribution: Novel — Mullen. PS and SRI from SC v2.2 §F/§G (Mullen, novel); $Q$ following Newman (2006).

Interpretation

  • PS_score: peaks at propShared = 0.25 (about a quarter of traits cross roles), drops away from there in both directions.
  • Q_score: peaks at modularity = 0.20 (some clustering, not too rigid).
  • SRI_score: higher when spillover risk is lower (linear inverse).

Null cases

  • aspectCount < 2 → INT = null. Integration is undefined with fewer than two aspects.

Mathematical notes — target shaping

  • Both Gaussian-shaped sub-components ($PS_{score}$, $Q_{score}$) reach 1.0 at the optimal value and decay smoothly. The widths (PS_WIDTH = Q_WIDTH = 0.20) determine how tolerant the score is to departures from optimal.
  • The optimal values (PS_OPTIMAL = 0.25, Q_OPTIMAL = 0.20) are based on Mullen's preliminary work and should be treated as provisional. Empirical calibration against outcome measures (well-being, identity coherence self-reports) is a near-term research priority. Report INT with a footnote indicating the calibration is provisional.
  • The optimization-around-a-target shape is what makes INT distinctive among ISI dimensions. None of the other five dimensions has a theoretical optimum — they're "more is better."

Limitations

  • Sensitive to the PS_OPTIMAL and Q_OPTIMAL choices. A different research program might justify different targets.
  • The three sub-components are weighted equally (arithmetic mean). Future revisions could replace this with a published weighting derived from external validity studies.
In plain language

Are your roles connected enough to function as one self, but distinct enough to buffer each other? Too independent and you fragment — your roles never talk to each other. Too connected and a bad day at work bleeds into your evening at home, and into your weekend tennis game. Integration is the goldilocks zone: porous walls between rooms, not no walls and not concrete.

5. Commitment (COM)

Strength of identity endorsement. Mean of the importance and certainty ratings, normalized to $[0, 1]$ from the underlying 1–11 scale.

$$imp_{norm} = \frac{avgImportance - 1}{10}, \quad cert_{norm} = \frac{avgCertainty - 1}{10}$$
$$COM = \frac{imp_{norm} + cert_{norm}}{2}$$

Inputs:

  • avgImportance (mean 1–11)
  • avgCertainty (mean 1–11)

Dashboard key: isi.commitment

Attribution: Novel — Mullen.

Interpretation

  • High → participant rates their aspects as both important and certain (confident endorsement).
  • Low → participant treats their aspects with low importance or low certainty (weak endorsement).

Null cases

  • Either avgImportance or avgCertainty is null → COM = null.

Mathematical notes

  • Assumes a fully populated 1–11 rating scale. If the participant skipped ratings on some aspects, avgImportance and avgCertainty are computed only over rated items.
  • The two ratings are weighted equally. Could be revised to weight certainty more heavily for identity research (where rated certainty arguably better captures "endorsement" than importance does).
In plain language

Do you really mean your roles, or are you just listing things? High commitment means you rate the roles as important and feel certain about them — your athlete-self and your parent-self are things you stand behind. Low commitment can flag tentative, exploratory, or ambivalent identity reporting.

6. Positivity (POS)

Overall valence of self-view — proportion of aspects + attributes the participant rated with positive valence. Passthrough of positivityRatio.

$$POS = positivityRatio$$

Inputs:

  • positivityRatio

Dashboard key: isi.positivity

Attribution: Novel — Mullen.

Interpretation

  • High → predominantly positive self-view.
  • Low → predominantly negative or mixed self-view.

Null cases

  • positivityRatio === null → POS = null.

Mathematical notes

  • This dimension is a passthrough — POS = positivityRatio with no transformation. The reason it's included in the ISI profile (rather than just reported on its own) is that valence is theoretically inseparable from identity strength. A coherent, well-differentiated, balanced identity composed entirely of negative roles is not a "strong" identity in any clinically useful sense.
In plain language

Of everything you listed about yourself, what fraction reads as positive? A direct affect measure of how your identity reads to you — your parent-self might be tagged "patient" and "warm," or it might be tagged "tired" and "short-tempered," and the mix matters.

II. Change Pattern Detection

The dashboard's IdentityStrength.detectChangePattern(deltas) classifies first-vs-latest ISI changes into seven discrete patterns based on direction and magnitude.

PatternTrigger
crystallizingCOH↑ AND COM↑ AND DIF stable
fragmentingDIF↓ AND INT↓
strengthening≥3 dimensions increased ($\Delta > 0.05$)
weakening≥3 dimensions decreased ($\Delta < -0.05$)
rebalancingmixed up and down, with ≤2 stable
stabledefault (no other pattern matched)
insufficientfewer than 3 valid dimensions in the delta input

These categories are diagnostic, not formal hypotheses. Use them for descriptive coding in longitudinal reports and for participant-level narratives in clinical or coaching contexts.

Interpretation Summary

A compact lookup view across the six ISI dimensions, with dashboard keys and plain-language gloss.

DimensionKeyRangeInterpretation
Differentiationisi.differentiation0–1How distinct aspects are
Coherenceisi.coherence0–1How tight each aspect is internally
Balanceisi.balance0–1How evenly weighted aspects are
Integrationisi.integration0–1Goldilocks-zone connectivity (target-shaped)
Commitmentisi.commitment0–1Strength of importance + certainty endorsement
Positivityisi.positivity0–1Proportion of positive-valence items

Methodological Notes

  • ISI is a profile, not a scalar. Report all six dimensions; do not collapse to _overallMean for inferential analyses.
  • Integration is the only target-shaped dimension. Its calibration (PS_OPTIMAL = 0.25, Q_OPTIMAL = 0.20) is provisional and should be empirically refined against external validity criteria.
  • Some dimensions can return null on otherwise-valid journeys (COH when all aspects are singletons; INT when $K < 2$). Report _validDimensions alongside the profile and exclude _validDimensions < 4 from inferential models.
  • ISI is computed at every period (past, present, future) wherever the dashboard's SC suite is computed. Change pattern detection (§II) operates on first-vs-latest deltas; longitudinal multi-snapshot analyses should average per-snapshot ISI rather than computing change-pattern strings.
  • The 1–11 rating scale assumption (Commitment) is hard-coded. If the participant app changes its scale, the normalizers $(x-1)/10$ must change in lockstep.

Reporting Guidelines

  • Present ISI as a six-dimensional radar plot or table; avoid bar plots that imply scalar comparison.
  • For longitudinal reports, present a $2 \times 6$ or $3 \times 6$ table (periods × dimensions) and a change-pattern label per participant.
  • When a dimension is null, report it as such; do not impute.
  • When PS_OPTIMAL or Q_OPTIMAL is recalibrated based on new evidence, bump this Specification to v1.1 and document the recalibration source.
  • Do not treat ISI as orthogonal to SC. ISI dimensions are derived from SC quantities (DIF from Scott $H$ + OL + $\bar{B}$; COH from $W$; BAL from entropy + balance; INT from propShared + $Q$ + SRI). Reporting ISI alongside SC requires acknowledging this dependence.
  • DifferentiationScott $H_{norm}$ · $OL_{inv}$ · $\bar{B}_{norm}$
  • Coherence$1 - W_{norm}$
  • Balance$H_{norm}$ · balanceIndex
  • Integration$PS_{score}$ · $Q_{score}$ · $SRI_{score}$
  • Commitment$imp_{norm}$ · $cert_{norm}$
  • PositivitypositivityRatio

Companion Specifications

Related framework documents that extend or interface with this specification.

Self-Complexity Measurement Specification The 22-metric measurement framework from which ISI's SC inputs (Scott $H$, OL, $\bar{B}$, $W$, propShared, $Q$, SRI, entropy, balance) are drawn. Salience Specification Salience weighting and per-aspect importance modeling used alongside the core metric set. Identity Archetypes Profile-based classification scheme that interprets ISI and SC combinations as recognizable identity structures.

How to Cite

This is a living specification. Always include the version number when citing.

ISI specification

Mullen, S. P. (2026). Identity Strength Index specification (Version 1.0). Self-Complexity Research Network. https://selfcomplexityresearch.org/docs/identity-strength-index.html

@misc{mullen2026isispec,
  author       = {Mullen, Sean P.},
  title        = {Identity Strength Index Specification},
  year         = {2026},
  version      = {1.0},
  publisher    = {Self-Complexity Research Network},
  url          = {https://selfcomplexityresearch.org/docs/identity-strength-index.html}
}

In-preparation manuscript

Mullen, S. P. (in preparation). The Identity Strength Index: A multidimensional profile of self-aspect system strength. University of Illinois Urbana-Champaign.

Citation note

ISI is a novel composite construct. It builds on:

  • Scott (1969) — $H$ statistic
  • Linville (1985, 1987) — self-complexity theory
  • Rafaeli-Mor et al. (1999) — component decomposition (NSA + OL)
  • Schleicher & McConnell (2005) — distance-based AST approach
  • McConnell (2011) — multiple self-aspects framework
  • Sakaki (2004) — composite SC = $K$/OL

Novel composite construct developed by Sean P. Mullen, PhD, University of Illinois Urbana-Champaign.

References

ISI v1.0 carries forward the SC Specification v2.2 reference list; it introduces no new outside references.

  1. Blondel, V. D., Guillaume, J.-L., Lambiotte, R., & Lefebvre, E. (2008). Fast unfolding of communities in large networks. Journal of Statistical Mechanics: Theory and Experiment, 2008(10), P10008. https://doi.org/10.1088/1742-5468/2008/10/P10008
  2. Jaccard, P. (1912). The distribution of the flora in the alpine zone. New Phytologist, 11(2), 37–50. https://doi.org/10.1111/j.1469-8137.1912.tb05611.x
  3. Linville, P. W. (1985). Self-complexity and affective extremity: Don't put all of your eggs in one cognitive basket. Social Cognition, 3(1), 94–120. https://doi.org/10.1521/soco.1985.3.1.94
  4. Linville, P. W. (1987). Self-complexity as a cognitive buffer against stress-related illness and depression. Journal of Personality and Social Psychology, 52(4), 663–676. https://doi.org/10.1037/0022-3514.52.4.663
  5. McConnell, A. R. (2011). The multiple self-aspects framework: Self-concept representation and its implications. Personality and Social Psychology Review, 15(1), 3–27. https://doi.org/10.1177/1088868310371101
  6. Newman, M. E. J. (2006). Modularity and community structure in networks. Proceedings of the National Academy of Sciences, 103(23), 8577–8582. https://doi.org/10.1073/pnas.0601602103
  7. Pilarska, A., & Suchańska, A. (2014). Self-complexity and self-concept differentiation: What have we been measuring for the past 30 years? Current Psychology, 34, 723–743. https://doi.org/10.1007/s12144-014-9285-7
  8. Rafaeli-Mor, E., Gotlib, I. H., & Revelle, W. (1999). The meaning and measurement of self-complexity. Personality and Individual Differences, 27, 341–356. https://doi.org/10.1016/S0191-8869(98)00247-5
  9. Sakaki, M. (2004). Effects of self-complexity on mood-incongruent recall. Japanese Psychological Research, 46(2), 127–134. https://doi.org/10.1111/j.0021-5368.2004.00244.x
  10. Schleicher, D. J., & McConnell, A. R. (2005). The complexity of self-complexity: An associated systems theory approach. Social Cognition, 23(5), 387–416. https://doi.org/10.1521/soco.2005.23.5.387
  11. Scott, W. A. (1969). Structure of natural cognitions. Journal of Personality and Social Psychology, 12(4), 261–278. https://doi.org/10.1037/h0027734
  12. Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379–423. https://doi.org/10.1002/j.1538-7305.1948.tb01338.x
  13. Tversky, A. (1977). Features of similarity. Psychological Review, 84(4), 327–352. https://doi.org/10.1037/0033-295X.84.4.327

Methods become infrastructure when others can inspect, cite, and use them.

This specification is part of a broader open-science direction that includes tools, documentation, and version tracking.

Explore the Platform Join the Network