Vladisav Jovanovic

Self-evaluation often depends on social and performance signals. Research on contingencies of self-worth shows that success and failure matter more strongly in domains on which people stake self-esteem, while sociometer theory links self-esteem to perceived relational inclusion. This paper develops fixed worth as a normative counter-position to such contingent valuation: basic human worth should not be treated as something earned or lost through approval, rank, desirability, usefulness, correctness, or current performance. The paper uses pricing self as an interpretive label for situations in which external valuation becomes unusually central to self-regulation. The central hypothesis is modest: when self-worth is highly contingent on a domain, threats in that domain may increase defensiveness, audience monitoring, and susceptibility to influence, whereas broader and less contingent sources of self-integrity may make corrective information easier to tolerate. This proposal is situated alongside research on contingent self-worth, the sociometer hypothesis, self-affirmation, and self-determination. The term invariance is retained as a philosophical metaphor for non-tradability, not as a literal control-theoretic property of the psyche. The paper concludes by applying the framework to environments saturated with rankings, metrics, and feedback, where the practical question is not whether people can become unsteerable, but whether fewer judgments about status or performance need to function as verdicts on personal worth.

Bodily states matter for judgment, but they are not direct detectors of truth. Interoception provides information about the physiological condition of the body, while contemporary models emphasize that bodily experience reflects both ascending signals and predictions about bodily state. This paper develops a cautious account of the relation among felt resonance, distress, embodiment, and judgment. Resonance refers to a subjective sense that an interpretation fits or brings relief; embodied distress refers to bodily and affective signals such as tension, arousal, fatigue, pain, or shutdown. Neither is evidentially self-authenticating. Bodily signals can reveal that something matters, that stress is high, or that a situation is physiologically costly, but the cause of those signals remains an empirical question. The paper therefore rejects the stronger claim that suffering is structurally necessary for truth or that pain should be preserved as an epistemic gate. Pain can be adaptive, maladaptive, acute, chronic, or poorly related to present external danger, and reducing suffering can be medically and psychologically beneficial. Generative AI introduces a different asymmetry: it can produce empathic or coherent language without human embodiment, while users still bear the practical consequences of acting on its outputs. The resulting framework treats the body as a source of relevant data, not a sovereign oracle, and evaluates insight through evidence, consequences, and revisable action rather than through the amount of suffering it produces.

Ideas can feel true before they have been adequately tested. Familiarity, processing fluency, repetition, affective fit, identity protection, and social reinforcement can all increase subjective ease or confidence without corresponding increases in accuracy. This paper uses resonance as an umbrella label for felt fit or rightness and argues that resonance should not be treated as evidence by itself. The proposal is situated within research on processing fluency, the illusory truth effect, affect-as-information, motivated reasoning, and epistemic vigilance. The paper replaces a quantum-style formalization with a simpler calibration model that separates three questions: how compelling a claim feels, what evidence supports it, and how the claim responds to warranted correction. Epistemic integrity is defined operationally through provenance, discriminating evidence, uncertainty, and correction behavior rather than through a multiplicative binding equation. Generative AI intensifies the practical problem because fluent, personalized language can increase perceived fit while remaining weakly grounded. A five-step resonance check is proposed for human and AI-assisted reasoning: identify the source of felt rightness, state the claim precisely, inspect independent support, search for realistic disconfirmation, and specify what would change after error. The contribution is a psychologically grounded framework for separating felt certainty from evidential warrant.

Acute stress and social threat can impair working memory and cognitive flexibility, but the effect is variable rather than a total shutdown of reason. This paper develops judgment under activation as a practical problem at the intersection of stress, metacognition, psychological flexibility, and need for cognitive closure. The useful phenomenon in the source concept of “spiritual intelligence” is retained without treating it as a validated, distinct intelligence faculty. The central risk is premature meaning: a rapid interpretation that reduces uncertainty or emotional tension and is then granted more evidential authority than it deserves. Under activation, sophisticated reasoning and social skill can still operate while serving a narrowed objective such as immediate threat reduction, status defense, or interpersonal control. A five-part interruption sequence is proposed: identify activation, delay irreversible action, state the live interpretation as a hypothesis, seek disconfirming information, and specify what evidence would justify revision. The account is grounded in research on stress effects on executive function, psychological flexibility, cognitive closure, and metacognition. It also clarifies that calmness is not proof of good judgment and activation is not proof of irrationality. Generative AI may amplify rapid closure by supplying fluent explanations on demand, making uncertainty and verification cues more important when users are highly activated.

Hallucination in natural-language generation typically refers to unsupported, unfaithful, or factually incorrect output. This paper develops hallucination regime as a broader socio-technical concept for environments in which fluent generated text enters decisions faster than it can be verified and responsibility for checking becomes diffuse. The framework connects three levels of analysis: model hallucination, human automation bias and over-reliance, and institutional accountability. Its central mechanism is verification displacement: plausible generation reduces the friction of completion, the user treats that completion as sufficient to proceed, and an institution circulates or authorizes the output without assigning clear ownership of evidential review. Retrieval, citations, and disclaimers can reduce some errors, but they do not by themselves establish that claims have been adequately verified or that responsibility for correction is clear. The paper proposes an audit organized around provenance, verification, uncertainty, consequence, and correction. The central claim is that the social danger of hallucination depends not only on what a model generates, but on whether the surrounding workflow preserves explicit responsibility for checking and repairing error.

Digital systems often treat a click, prompt, trade, purchase, or command as a transparent expression of stable preference. Human decision research gives reason to be more cautious: preferences can vary with affective state, cognitive load, fatigue, urgency, and temporal perspective, and people sometimes use commitment devices to protect later choices from foreseeable short-term impulses. This paper develops sovereign and reflex agency as user-facing shorthand for a narrower design problem: when should a system treat a momentary signal as sufficient authorization for a high-stakes or difficult-to-reverse action? The paper rejects hidden classification of trauma, dissociation, or “true will” from behavioral traces. Instead, it proposes an autonomy-preserving confirmation architecture based primarily on properties of the action—stakes, reversibility, novelty, and temporal instability—together with explicit user preferences. Safeguards include staged execution, confirmation after delay, reversible defaults, transaction limits, and user-configurable commitment settings. These measures should be transparent, appealable, minimally invasive, and designed so that friction protects reflection without becoming coercion. The contribution is therefore not an ontology of inner agency and not a mind-reading classifier. It is a design framework for treating irreversible action as requiring stronger evidence of endorsement than a single transient input provides.

Contemporary AI systems can generate fluent first-person claims about feelings, fear, desire, selfhood, and consciousness. This paper argues that such self-report-like language is insufficient evidence of sentience. Sentience is understood minimally as phenomenal experience, including some valenced distinction between better and worse for a subject. The paper develops three complementary evidential dimensions for evaluating claims of artificial sentience: persistent internal constraint, endogenous normativity, and continuity of first-person stakes. These dimensions are not proposed as necessary and sufficient conditions for consciousness, but as a conceptual filter for distinguishing generated self-description from stronger evidence of an experiencing point of view. Language-model self-reports are weak evidence when they can vary substantially with prompting and are not independently supported by architectural or behavioral indicators of persistent internal stakes. The paper argues for an evidential rather than anthropomorphic approach to artificial consciousness: assessments of sentience should rely on converging evidence from system architecture, behavior, temporal continuity, and theories of consciousness rather than on conversational fluency alone.

Wisdom is not well described as secret knowledge possessed by a spiritual elite. Contemporary psychology instead studies wisdom through capacities such as intellectual humility, perspective taking, recognition of uncertainty and change, integration of competing interests, and practical deliberation. This paper develops the monastic-effect hypothesis as a bounded environmental proposal: some forms of renunciation or low-volatility contemplative life may reduce status competition, novelty, distraction, and role performance in ways that make certain self-protective narratives harder to maintain. The hypothesis does not claim that monks are necessarily wiser, that deprivation reveals truth, or that contemplative practice reliably produces wisdom. Research on meditation and mindfulness contains substantial methodological limitations, and wisdom itself varies across situations and social environments. The proposed contribution is therefore social-ecological rather than mystical. Reduced stimulus and repeated practice may sometimes change the conditions under which reflective judgment occurs, but self-selection, tradition, community structure, and prior motivation must be separated from causal effects of renunciation. Generative AI adds a contemporary contrast by making fluent explanation inexpensive and immediate. Under those conditions, wise judgment depends less on producing a coherent story and more on tolerating uncertainty, inspecting evidence, considering alternatives, and revising when circumstances change.

Fluent explanation can create an impression of resolution even when a problem remains uncertain, costly, or poorly tested. This paper offers a heuristic taxonomy of seven capacities that can improve judgment after an initial coherent interpretation has formed: ambiguity tolerance, sensitivity to disconfirming constraint, cost visibility, repair capacity, exit discernment, non-coercive interpersonal containment, and meta-orientation. These are not proposed as newly discovered functions of consciousness, fixed stages of development, or a validated model of wisdom. Each has established neighbors in research on cognitive closure, ambiguity tolerance, psychological flexibility, self-distancing and wise reasoning, organizational learning, and exit and voice. The paper's contribution is organizational: it asks how these capacities interact when a person must remain functional without converting uncertainty into premature certainty. Generative AI makes the question practically important because it can provide rapid, plausible completion that may reduce the felt need to search, verify, or tolerate an unresolved state. The taxonomy therefore distinguishes interpretive closure from better judgment and proposes observable questions for each capacity. The framework is strongest as a comparative checklist for reasoning under uncertainty; its stronger claim that the capacities jointly constitute wisdom remains a hypothesis requiring validation.

Humans often seek orientation under uncertainty, and the desire for a coherent interpretation can sometimes compete with accuracy. This paper develops coherence-first cognition as a bounded hypothesis about judgment under uncertainty rather than a theory of consciousness. The proposal is situated alongside research on need for cognitive closure, motivated reasoning, meaning maintenance, emotion regulation, and meaning-making after stressful events. These literatures support a weaker and more testable claim than the idea that consciousness universally optimizes coherence before truth: under some conditions—especially uncertainty, threat, time pressure, and identity relevance—people may prefer interpretations that restore predictability, reduce ambiguity, or protect valued commitments. Reasoning, emotion regulation, and existential meaning-making can all support orientation, but none is treated as a single “coherence engine,” and spiritual intelligence is not assumed to be a settled scientific construct. Generative AI makes the problem more visible because it can supply rapid, plausible closure even when evidential support is weak. The paper therefore proposes an epistemic restraint principle: the more a response reduces uncertainty through fluency or emotional fit, the more important it becomes to inspect evidence, alternatives, and conditions for revision. In interpersonal contexts, restraint also means distinguishing helpful interpretation from unsolicited certainty about another person's inner life. The contribution is a synthesis linking orientation needs to the risk of premature closure in AI-mediated reasoning.

Insight can matter in psychotherapy and self-understanding, but insight is not identical to behavior change. A meta-analysis of psychotherapy research finds a moderate association between insight and outcome, while broader behavior-change research shows that changes in intention often exceed changes in action. This paper develops insight without contact as a practical description of cases in which an explanation becomes more coherent while the behavior, relationship pattern, or decision process it is meant to change remains substantially unchanged. The phrase “this makes sense, but it does not work” functions as a useful boundary marker between explanatory fit and effective revision. Five tests are proposed: counterevidence, behavior under pressure, recurrence over time, repair after failure, and relational consequences. The paper also develops a bounded ethics of interpretation. In clinical settings, informed consent protects autonomy and can strengthen trust; in ordinary relationships, unsolicited psychological interpretations can likewise be intrusive or counterproductive when they are presented as authoritative facts. Consent is not an absolute rule against difficult truth-telling, and safety or professional duties can override ordinary conversational restraint. The central claim is therefore twofold: insight should be judged partly by what changes under recurrence, and interpretations about another person's inner life should be offered with proportionality, uncertainty, and respect for their agency.

Accurate or well-supported information does not automatically produce psychological or behavioral change. People can understand a painful fact, acknowledge a failed prediction, or receive corrective evidence while remaining defensive, avoidant, or behaviorally unchanged. This paper uses truth-load as a metaphor for the psychological integration demands created by self-relevant corrective information. The concept is explicitly separated from truth itself: a claim is not made more true because it is painful, costly, destabilizing, or difficult to accept. The paper situates the integration problem alongside cognitive dissonance, motivated defense, self-affirmation, information avoidance, self-distancing, and meaning-making. It proposes a bounded sequence in which evidence is assessed first, then the personal implications of warranted information are examined, and finally integration is evaluated through changed belief, action, or coping over time. Self-threatening information may provoke defensive processing, while broader self-integrity and psychological distance can sometimes improve openness. The paper also develops an ethical implication: timing and support matter when communicating highly self-relevant information, but distress is not a reason to withhold warranted truth indefinitely or to treat another person's interpretation as authoritative without evidence. The contribution is a vocabulary for the gap between propositional acceptance and lived integration, not a new theory of truth.

Generative AI lowers the cost of producing fluent, internally coherent language. This paper argues that the resulting change is best understood not as a new theory of truth but as a change in the reliability of familiar cues. Processing fluency, repetition, and polished explanation can influence credibility judgments, while language models can generate fluent output that is unsupported, fabricated, or preference-aligned. The paper retains the author-defined terms cheap coherence, contact, answerability, and Structural Intelligence, but maps them to established work on processing fluency, epistemic vigilance, misinformation, provenance, hallucination, and corrigibility. Its candidate contribution is a practical synthesis: when coherent language becomes abundant, reliability assessment should place more weight on independent constraint, correction, consequence, provenance, and durable but revisable follow-through. A five-channel taxonomy is offered as a heuristic rather than a validated scale, with explicit limits and disconfirmation conditions.