NOTE-553: Huebner — Genuinely collective emotions

Contribution

It is the first sustained argument that groups can be in emotional states in the strong, “collectivist” sense: states of the group that are not its members’ emotions in aggregation. Earlier defences (Gilbert, May, Tollefsen) rested on how we talk and on practices of collective apology. This one starts from a computational account of individual emotion and asks whether a distributed system could implement it. It also names exactly where such a claim fails: in a well-trained crew, danger is detected without fear.

Key insight

An emotion is not a feeling but an integration. Fear is what happens when danger detection, the redeployment of attention, preparatory bodily changes and higher-level appraisal are coordinated into one representation that says both “this is dangerous” and “do something now”. Nothing in that description mentions neurons. If a group’s routines are integrated so that the group’s state does that work, and no member’s state does, the group is afraid, whether or not it feels anything. The same description also says what to look for when a group seems emotional: agitation, a whole-system reorientation that the members’ own varied emotions do not explain.

Assumptions

• Computational theory of mind and homuncular decomposition (§2, p. 5). Psychological explanation specifies a task, decomposes the system that solves it into simpler computational systems, and repeats down to a mechanical level (Dennett 1978). Implementation is plastic: any medium that maintains the relevant counterfactual invariances will do.
• Emotions are representational states (p. 6). They have at least the function of carrying information about the world (Dretske, Millikan, Prinz). Fear carries information about danger.
• Formal and particular objects (p. 7). Emotions represent a formal object (danger) and a particular one (this raccoon), following Kenny and Prinz. Lazarus’s core-relational themes are taken as the formal objects (n. 6).
• Indicative content and imperative force (p. 8). Emotions are “action-oriented representations that reflexively reorient behavior” (Griffiths & Scarantino 2005).
• Collective representation is possible (n. 4). Defended elsewhere; here only sketched (Haugeland’s and Clark’s criteria; collective misrepresentation is possible). Huebner says sceptics may read the paper as establishing only that if there are collective representations, there can be collective emotional representations.
• Collective includes technology (n. 1). “Collective mentality” covers groups together with the technological structures they are embedded in, following distributed cognition.

Key results

All are arguments; there are no data of the author’s own.

• Holism is not collectivism (pp. 3–4). Many defences of collective emotion, such as Jaspers’s and May’s “metaphysical guilt” and Tollefsen’s collective guilt as “the guilt one feels in response to the harms committed by one’s group”, establish only that individuals’ emotions are socially relational. They do not show a group-level state.
• Gilbert’s plural-subject account is reducible (p. 4). Gilbert distinguishes personal, membership and collective guilt feelings, and defines the last by joint commitment to feel guilt “as a body”. Huebner: it is unclear how this is a collective emotion “as opposed to a distribution of individual states that are directly sustained by the causal and conceptual connections between the individuals”. Following Rupert (2005), positing such states adds no explanatory power; a plural subject’s emotion reduces to individual commitments and aggregation rules, and an individual’s emotion admits no such reduction.
• The architecture of fear (p. 10). At least (1) a capacity to perceive things as dangerous (sensory input, salience detectors, conceptual interpretation); (2) attentional mechanisms aimed at the threat and at planning; (3) systems that produce and monitor system-wide preparatory changes; (4) coordination of these outputs into a single state of fear, monitored to guide system-level coping. Huebner stresses (n. 12) that this is a requirement for representing as emotions do, not a definition of emotion.
• The USS Palau as candidate (§4, pp. 11–13). After a loss of steam pressure (Hutchins 1995), the engineer’s report is a low-level danger representation, centring the rudders plays the role of an action tendency, probability, capacity and urgency of the threat are evaluated, and attention shifts to the chart. The resulting representation is a state “of the navigation crew as it is situated in the ship”, not of any member or device. No crew member showed fear.
• Non-conscious emotion (§5, pp. 13–18). Huebner concedes the crew cannot be conscious: it is “too spatially distributed, too insufficiently integrated”, with low-bandwidth channels (p. 14). Psychoanalytic cases of unconscious fear (the Natalja example) do not help, pace Gilbert, because they may be phenomenally conscious but inaccessible (Prinz). He separates (1) a felt but unreportable emotion from (2) a non-conscious emotion in a person who feels only its scattered bodily effects (p. 15). Following Sizer on moods, he argues for (2). The feeling of fear needs the bodily changes to be integrated into a state introspection can access, and the subpersonal routines often run without recruiting introspection (Damasio, LeDoux, Tsuchiya & Adolphs). If phenomenal consciousness could instead be wholly inaccessible (Block), he says, “all empirical bets are off”, and collective phenomenal consciousness could not be ruled out either (p. 17).
• Against the homology requirement (§6, pp. 18–22). Objection, after Griffiths: conscious and non-conscious human fear share homologous mechanisms, groups do not, so a group’s danger representation is not an emotion. Reply: requiring homology (i) makes artificial emotion uninteresting by fiat, (ii) excludes aliens and convergent evolution, and (iii) forces Griffiths’s own eliminativism about emotion as one kind even within humans. Psychological kinds are fixed at the computational level. Emotions are “like vodka” (any sugar source that yields the right product), not “like bourbon” (51–79% corn). Against Rey’s demand for neurotransmitters: “neurotransmitter” is itself a functional role, which memos, emails and vocalizations could fill in a group (p. 22).
• The crew is not afraid after all (§7, pp. 23–25). Hedahl’s objection: training that removes individual fear also removes it from the distributed system. The Palau case is “collective detection of danger in the absence of a fear response”; what is missing is agitation.
• The McCain–Palin campaign (pp. 23–25). In autumn 2008 the campaign redeployed attention to its opponent, diverted resources, re-evaluated threats, and its members were agitated. Yet members’ emotions varied (hopeful, angry, saddened), so the campaign’s fear is not an aggregate of members’ fear. There is “a real pattern” missed by attending to individuals. This is offered as an “intuitively plausible case” until “the relevant anthropological data is collected”.

Claims

id	claim	strength	support
C1	Arguments from collective guilt and apology establish only holism (members’ emotions are socially relational), not collective emotion	moderate	conceptual, pp. 3–4, on May, Jaspers, Tollefsen
C2	A plural-subject collective emotion reduces to individual commitments plus aggregation rules, so it is not genuinely collective	weak–moderate	argued from Rupert’s explanatory-superfluity point, p. 4; Gilbert’s reply is not considered
C3	Representing as fear does requires four integrated capacities, and a distributed system can implement them	moderate	homuncular analysis, pp. 6–10; the Palau trace, §4
C4	Some individual emotions are non-conscious in the strong sense (no felt emotion, only felt bodily effects), so collective emotions need not be felt	weak–moderate	Sizer’s mood argument transposed, cited neuroscience of unconscious fear, §5; the key case is described, not demonstrated
C5	Emotion kinds are functional, so lacking homologous mechanisms does not exclude groups	moderate	three reductio arguments and the vodka analogy, §6
C6	Some actual groups have been in collective emotional states	weak	the campaign case, admittedly without data, §7
C7	Where a group lacks an emotion, the reason is its organisation, not that it is a group	weak	conclusion, p. 25; follows from C3 and C5

Concepts

• Holism / collectivism (after Pettit 1996) — holism: an individual’s mental states depend on social relations; collectivism: groups themselves are in mental states (p. 3).
• Genuinely collective emotion — an emotional state of a collectivity that cannot be exhaustively characterised as its members’ intentional states in aggregation (p. 4).
• Collective mentality — states of groups, or of groups plus their technological environments (n. 1).
• Emotional representation — a representation with a formal object (e.g. danger), a particular object, indicative content and imperative force, produced by integrating subpersonal routines (pp. 7–10).
• Collective detection of danger — the Palau state: integrated danger representation without the agitation of fear (p. 23).

Connections

• Gilbert (LIT-701). The rivals relation is declared on the LIT. Huebner and Gilbert agree that collective emotions need not be felt; they disagree about what bears them. Gilbert’s bearer is a plural subject constituted by joint commitment; Huebner’s is an integrated computational system that may include technology. Huebner’s claim that her account reduces to individual commitments is the core of the disagreement. Gilbert’s earlier On Social Facts (LIT-492) and “Walking Together” (LIT-477) are held unread.
• Schmid (LIT-697), unread. Schmid, by his chapter’s abstract, gives groups consciousness as plural pre-reflective self-awareness and allows them “slow” emotions only. Huebner denies group consciousness (p. 14) but finds his own fast case fails and keeps a slow one. On consciousness, Gilbert and Huebner stand together against Schmid. On bearers, all three differ.
• List & Pettit (LIT-705) and List (LIT-700). Huebner’s test for a genuinely collective state is the informal counterpart of their failure of proposition-wise supervenience. The navigation crew’s structure, routines each sensitive to one constraint and integrated on a chart, matches their distributed premise-based procedure, where no member need judge the conclusion. In List’s taxonomy, Huebner’s “holistic” emotions are aggregate or common attitudes and his “genuinely collective” ones would be corporate attitudes. Neither List paper treats emotion.
• Huebner’s Macrocognition (LIT-426), unread. The book’s published criterion, specialised subroutines integrated into goal-directed behaviour sensitive to a group’s concerns, is this paper’s criterion. A reading should check whether the book keeps the campaign case.
• Emotion theory. Frijda (LIT-546, unread; LIT-540, read) is cited among componential accounts (n. 12). Huebner’s functional emotion kinds, with appraisal-like formal objects, are what Barrett’s constructionism (LIT-542, THEORY-057) would deny is a natural kind. Huebner’s §6 accepts that cost for Griffiths’s view instead of answering Barrett, who is not cited.
• Schwitzgebel (LIT-159) and List (LIT-401). Footnote 22 states the conditional Schwitzgebel later argues for: on representational or higher-order theories, groups can be conscious given the right organisation. List’s 2016 reply quotes Huebner’s book on the absent-qualia intuition.

Bearing on the record

• THEORY-043 (composed and decomposed minds) has a direct new input. Huebner’s criterion is an architectural one, and his concession that the Palau crew is not conscious “because too spatially distributed, too insufficiently integrated” is an informal version of List’s low-Φ reasons in LIT-401. The two agree from opposite starting points.
• Candidate THEORY (proposed, not filed): collective emotions, where they exist, are corporate attitudes in List’s sense, not aggregate or common ones. What the social-science literature measures as collective emotion, convergence of members’ affect (von Scheve & Ismer, LIT-696), is an aggregate or common attitude, and so a different thing. Sources: LIT-700, this paper, LIT-696. It would be promoted when a read text applies the aggregate/common/corporate split to emotion and finds a corporate case with data.
• No instruction for machine-learning practice; nothing for the anthology.

Limitations

• No instance is established. The only worked case is withdrawn, and its replacement is described from news-level knowledge of a campaign, by the author’s own admission.
• Collective representation is assumed from other work (n. 4). The paper is, on its own statement, conditional on it.
• The non-conscious-emotion argument rests on describing a case (a person afraid without feeling fear), not on evidence that such states exist with no felt emotion at all. The cited neuroscience shows unconscious processing of fear stimuli, which is weaker.
• Gilbert’s reply to the reduction charge is not considered: a joint commitment is meant to be a single commitment of the group, not a sum of individual ones. Whether her account reduces is what is in dispute.
• “Agitation”, the property the Palau case lacks and the campaign has, is never defined in computational terms, though the whole argument is that emotions should be.

Open questions

• What, computationally, is agitation, and how would one detect it in a group without relying on members’ felt emotions?
• Can a group be in a fast emotion at all, or only slow ones, as Schmid holds? The Palau failure suggests the answer depends on training, not on speed.
• Does the distributed premise-based procedure of List and Pettit give a formal test for Huebner’s “not individual states in aggregation”? A group emotion would then be one whose occurrence is not a function of members’ emotions about the same object.

Corrections

None. The bibliographic details given in the request (2011, European Journal for Philosophy of Science) are correct for the issue; the paper first appeared online on 30 November 2010.