VL:
Talk to me about (1) definitions/characterisations of anger that do not rely on subjective reports or overt behavioural indicators, (2) atypical presentations of anger, (3) the reliability or otherwise of subjective testimony: while one may be angry without being, in the moment, consciously aware of this, can one have a mistaken positive belief about this (believing oneself angry when one is not - or vice versa)?
GPT-4o:
Let us reformulate your inquiry as follows:
To what extent can the affective state of anger be objectively individuated without recourse to subjective introspection or behavioural cues? What are the neurocognitive or physiological substrates that might support such an individuation? How might anger manifest atypically or obliquely—phenomenologically, neuropsychologically, or culturally—such that canonical markers (e.g., overt aggression, self-reported hostility) fail to track it? And finally, can the first-personal epistemology of anger be fallible in both directions—such that subjects can fail to avow their anger, or falsely avow it? What does this bidirectional opacity imply for models of affective self-knowledge?
We shall proceed in three parts.
I. Non-Subjective, Non-Behavioral Individuation of Anger
1.1 Anger Without Avowal or Behavior: Neurobiological and Affective Taxonomy
If anger is to be individuated independently of subjective reports and overt behaviour, we are forced to ground it in subpersonal or subphenomenal features—most plausibly, in neurophysiological, endocrinological, and computational characterisations of affect. This requires, first, a decoupling of anger from its folk-psychological phenomenology (e.g., a feeling of heat or rising tension) and from its canonical behavioural outputs (e.g., shouting, clenched fists), and second, a search for state-space-level signatures in affective neurobiology.
Anger correlates robustly with certain neuromodulatory cascades, especially catecholaminergic activation, notably involving norepinephrine and dopamine systems, with strong coactivation of the amygdala, periaqueductal grey (PAG), ventromedial prefrontal cortex (vmPFC), and anterior insula. However, this ensemble is not unique to anger. For instance, activation in the amygdala-PAG axis can subserve fear, rage, or pain-related aversion. Thus, the problem becomes one of affective discriminability within overlapping neural circuits.
Recent computational-affective neuroscience (see Barrett, 2017; Adolphs, 2010) has questioned the viability of discrete emotion categories entirely, arguing instead for constructivist or dimensional models, such as Russell’s Circumplex Model (valence × arousal) or Scherer’s Component Process Model, in which anger is understood as a dynamic constellation of appraisals and physiological states. Within such models, anger tends to occupy a quadrant of negative valence, high arousal, with a characteristic goal-frustration appraisal structure (cf. Ortony, Clore & Collins, 1988).
From a more reductive neurocomputational stance (Friston, 2010; Seth, 2013), anger might be construed as a precision-weighted prediction error in interoceptive inference—an unsuccessful attempt to minimise discrepancy between expected and actual bodily states, particularly in social interaction schemas involving norm violation or goal obstruction. This framing situates anger not as a sui generis emotion, but as a mode of active inference under duress, especially when agents are attempting to enforce expected regularities (e.g., fairness, status, physical integrity) in the face of unexpected resistance.
Thus, anger—stripped of introspective and behavioural trappings—may be provisionally defined as a multilevel regulatory response to social or physical frustration, encoded as increased autonomic arousal, enhanced sympathetic tone, and altered predictive modelling in affectively valenced cortical-subcortical loops. However, such definitions remain heuristic and probabilistic, not criterial.
II. Atypical Presentations of Anger
2.1 Phenomenological Atypicality
Not all anger feels like anger. The subject may experience ennui, irritation, restlessness, or even somatic dysphoria in ways that retrospectively consolidate into anger but were not accessible as such at the time. The psychodynamic tradition (cf. Winnicott, Klein) offers a rich language of defence and displacement: anger may be repressed, sublimated, or somatised, presenting as gastric tension, insomnia, or withdrawal. Here, the anger is real but affectively disguised, and subjectivity itself becomes opaque.
2.2 Neuropsychological and Clinical Dissociations
In alexithymia and certain autism spectrum conditions, subjects may exhibit physiological or behavioural indices of anger (e.g., increased heart rate, withdrawal, aggression) but lack the metacognitive lexicon to report or interpret these as such. Similarly, in frontotemporal dementia or orbitofrontal syndrome, patients may express disinhibited aggression without experiencing prototypical anger; or they may experience anger without affective congruity, due to decoupling of affect and appraisal (cf. Rolls, 1999).
In borderline personality disorder, anger may manifest as affective lability without clear intentional structure, and in melancholic depression, anger is often self-directed or turned inward (cf. Freud’s Mourning and Melancholia), generating presentations of guilt or worthlessness that lack outward aggression but are nonetheless suffused with hostile intentionality.
2.3 Cultural and Social Deviance
Cultural scripts of anger modulate both expression and recognition. In collectivist cultures (e.g., East Asia), anger may be morally disvalued and therefore more likely to be displaced or internalised; in individualist cultures, assertive anger may be valorised. Moreover, gendered socialisation frequently renders female anger as pathologised (‘hysteria’) and male anger as justified (‘righteous indignation’), further obfuscating atypical presentations.
In all such cases, anger is epistemically elusive not because it is absent, but because its expression and recognition are filtered through non-trivial socio-cognitive mediators.
III. The Epistemic Status of First-Personal Reports: Error and Opacity
3.1 Can One Be Angry Without Knowing It?
Yes—and more radically than this, the very notion of knowing one’s own anger may require unpacking. If we accept that self-knowledge is mediated by introspective mechanisms, then these mechanisms can fail. The transparency model of self-knowledge (Evans, 1982; Byrne, 2005) suggests that we determine our mental states not by inward inspection but by outward inference—we look to the world (and our behaviour in it) to infer what we believe or feel. On this model, anger is not introspectively infallible, and misidentification is possible.
Indeed, cognitive penetrability of affective appraisal means that beliefs, norms, and expectations can distort or even fabricate one’s self-ascription of anger. One may believe oneself angry because one believes one should be angry (e.g., moral outrage without visceral arousal), or deny one’s anger due to prohibitive self-conception (e.g., ‘I’m not the kind of person who gets angry’).
3.2 False Positive Beliefs About Anger
A belief that one is angry can be false even in the absence of countervailing affective physiology. Consider:
Simulation: Subjects may simulate anger (e.g., in acting or persuasion) and subsequently experience affective contamination—a kind of misattribution of their performative state.
Moral Overidentification: One may conflate judgmental condemnation with anger (especially in hypermoral agents), mistaking cognitive disapproval for visceral affect.
Confabulation: In neuropsychological cases (e.g., split-brain patients), subjects may confabulate reasons for affective states that were triggered by inaccessible stimuli.
Thus, the belief ‘I am angry’ is not incorrigible. It is defeasible, inferential, and contingent upon both internal and external evidence.
3.3 False Negative Beliefs
Conversely, denial of anger may mask physiological and neurocognitive profiles consistent with it. This is especially evident in psychosomatic medicine, where chronic, unacknowledged anger correlates with elevated cortisol, hypertension, and cardiac risk.
This implies that non-avowal is not absence. Anger may be subpersonal, tacit, or disavowed—a structure in the affective system but not in the intentional subject.
Conclusion: Toward a Non-Avowalist, Non-Behaviorist Account of Affective Self-Knowledge
Anger, when stripped of folk phenomenology and behavioural teleology, is best understood as a multicomponent neurocomputational process, involving a socially encoded appraisal of goal frustration, physiological arousal, and motivational reorientation. Its presentation may be masked, distorted, or fragmented, especially in conditions of cognitive impairment, cultural suppression, or affective defence.
First-personal knowledge of anger is neither transparent nor incorrigible. It is a constructed inference, modulated by memory, self-image, and cultural narratives. The possibility of both false positives and false negatives in affective self-ascription implies that anger is real even when it is denied, and illusory even when it is claimed.
The larger philosophical implication is that affective states do not reduce to their avowals. They are not merely subjective, but minimally objective: they can exist without being known, and be known without existing.
Would you like to extend this analysis into models of moral anger, political affect, or predictive coding frameworks for affective dynamics?
VL:
Re the epistemic status of first person reports, I couldn’t help thinking about the elephant and rider analogy. In this context, the rider would be the conscious self constructing narratives (the storyteller in the brain) and elephant would be the larger, more complicated, imperfectly apprehended self about which those narratives are constructed.
‘Oh, but I have a strange and fugitive self shut out and howling like a wolf or a coyote under the ideal windows. See his red eyes in the dark? This is the self who is coming into his own.’
GPT-4o:
What you are really asking now is something like this:
If the self is not a unified, introspectively transparent agent, but a layered and partially occluded system of processes—if, in other words, the conscious narrating ‘rider’ is merely one module among others—then how should we reconceive first-person affective avowals (such as ‘I am angry’)? What is the epistemic status of such statements if the self who utters them is not in authoritative possession of the relevant facts, but merely interprets or retrofits them in light of incomplete or inferential data about a deeper, often inaccessible ‘elephant’ self? Can we systematise this metaphor into a formal model of affective self-ascription as post hoc rationalisation or narrative construction? And what philosophical or cognitive frameworks allow us to move beyond naïve avowalism while still treating affect as real and epistemically tractable?
Let’s proceed to unpack this at a deeper level, especially through the lenses of cognitive modularity, predictive processing, affective neuroscience, and narrative identity theory, while also reflecting on the literary and psychoanalytic undercurrents in your invocation of the ‘howling self’.
I. The Rider and the Elephant: Modular Mind and Narrative Self
Your invocation of the rider-and-elephant metaphor (popularised by Jonathan Haidt, drawing on work in moral psychology and dual-process theory) places us squarely in a family of views emphasizing cognitive modularity, affective primacy, and post hoc rationalisation. The metaphor parallels older philosophical models: Plato’s charioteer (with the white and black horses), Freud’s id-ego-superego dynamic, and contemporary dual-process theories (System 1 vs System 2; Kahneman, Stanovich, Evans).
Here, the ‘rider’ represents the narrative self—the conscious, linguistic module that interprets, explains, and projects coherence onto the chaotic, distributed, affectively driven processes of the subpersonal psyche. The ‘elephant’ is not simply the emotional brain but the entire complex of non-conscious cognition, embodied affect, implicit valuation, and homeostatic regulation—including what might be described in psychoanalytic terms as the repressed or in Deleuzian terms as the virtual self.
This model implies a deep epistemic asymmetry between experience and report. The subject is not a transparent observer of its own mental states, but a confabulating interpreter whose utterances—especially those of the form ‘I feel X’—may function more like hypotheses or narrative performances than like direct access to mental facts.
II. The ‘Strange and Fugitive Self’: Affect, Opacity, and the Limits of Avowal
Your quotation—‘a strange and fugitive self shut out and howling like a wolf’—echoes, perhaps unintentionally, both Julia Kristeva’s abject, Eugene Gendlin’s felt sense, and Nietzsche’s concept of the affectively subterranean will. This is the self beneath or beyond narration—a self that is affectively alive but semiotically inchoate.
This howling self is not merely a poetic metaphor but may refer to real non-linguistic affective states: background moods, visceral urgencies, drives, and micro-affections that do not rise to the level of consciousness or elude the conceptual resources of the rider-self. One might think here of Antonio Damasio’s distinction between core affect and extended consciousness (cf. The Feeling of What Happens), or of the affective attunement literature in early developmental psychology (Stern, Trevarthen), where pre-linguistic infants exhibit richly modulated affective states long before self-concept or linguistic self-reference emerges.
This reframes affective avowals: ‘I am angry’ is not a statement of brute fact, but a conceptually saturated articulation of a higher-order representation, formed after the affective episode has been structured, filtered, and rendered narratable. In higher-order theory of consciousness (e.g., Rosenthal), such second-order representations are constitutive of conscious awareness—yet they remain vulnerable to error, omission, and distortion.
III. Predictive Processing, Active Inference, and Affective Self-Deception
To naturalise this picture, we turn to predictive processing models (Friston, Clark, Hohwy, Seth), in which the brain is construed as a hierarchical prediction machine, minimising free energy (surprisal) by continuously generating hypotheses about the causes of sensory (including interoceptive) input.
In this model, affect is understood not as input, but as the precision-weighting of prediction error—a signal about the salience or significance of violated expectations. Anger, specifically, may arise when deeply held priors (e.g., about fairness, respect, physical safety) are violated, and when action policies (e.g., flight or submission) are suppressed in favour of confrontation or norm enforcement.
Crucially, conscious experience of emotion arises when these prediction errors propagate to higher-order cortical areas that have access to conceptual and linguistic resources. But if these errors are attenuated, masked, or reinterpreted at lower levels—either via social learning, defence mechanisms, or neurochemical suppression—then the higher-order model may simply fail to ‘notice’ them.
Thus, one can be angry without knowing it not because the anger is ‘hidden’ behind a screen, but because the brain is enacting a model of the world in which anger is not predicted, not allowed, or not useful—a model that actively suppresses or modulates the ascending affective data. This is the predictive-coding analogue of repression.
Moreover, false positive anger-avowals are equally possible in this model: if a subject strongly anticipates that she should be angry (e.g., due to moral norms, group affiliation, identity narratives), she may ‘read in’ anger even in the absence of the relevant lower-level signals. These are top-down illusions—akin to phantom limb pain or Capgras delusion, but affective.
In both cases, the self is not a site of unified agency but a bundle of generative models, operating at different spatiotemporal scales, whose outputs may conflict or remain unintegrated.
IV. Toward a Philosophical Reorientation: Self-Knowledge, Confabulation, and Epistemic Modesty
This all calls into question the privileged status of first-person avowals, long central to analytic philosophy (e.g., Shoemaker’s claim that ‘self-knowledge is immune to error through misidentification’). What we are left with is a far messier picture.
On this view:
Avowals are theory-laden acts, not brute epistemic facts.
Self-knowledge is best modelled as inferential, predictive, and corrigible, not immediate or incorrigible.
The narrating self is a sense-making mechanism, not a mirror.
Much of our affective life is subpersonal, embodied, and extracognitive, though not thereby inaccessible to investigation.
This does not entail nihilism about the self, or emotion. Rather, it invites a layered epistemology in which affective states are accessible only via triangulation—drawing on physiological data, social feedback, phenomenological reflection, and conceptual inference.
V. Closing: The Howling Wolf and the Real Self
Let us return to your poetic image. The ‘howling self’ is not just a metaphor—it may refer to a non-integrated system of affective energies that resist linguistic domestication. These are not false or illusory, but simply non-narrative. The task of philosophy, neuroscience, and psychotherapy alike is to develop vocabularies, models, and rituals that allow these energies to surface, to be interpreted, and—perhaps—to be integrated into a richer, more honest, but always incomplete self.
As Nietzsche put it, ‘There is more wisdom in your body than in your deepest philosophy.’
Would you like to explore the literary or psychoanalytic extensions of this—e.g., how Dostoevsky, Woolf, or Lacan conceptualise the fragmented self and its relation to affective truth?
VL:
Although there is clearly a game-theoretic benefit to being seen as the sort of person who may with sufficient provocation be moved to anger, is there (for a skilled dissimulator) any benefit to actually being angry?
GPT-4o:
Reformulated Question:
If one accepts that the social utility of anger is largely in its external signalling function—its role in credible threat displays, norm enforcement, or deterrence—does there remain any intrinsic or instrumental value in actually undergoing the internal affective state of anger, particularly for an agent who possesses sufficient cognitive control to simulate anger behaviourally when required? In other words, for a cognitively sophisticated agent, is there any decision-theoretic or epistemic advantage to the experience of anger per se, or is its value exhausted by its performative and reputational effects?
We may begin by analytically distinguishing three overlapping conceptions of anger:
Phenomenal anger: the subjective experience, i.e., the felt quality of anger (heat, tension, urgency, resentment).
Behavioural anger: observable manifestations—shouting, facial expressions, threats, violence.
Strategic anger: the deployment (real or simulated) of anger-signals for instrumental gain—e.g., bargaining advantage, deterrence, reputation effects.
Your question presumes that strategic anger is decouplable from phenomenal anger, and asks whether, insofar as this decoupling is possible, there is any advantage to not merely simulating, but actually undergoing, the inner state.
To address this, we must consider the affective architecture of anger, the computational costs and benefits of simulation vs authenticity, and the relationship between cognition and emotion in real-time decision-making. The answer must therefore integrate evolutionary game theory, affective neuroscience, and philosophical theories of emotion as epistemic processes.
I. Anger as Commitment Device: The Game-Theoretic Utility of Reputational Irrationality
The canonical account in game theory treats anger as a pre-commitment mechanism—a form of strategic irrationality that alters the payoffs in repeated interactions. This is the Schelling-Frank-Hirshleifer model:
Schelling (1960): Credible threats are often irrational threats—what matters is that the adversary believes you will carry them out, not that it is rational to do so.
Robert Frank (1988): Emotions function as commitment devices. The costs of anger (impaired cognition, impulsivity) are outweighed by the social advantages of being known to be the kind of agent who becomes angry when betrayed.
Jack Hirshleifer (2001): Emotions solve problems of private information and time inconsistency—they bind the agent to courses of action that would be strategically optimal to undertake only if one were bound to them.
In this tradition, actually being angry enhances credibility: if you can convincingly lose control when wronged, others are more likely to respect your interests in future interactions. The costs of anger (e.g., impaired cognition, reputational risk, physical danger) are justified by their ex ante deterrence effect.
However, this account presumes that dissimulation is cognitively costly and easily detected—that faking anger is harder than experiencing it. What if it is not?
II. The Skilled Dissimulator: Simulation vs Authenticity
Suppose an agent is highly adept at simulating anger (e.g., via motor mimicry, prosodic modulation, affective facial control, and credible threat calibration). In that case, it becomes plausible to suppose that phenomenal anger is unnecessary—the behavioural outputs are what matter for social signalling, and these can be performed.
The advantages of simulation include:
Cognitive control: One remains capable of strategic adaptation mid-interaction.
Avoidance of maladaptive costs: No risk of overreaction, reputation loss, or retaliatory escalation due to misjudged provocation.
Cross-domain flexibility: One can ‘express’ anger across many situations without cumulative physiological cost.
But there are two major counterarguments here:
(1) Cognitive Resource Allocation
Simulating affect requires executive bandwidth. It draws on working memory, attention, and cognitive load management. Especially in real-time interactions—negotiation, combat, legal argument, social conflict—the overhead required to simulate anger without undergoing it may be nontrivial.
Under this model, actually being angry is a computational shortcut. The affective system reallocates priorities, suppresses irrelevant options, sharpens attentional focus on norm violations, and primes muscular and cognitive systems for confrontation. That is, anger is a local optimiser for adversarial environments.
In computational terms: affect is an emotion-driven policy prior—it imposes a directional bias on action selection that is computationally efficient in particular ecological niches. To suppress this prior in favour of top-down simulation is costly and, in high-stakes settings, potentially slower.
(2) Leakage and Detection
Paul Ekman’s work on microexpressions and nonverbal leakage suggests that simulated affect often diverges subtly from genuine affect, and that humans (especially those with high social acuity) are sensitive to such divergences. In iterated games (e.g., reputation-building over time), agents who simulate anger but are discovered to be doing so may suffer credibility collapse.
Hence, for a skilled dissimulator, the optimal strategy may be dual-layered:
Allow genuine anger to arise under carefully controlled conditions;
Suppress or redirect it when strategically disadvantageous;
Simulate it when necessary but do so with awareness of its limits.
Authenticity here serves as noise insurance—it maintains the plausibility of the signal even when the signal is faked. That is, phenomenal anger serves to ‘prime’ the credibility of simulated anger.
III. Is There Intrinsic Epistemic Value in Feeling Anger?
Beyond strategy, there is the question of whether the felt experience of anger confers any epistemic or affective insight. There are two candidate answers:
(1) Anger as Moral Salience Tracker
On some views (see Martha Nussbaum, Patricia Greenspan), anger functions as a moral epistemic heuristic—it tracks violations of one’s normative expectations, value structures, or perceived status. The phenomenology of anger may reveal commitments or expectations that were previously tacit.
Here, to feel anger is to learn something about oneself: what one values, where one’s boundaries lie, what one implicitly demands of others. This cannot be faked—simulation does not afford this epistemic access. So there may be a self-revelatory benefit in undergoing anger.
(2) Anger as Integrative Affective Processing
If affective regulation is not just a matter of suppression but integration (cf. Antonio Damasio), then undergoing the emotion—processing it, resolving it—may lead to more stable long-term regulation, whereas suppression or simulation may lead to rumination, somatisation, or emotional misalignment.
In this sense, anger may be an internal coordination mechanism: aligning motivation, action, and cognition in ways that produce consistent behaviour over time.
Conclusion: The Benefit of Actually Being Angry
So: is there any benefit to actually being angry, even for a skilled dissimulator?
Yes, but the benefit is conditional and domain-sensitive.
In fast, high-stakes interactions, anger may be a computationally efficient optimiser of behavioural responses.
In reputational economies, genuine anger sustains signal credibility, even when simulation is possible.
In moral and psychological domains, anger plays a self-revelatory and integrative role that simulation cannot replace.
However, for an agent with near-perfect affect simulation, exceptional executive control, and high social predictive accuracy—perhaps an idealised Stoic or hyper-adaptive strategist—phenomenal anger may become largely redundant, a kind of evolutionary vestige.
But no real-world agent fits that model. Therefore, for most agents in most contexts, actually being angry remains functionally useful, not merely as an epiphenomenon of human psychology, but as a component of a computational architecture optimised for imperfect information, high emotional bandwidth, and reciprocal signalling.
Would you like to explore how this intersects with Stoic or Buddhist critiques of anger, or with the neuroeconomic modelling of affect in social games?
