Contribution
A theoretical article in an Emotion Review special section on action and emotion. It does three things. It argues that the usual dual-process mapping, with stimulus-driven processes automatic and goal-directed ones slow, is wrong. It carries the argument over to emotion, where appraisal theories turn out to posit stimulus-driven links. And it sets out what evidence would decide between the standard and the alternative model.
Key insight
What is usually called the emotion’s push is a choice. Once goal-directed processes are allowed to be fast, “most of the competition does not occur between stimulus-driven and goal-directed processes, but instead between different goal-directed processes” (p. 11). An emotional action tendency wins its control precedence the way any action tendency wins: its goals are highly valued and the expected utility of acting on it is highest.
Assumptions
- A content-based definition of the process types: goal-directed means represented [S:R–Ov], with at least one action option and its expectancy for one valued outcome (p. 5). Stimulus-driven means [S–R].
- Expected utility as the selector. The action option with the highest utility activates its action tendency (p. 4), and values and expectancies decide which process wins (p. 11).
- A cybernetic control architecture with nested goals, after Carver and Scheier (1990, LIT-1441). Several cycles run in parallel, and the balance among lower-order goals is kept by a higher-order goal (pp. 5–6).
- Frijda’s control precedence as the mark of emotional action (p. 15), explained by the value of the goals at stake.
Key results
Argument, with evidence reviewed, not new results:
- Three reasons goal-directed processes can be automatic (pp. 8–9): simple ones need few conditions, goals can stand in for missing conditions (subliminal priming works for goal-relevant primes), and goal-directed processes can be associative.
- The default reversed (p. 10): “we submit that goal-directed processes should win the competition instead”. Stimulus-driven control is left two niches. One is a goal specified only at a superordinate level, as in action slips. The other is operating conditions too poor for even simple goal-directed processes (pp. 10–11).
- Appraisal theory as stimulus-driven (p. 13). Appraisal patterns linked to emotion-specific action tendencies are “sophisticated types of stimulus-driven processes”. Fleeing under low control is usually optimal but not when flight is impossible, which only a goal-directed process that weighs affordances will catch (pp. 13–14).
- Emotion as decision (pp. 15–16): emotions “can themselves be regarded as decision processes, ones that rely on the very same mechanisms as non-emotional decision processes, but in which the stakes are higher”. “Episodes can be called more or less emotional” (p. 15).
- Evidence for habit is weaker than it looks (pp. 16–20). Resistance to devaluation after overtraining vanishes when there is a choice between two actions (Colwill and Rescorla 1985, pp. 17–18). Null results leave room for hidden goals, and four complexities hide them. Costly aggression, for instance, may serve a higher-valued outcome such as repaired self-esteem (p. 18).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Goal-directed processes can operate under poor conditions (little time or attention) | moderate | reviewed evidence (Bechara et al. 1997; subliminal goal-relevant priming) |
| C2 | Most behaviour-determining competition is among goal-directed processes, decided by outcome values and expectancies | weak–moderate | argued from C1 and from optimality; not tested here |
| C3 | Appraisal-pattern-to-action-tendency links are stimulus-driven | moderate | conceptual analysis of appraisal theories |
| C4 | Emotional action tendencies can be caused by goal-directed processes | weak–moderate | a handful of studies (Eder and Hommel 2013; Parkinson 2001); one-step causation not yet shown |
| C5 | Evidence for stimulus-driven control rests on null results open to hidden goals | strong (as a methodological point) | the logic of revaluation tests |
Concepts
- goal-directed process: a process whose representations hold action options with expectancies for valued outcomes, [S:R–Ov]; it activates the action tendency with the highest utility.
- stimulus-driven process: activation of a stored stimulus–action link, [S–R], whether innate, overtrained or associatively formed.
- parallel-competitive architecture: both process types run at once and compete for behaviour (Daw, Niv and Dayan 2005, LIT-580).
- control precedence: the priority an action tendency takes over others (after Frijda), read here as a product of the value of the goals at stake.
Connections
Builds on Heyes and Dickinson on goal-directed action, Carver and Scheier’s cybernetic cycle (LIT-1441, read through LIT-1449), Frijda’s control precedence (THEORY-055), and the arbitration models of Daw, Niv and Dayan (LIT-580, THEORY-062), whose default it reverses. It is set against appraisal theories (Roseman, Scherer) and affect-program theories. Its book-length development is Moors (2022, LIT-1433), unread, whose Chapter 7 is “dedicated to response evaluation theories, in particular my own version of them, the goal-directed theory”.
Bearing on the record
On the novelty of the negotiation reading (CLAIM-440, part 1). Part of the proposal is not new, and the claim should say so as it already does for Lewis and Scherer. Moors holds that what decides action, emotional action included, is a competition among goal-directed processes. She also holds that control precedence is not a property the emotion brings but the outcome of that competition when highly valued goals are at stake (p. 15). This is the record’s sentence “control precedence is not a property an emotion brings with it but the outcome of their negotiation”, stated in 2017 for an individual. Its source is Frijda, as the record’s is. The proposal cannot claim this reframing as its own.
Where Moors stops, and the proposal goes on. There are four points, and the first two carry the weight.
- She presupposes the currency the proposal makes a question. Her competition is settled by “the values and expectancies of outcomes” (p. 11), and the winner is “the action option with the highest utility” (p. 4). That is a negotiation that always yields an ordering, and a numerical one. Part 2 of CLAIM-440 holds that a negotiation need not yield an ordering (Sen’s theorem), and that an ordering need not be a currency (CLAIM-436). In the terms of CLAIM-435, Moors describes only the compensatory regime of potential, and describes it as the only one there is. The override cases she discusses (aggression costing retaliation and repairing self-esteem, the martyr’s suffering, p. 18) she reads as a hidden outcome with higher value, recovering the trade-off. A lexicographic reading of the same cases is open, and the record has a reason to prefer it in some of them.
- Her competitors are options, not perspectives. What competes is goal-directed processes or action options within one agent, ranked on one utility. The proposal’s competitors are perspectives, each with its own ranking and its own authority (CLAIM-283’s distinction between the precedence a motive claims and the authority it is granted). Moors has no room for a perspective whose ranking is not commensurable with another’s. Her balancing of multiple goals is done by “a higher-order goal” (pp. 5–6), a regulator above the competitors. That is what Frijda, Ridderinkhof and Rietveld deny (THEORY-056), and what the proposal’s negotiation lacks.
- She gives no account of valence. Emotionality on her account is a matter of degree, set by the stakes. Nothing in it marks out an ordering of the system’s own states, so it says nothing about part 3, the proposal’s own contribution. Her graded emotionality does fit consequence 3 (intensity as the strength of potential), since higher stakes give stronger control precedence.
- She would read the proposal as goal-directed theory, not appraisal theory. On her taxonomy an account on which concerns negotiate over action options, and the strongest wins, is no longer an appraisal theory. Its appraisal-to-tendency links would be goal-directed, not stimulus-driven. If the record keeps the word “appraisal” for the negotiation, it should say that it uses the word in a wider sense than appraisal theory does, and that Moors would class the view as a response evaluation theory.
What the record can take. Moors’s methodological point bears on QUESTION-268. A record of which conflicts a ranking wins is confounded by hidden outcomes. A choice that seems to follow one perspective’s ranking may follow another’s, valuing an outcome the observer did not model (pp. 18–19). Outcome revaluation and contingency degradation are her tests for which representation controls a response, and they are candidates for an interventional test of which ranking is potential. Devalue what ≿_v ranks high and see whether conduct follows.
Candidly. This is the strongest prior statement found so far of the competitive core of the proposal, stronger than Lewis’s because it is about selection among goals, not about attractors. The proposal’s novelty now rests on part 2 (no currency presupposed), on the plurality of perspectives with authority, and on part 3 (the kind of ordering that makes the regime valence). It does not rest on part 1. The book should be read before the claim says more than that.
Limitations
A position paper. The central claim, C2, is argued from optimality (“The system should prioritize processes that are most likely to lead to optimal behavior”, p. 10) and from the weakness of rival evidence, not tested. The authors grant that future research must show goal-directed processes setting the initial action tendency, not only intervening later (p. 20). Expected utility is assumed throughout and never questioned as a representation of what the competing goals rank.
Open questions
Whether “values and expectancies” settle conflicts between goals that cannot be traded off, and what Moors’s account does when the highest utility is not defined. How the higher-order balancing goal is itself selected without a regress. Whether the book’s hierarchy of goal-directed cycles (its Fig. 7.2) changes either answer.