NOTE-173: Griesemer & Shavit — Scaffolding individuality (3Cs)

Contribution

The paper offers a vocabulary, not a model. It orders coordination, cooperation and collaboration as an idealized sequence of stages in an evolutionary transition in individuality (ETI). The stages are paired with three grades of group structure: collection, conexus and community. The authors read the MLS1/MLS2 contrast as a developmental progression, “from potentially unstable MLS1 situations into more stable MLS2 situations” (§2), rather than as rival formalisms. They extend Griesemer’s sampling account of inheritance (propagules as generative samples of parental organization) from organisms to migrant social groups (§3). What is new is the combination: a staged interaction typology in which the last stage is marked by generative entrenchment and organizational closure. That typology is joined to a reproducer-based account of how such a community could be inherited.

Key insight

Group-level individuality is not a matter of how strongly members interact but of what kind of interaction dominates. In coordination and cooperation, benefits can still be booked per member in the members’ own currency. The lions’ meat per lion is the paper’s example: “the currency of cooperative interaction is the same as for individual action” (§1). In collaboration the group benefit (reliability, a team’s win) “cannot even be recognized as group benefits without reference to the special interactions engendered by the group”, and marginal attribution becomes “effectively meaningless” (§1). A community is a collective whose core interactions have become generatively entrenched, so that removing them is catastrophic. It therefore has organizational closure, and reproducing it requires exporting an organized sample of roles and interactions, not just of members (§3).

Assumptions

• The sequence is idealized and normative only in the embryologist’s sense. Coordination → cooperation → collaboration is “one possible pathway”, like a normal table, and deviations are “heterochronic”. How many real ETIs follow it is “an open empirical question” (§1).
• It fits “egalitarian” (get-together) transitions better than “fraternal” (stick-together) ones. Fraternal cases may run in other orders or in reverse (§1, after Queller & Strassmann 2009 and Szathmáry 2015).
• Diversity ≠ heterogeneity (after Shavit & Ellison 2021). Diversity is the relative abundance of member types. Heterogeneity is the variety of interactions and structural positions. Community needs the latter (§1).
• Individuality is marked by organizational closure in the sense of Moreno & Mossio (2015), which is adopted, not argued (§1, §3).
• Inheritance is propagation by sampling (Griesemer 2000). Offspring develop from generative samples of parental material parts, and an inheritance system is a sampling process evolved to make that development reliable. It need not propagate genes (§3).
• Trait groups vs demes. Only demic communities, not ephemeral trait groups, are candidates for higher-level reproducers (§2, citing Griesemer 2000 “units of evolutionary transition”). Which kind a given human grouping is, is treated as empirical.
• Human focus with a general hope. Humans are the target case “yet our hope is that the conceptual model applies to all life histories” (§1).
• No formal setting. There are no equations, no simulation and no data (“No data were reported”).

Key results

This is a conceptual paper, so the results are the positions argued or stated.

• The 3Cs (§1).

  • Coordination is “ordering in space and time” among members of a collection who co-occupy a place. The interaction is minimal, contributions are “mainly ‘aggregative’”, and the order may be self-organized, mutually imposed or externally scaffolded (queues, crosswalk stripes).
  • Cooperation is “operating together”. Members enter the arena because individual goals need a group context, but there need be no shared goal, and effects are “interactive and non-aggregative … but only in limited ways”. Benefit is still measurable per individual.
  • Collaboration is “labouring together” for a shared goal. In a community it yields a “sense of belonging” or trust, and “reliability is thus an emergent community property”.

• Table 1. The columns are Stage 1 (“Low intensity of individual interaction”), Stage 2 (“Medium–high intensity”), and Stage 3 (“Intra-group interaction”), with a vertical label, “Transition to a new level-of-individuality perspective”, between Stages 2 and 3.

  	Stage 1	Stage 2	Stage 3
  Interaction	Coordination (co-ordination)	Cooperation (co-operation)	Collaboration (co-labour)
  Entity	Collection (“Co (two)-location”)	Conexus (“Co (two)-nexus”)	Community (“Com (together)-unity”)
  Action	Ordered position	Joint operation	Shared labour
  Cause of action	Follow a rule, force or standard, externally or mutually imposed	Reconciliation within existing state, providing individual win-win situations	Maintain a transition through participation in a shared goal/aim/entity

• The structural sequence (§2). A collection is pairwise binary interactions. A conexus (“connected collection”) is a system of binary interactions forming “a web of junctions and meeting points”. A community is “intersecting reference groups” in “a rich network of heterogeneous multi-way interactions where some interaction subsets have become entrenched so that most other interactions depend on them”. Each step “further increases density and complexity of interaction structure”.

• MLS reinterpreted (§2). There is “a rough positive analogy” between individual selection, MLS1 and MLS2 and the three grades. The disanalogy is that “the 3Cs are not necessarily about levels of selection. Rather, they are about levels of individuality and reproduction.” In the paper’s variant of classical group-selection models, migration can happen within generations and groups “develop” during the selection phase, so “selection and reproduction are more or less intertwined”.

• Two instabilities and three outcomes (§2). Individuals move among reference groups, and reference groups move in trait space away from individuals. A conexus may (1) dissolve into coordination, (2) persist as an unstable conexus held by external scaffolding and high-connectivity hubs, or (3) self-stabilize into a community.

• Generative entrenchment drives conexus → community (§2). “Inter-actions” among conexuses become “intra-actions” (after Barad). The transition is “highly contingent, even if contingently irreversible”. In human cases a transition “is marked by a change in the way (human) communication is organized (and propagated)”.

• Community reproduction (§3). Following Caporael, a community “must be exported as an organized generative package larger than a single individual member”, because no migrant can carry the whole structure “in their head”. Integration means that removing node or interaction types breaks organization. The chain of inference is: community needs organizational closure, “Hence, community constitutes a new level of individuality.”

• The Neolithic thought experiment (§3). Take a society of hunters, gatherers, farmers and shamans with heterogeneous interaction. A migrant group “might require” at least one of each type, the ongoing interaction types, and new group-specific capacities (fencing farmers, temple-founding shamans) to reassemble the culture at a new site.

• Individuality and inheritance (§3). Inheritance systems at new levels “seem likely to evolve in tandem” with new levels of individuality “in so far as the latter involve diverse members with heterogeneous interactions”. The paper sets this against those who separate the two. It also suggests that Maynard Smith & Szathmáry’s limited vs unlimited heredity “might be further characterized” by the mechanisms of the 3Cs. No such characterization is given.

Claims

id	claim	strength	support
C1	ETIs can be described non-circularly, via interactions among community members that are not instances of interactions within members	assertion	Abstract only; not returned to or argued in §1–3; the among/within criterion is never applied
C2	There are explanatory benefits to treating individuality as a special case of collectivity	assertion	Abstract (“We argue for …”); the body gives a graded scale of collectivity (§2) and “community constitutes a new level of individuality” (§3), but no argument for the benefit and no comparison with rival framings
C3	Coordination, cooperation and collaboration are distinguished by aggregativity, the presence of a shared goal, and whether benefit is computable per member	informal argument	Definitions with examples (queue, crosswalk, firm, lion hunt, sports team) (§1); endnote 1 on incompatible biological definitions of cooperation
C4	The transition to a new level of individuality lies between conexus and community	informal argument	Table 1’s placement; the organizational-closure argument (§3)
C5	Generative entrenchment of core interactions drives conexus → community, and the result is contingently irreversible	informal argument	Imports Wimsatt’s (2007) entrenchment concept (§2); no case or model
C6	MLS1 → MLS2 is better read as a developmental progression than as a static contrast	informal argument	Analogy plus a sketch of models with within-generation migration and developing groups (§2); no model is given
C7	Bourrat’s scaffolding models neglect the Wrightian shifting-balance tradition and miss community “reliability”	informal argument	Two paragraphs on Bourrat (in press) and Black et al. (2020) (§2); the alternative model is deferred to “future work”
C8	Reproducing a community requires exporting an organized sample of roles and interactions, not just members	informal argument	Derived from Caporael’s concept of a social group and Griesemer’s sampling account of inheritance; illustrated by a conditional Neolithic thought experiment (§3)
C9	New levels of individuality and new inheritance systems “go hand in hand”	informal argument, hedged	§3 (“seem likely … in so far as”); in tension with §2’s allowance that individuality may emerge “not yet” with a new inheritance system
C10	The 3Cs may refine the limited/unlimited heredity distinction	assertion	One closing sentence (§3)
C11	The sequence applies “across a wide range of cases and scenarios” / “to all life histories”	assertion (stated as a hope)	Abstract and §1; the paper discusses only human cases

Method

There is no formal method. It is conceptual analysis by definition, analogy (population-genetic trait-group and deme models, embryological normal tables) and illustration (everyday examples, a Neolithic thought experiment). The main conceptual imports are generative entrenchment (Wimsatt), organizational closure (Moreno & Mossio), inheritance as sampling of propagules (Griesemer), truly social cognition and repeated assembly (Caporael), diversity vs heterogeneity (Shavit & Ellison), and scaffolding (Caporael, Griesemer & Wimsatt 2013).

Concepts

• Collection — a grouping of individuals with the potential for more interaction but interacting “in minimal ways”, through pairwise coordination. Membership lasts only while the interaction does (§1, §2).
• Conexus — “connected collection”. It is a network of binary cooperative interactions forming a web of junctions. Its persistence is an unstable dynamic equilibrium, and it corresponds roughly to MLS1 group structure (§1, §2).
• Community — intersecting reference groups in a heterogeneous multi-way network with entrenched core interactions and organizational closure. It is the new level of individuality (§2, §3).
• Coordination / cooperation / collaboration — “ordinating” in space and time / “operating together” without a required shared goal / “labouring together” for a shared goal (§1).
• Diversity vs heterogeneity — the abundance of member types vs the variety of interactions and structural positions (§1).
• Generative entrenchment — a core interaction that later interactions depend on, so that perturbing it is catastrophic while changing later ones is tolerable (§2).
• Reference group — a trait or task group an individual belongs to. Individuals can move among them, and the groups can move away from individuals (§2).
• Scaffolding — a temporary structure or interaction framework supporting a process, extended to development, evolution and cognition (endnote 2).
• Unithood — the paper’s generic word for the three grades collection/conexus/community, used alongside “phases of ‘individuality’” (§2).
• Endogenization — Bourrat’s term for bringing scaffolded features inside the collective. The paper relocates it to the “community rather than … a lower-level individual” (§2).

Connections

The paper’s nearest neighbours in this record are Bourrat’s individuality papers. It engages Bourrat’s endogenization-of-scaffolding account directly (the BJPS paper, now registered as LIT-1927) and Black, Bourrat & Rainey (2020), the model that LIT-005 names as a test case. It shares the scaffolding premise but pushes against a purely Darwinian-properties account of individuality, and that is where it differs from LIT-049 (individuals as coarse-grainings of lower-level processes) and LIT-005 (information-theoretic individuality made interventional). Its reproducer-based framing, taken from Griesemer (2000), is the ancestor of the distinctions in LIT-157 (reproducees, reproducers, Darwinian individuals). Its MLS1/MLS2 vocabulary cites Heisler & Damuth (1987), not the 1988 paper held as LIT-201, but it trades on the distinction LIT-201 draws. Its “levels of individuality and reproduction … have a complicated relationship to levels of selection” points at the units-of-selection map in LIT-160. For the broader individuality landscape see LIT-168, and for organism-centred process accounts see LIT-167. Its scaffolding/niche framing sits alongside LIT-110, LIT-169 and LIT-127, though it does not cite them.

Bearing on the record

• Mereology (LIT-124). The paper is tagged mereology, but it contains no mereology in the formal sense. It has no parthood relation, no composition principle and no discussion of when parts compose a whole. Its part–whole content is the ETI definition itself (“becoming part of a new reproducing collective while losing the capacity of independent reproduction”). Beyond that it offers a graded, interaction-based criterion for when a collective counts as one (entrenchment plus organizational closure), and the Barad-style re-description of inter-actions as intra-actions. That is a substantive answer to a special composition question for social and biological collectives, but it is given in process and organization terms, not mereological ones. A reader browsing mereology would find a case of the composition question, not a treatment of it. The tag is defensible under ANTH-ADR-046’s “justifiably appropriate” test, but it is secondary, and philosophy-of-science is right as the primary topic. Pair it with LIT-142 (Meincke), which does argue parthood and count explicitly, for contrast.
• Individuality cluster. It gives LIT-005, LIT-049 and LIT-157 a stated rival emphasis: community reliability and heterogeneity of interactions, against Darwinian properties plus individualizing mechanisms. It does not refute any of them, because no model or counterexample is given.
• ML practice. It carries nothing for ML practice. One paragraph (§1, and endnote 5 on Chalmers) speculates about a possible “AI–human transition” in which humans “find ‘community’ with AI”. The authors call this “an empirical question of great complexity” and do not pursue it. No ANTH- document bears on it.

Limitations

• The two most citable claims, non-circularity (C1) and individuality as a special case of collectivity (C2), are made only in the abstract. The abstract claims more than the body shows.
• The authors call it “not a fully-fledged model but only a first and rough sketch” (§1). They give no operational criteria that would sort an actual group into collection, conexus or community, so the empirical questions they pose (how many ETIs follow the path; trait group vs deme) cannot yet be put to data with this framework.
• Stage 3 is defined partly in psychological and intentional terms (shared goal, sense of belonging, trust, identity). That limits transfer to “all life histories” (§1) and sits awkwardly with the non-circularity claim.
• It is internally inconsistent on individuality vs inheritance timing (§2 vs §3), and there is a garbled key sentence in §2.
• The paper says it will “illustrate their explanatory power for other ETI case studies (independently studied in detail across this special issue)” (§2). It does not do so. Only the Neolithic case is sketched, and it is conditional.
• It says “hominid evolution” is central in the abstract, but the body does not discuss it.

Open questions

• Can the among-members vs within-members criterion from the abstract be stated precisely enough to check the non-circularity claim? What would count as an interaction among members that is “not an instance” of one within them?
• Is there a measurable signature that separates a conexus from a community? Candidates are non-computability of per-member benefit, or the knockout effect of removing an interaction type. With such a signature the Neolithic case could be scored.
• Does the promised model of community “reliability” and endogenization into the community (§2) actually yield ETIs that Bourrat-style models do not?
• Which ordering holds, if either: individuality before inheritance (§2), or both in tandem (§3)? What case would discriminate between them?
• How much of the framework survives when the intentional vocabulary of Stage 3 is removed, as it must be for non-human transitions?

Corrections to the seeded skim

• The dossier, the NOTE-173 summary and the LIT-188 summary all state “can be described without circularity” as a finding. In the full text, non-circularity is claimed in one abstract sentence: “We describe ET in a non-circular way, noting kinds of interactions among community members … that are not instances of those same interactions within community members”. The body never mentions circularity again. It never uses the among-members vs within-members criterion. It gives no argument that the 3Cs avoid presupposing the higher-level unit. Stage 3 is in fact defined partly by a “shared goal”, a “sense of belonging” and “group identity” (§1), which presuppose the group as their object. The claim is asserted, not argued.
• “Individuality is a special case of collectivity” is likewise abstract-only (“We argue for the explanatory benefits of treating ‘individuality’ as a special case of ‘collectivity’”). The phrase does not recur in the body, and no explanatory benefit is set out or defended there. What the body supports is a reconstruction. Collectivity is a graded scale of “density, intensity, higher-level emergent ‘jointness’ and heterogeneity” (§2). Collections and conexuses are collectives but not individuals, and a community is the collective that reaches “organizational closure” (after Moreno & Mossio), hence “community constitutes a new level of individuality” (§3). The one body sentence that tries to state the thesis is garbled in the published text: “community can be seen as a different kind not of individuality, which is not only an issue of Darwinian properties at collective level” (§2).
• The skim treats “collection, conexus, community” as three levels of individuality. The text calls them “three levels of unithood or phases of ‘individuality’” (§2). Table 1, however, draws a vertical bar, “Transition to a new level-of-individuality perspective”, between Stage 2 (conexus) and Stage 3 (community). Only the community is a new individual. The paper also slips once: the closing summary says “coordination in collectives”, not “collections” (§3).
• The skim misses the central mechanism. The conexus-to-community transition is driven by generative entrenchment (after Wimsatt 2007): “core” interactions come to be ones “later-evolved interactions will depend on for their existence and character, so that perturbations of it will tend to be catastrophic”. The transition is “highly contingent, even if contingently irreversible” (§2).
• The skim also misses the §2 apparatus. It sets out two “unstable dynamic equilibria”: individuals migrating among reference groups, and reference groups moving their trait boundaries away from individuals. It also gives three outcomes for a conexus: dissolution to coordination, persistence as an unstable scaffolded conexus, or stabilization into a community.
• The “in tandem” conclusion is weaker and less consistent than the skim says. §3 says inheritance systems “seem likely to evolve in tandem … in so far as the latter involve diverse members with heterogeneous interactions”. Earlier, §2 says that a transition to a new level of reproduction “may indicate the emergence of a new level of individuality, but not yet of a transition in inheritance system”, which “may require further adaptive evolution”. So the paper itself allows the two to come apart in time.
• The abstract says “hominid evolution and learning” drive the argument, but the body has no discussion of hominid evolution. The human material is the Neolithic southern Levant (after Shavit & Sharon), modern firms, a sports team, political parties and academic credit. “Learning” is one sentence citing Watson & Szathmáry (2016) (§3).
• The Neolithic case does not “show” that founding groups must sample heterogeneous roles. It is a conditional thought experiment: sampling “might require” migrant groups to include each occupation type (§3).
• The skim’s open question treats discipline formation (§1) as an analogy from biology to academic communities. In the text it is an example of a reverse pathway in “fraternal” cases, running collaboration → cooperation → coordination. The paper’s focus is human socio-cultural evolution throughout, and biology enters mainly through population-genetic analogies (Wilson trait groups, Wright demes).
• The skim does not record that §2 engages Bourrat directly. The paper endorses his scaffolding and endogenization account (Bourrat, BJPS; Black, Bourrat & Rainey 2020) but faults his “reliance on classical island migration models” for neglecting Wright’s shifting-balance tradition. It then proposes a “missing ingredient”: community reliability, with endogenization “into the community rather than into a lower-level individual”.
• Minor: LIT-188’s summary reads “Griesemer et al. (2023)”. There are two authors, James Griesemer and Ayelet Shavit.

Re-reading of 2026-11-01: corrections

• The among/within criterion. The correction above says the body never uses it. That is too strong. §2 uses the contrast once: the transition “may … incorporate inter-actions among disparate collective conexuses as ‘intra-actions’ within an expanded, stabilized community” (after Barad). It describes the transition there, and is never applied as a test of non-circularity. So the substance of the correction stands.
• What defines Stage 3. Only the shared goal is definitional (“otherwise it is merely cooperation”, §1). Belonging and identity are what collaboration “may contribute to” in a community. The charge that Stage 3 presupposes the group is better grounded in §3. There, sustaining heterogeneity “becomes a focal goal of every participating member”, and the paper cites Caporael’s “cognition about the group itself”.
• Three orderings, not two. §2 also says demic groups are “candidates for higher-level reproducers that may, post-transition, evolve the developmental integration necessary to count them higher-level individuals”. That puts reproduction before individuality, beside §2’s later sentence and §3’s “in tandem”.
• Missed passages.
  • §3: “Reliability can involve a trade-off between developmental resources carried within and ecological scaffolding outside the propagule”, and migrants carry “capacities to engage scaffolding”.
  • §3: migration is “a means for the products of evolution to modify the mechanisms of the evolutionary process that created them”.
  • The abstract: “Reproducing ‘systems’ tend to hybridize with environmental components”.
• Minor. C7 drops the paper’s hedge: Bourrat’s reliance on island models “may neglect” Wright. Griesemer’s “Units of evolutionary transition” is dated January 2001 by Crossref.

Re-reading of 2026-11-01: circularity and individuating an organization

Circularity. The paper never argues the answer its abstract gives. The body has three moves.

1. It starts from lower-level individuals. They coordinate “to continue to pursue whatever process trajectories they were individually following” (§2). That is not circular for the transition in question, but it leaves the regress the abstract names untouched.
2. It identifies a reproducer before an individual. Demic groups are candidate higher-level reproducers that become individuals only “post-transition” (§2, after Griesemer 2000, LIT-1932, and “The Units of Evolutionary Transition”, LIT-1937).
3. It lets scaffolds carry early coherence. Crosswalk stripes, queues and “external ecological scaffolding” do this. Endogenization later moves “scaffolding effects into the community rather than into a lower-level individual” (§2; Bourrat, LIT-1927; Black, Bourrat & Rainey 2020, LIT-1940).

Generative entrenchment (Wimsatt 2007, LIT-1926) is the mechanism. The loop by which products modify their own mechanisms is treated as a feature (§3), not dissolved.

The modes as criteria. Each grade comes with a candidate criterion.

• Collection. Membership lasts only while the interaction does (“only ‘in’ the collection … so long as they interact”), and contributions are aggregative.
• Conexus. A connected web of binary interactions, with benefit that can be booked per member (the lions).
• Community.
  • Per-member accounting of benefit becomes “effectively meaningless”.
  • Core interactions are entrenched.
  • Removing a node or an interaction type disrupts the organization.
  • Closure is asserted.

The criteria fall into two kinds.

• Stated on interactions: per-member accounting, the order of entrenchment, the knockout test and release from a scaffold. Only these can individuate an organization without presupposing it, and they do so as a declared candidate whose verdict can come out false (CLAIM-529).
• Stated on members’ attitudes toward the group: the shared goal, belonging, the focal goal and group cognition. These individuate constitutively.

Knockout and entrenchment show dependence, not closure (CLAIM-531). None of the criteria is made operational. The companion Neolithic paper (Shavit & Sharon 2023, LIT-1935) is the promised application, and is registered unread.