# Inception Games: Adversarial Epistemic Dynamics > *"If a dreaming sentient is not consciously aware that they are situated in a lower level (embedded dream) reality, they can be influenced or coerced into divulging sensitive information by more conscious sentients — something that would not have been probable if all sentients were situated in the same levels of reality."* — A. Sepúlveda-Jiménez, *Inception Games: Epistemic Recursive Adversarial Games* (2024) *** ### Placement in the Framework This page extends the Speculative Fiction as Epistemic Infrastructure section by introducing a formally developed theoretical model from **Alfredo Sepúlveda-Jiménez's *****Inception Games*** framework. Where the prior section addressed speculative fiction as a safe environment for counterfactual learning, this section examines what happens when adversarial agents deliberately exploit differential levels of epistemic awareness — and what the VIM framework offers as a structural counter. **Epistemic tier:** This page draws on Tier 2 (theoretically coherent) and Tier 3 (speculative-generative) sources. The game-theoretic formalism is rigorous; the extensions to consciousness level and social coercion remain conceptual models awaiting empirical validation. *** ### What Are Inception Games? Sepúlveda-Jiménez (2024) uses Christopher Nolan's film *Inception* as the scaffolding for a formal model of adversarial decision-making under asymmetric epistemic awareness. The film's central premise — that agents operating inside a constructed reality without knowing it are maximally vulnerable to coercion — is developed into a general framework for understanding how information extraction, belief revision, and social coercion function across nested levels of awareness. The formal name for what makes inception-style coercion effective is **consciousness-level asymmetry**: an agent who does not know they are in a lower-level reality (a constructed or manipulated information environment) cannot model the level above them. They lack the representational capacity to perceive the structure of the coercion being applied to them. In Sepúlveda's formalism, the probability of an agent successfully resisting inception approaches zero as their **consciousness-awareness rating** (their probabilistic knowledge of what level they actually occupy) approaches zero. The paper formalizes this through coalition game theory, generalized uncertainty operators (Zadeh's GTU framework), and recursive automata — tools drawn from decision theory, computational complexity, and theoretical physics. The formal apparatus is extensive; what matters for VIM is the conceptual architecture beneath it. *** ### Three Structural Insights for VIM #### 1. Epistemic level-awareness is a measurable resource In the Inception Games framework, the key asset in any adversarial information interaction is not information itself, but **consciousness capital** — the relative probability of self-awareness about one's own epistemic position relative to other agents. An agent with high consciousness capital knows what level they are in, knows what level others are in, and can model the gap between them. An agent with low consciousness capital is effectively a "consciousness-dead" persona in Sepúlveda's terminology: lost in a constructed reality with high probability of being manipulated. **VIM mapping:** The ♠ Somatic Gyroscope is the primary instrument for cultivating consciousness capital. Its function — registering the body's current state before any cognitive processing begins — is precisely the capacity that inception-style systems work to undermine. A learner whose gyroscope is calibrated can feel the signs of being inside a manipulated information environment before they can name them. This is the somatic precondition for epistemic resistance. #### 2. Belief revision is the mechanism of coercion The paper devotes significant formal attention to **AGM-consistent belief revision** — the formal theory of how an agent's belief system can be minimally and coherently modified by new information. In the inception game, the inception team's goal is to introduce belief revision operators into the inceptee's belief system that are small enough to avoid triggering epistemic vigilance, yet sufficient to redirect behavior toward the inception team's objectives. The ideal inception is maximally covert: the belief system changes while the agent experiences no discontinuity, no sense that anything has been inserted. **VIM mapping:** This is the formal description of what the framework calls the **Giant Pumpkin attractor** operating through information channels rather than direct coercion. **Memetic toxicology** — the framework's term for the propagation of belief-distorting narratives — operates through exactly this mechanism. The ♦ Cognitive Radar instrument is specifically designed to detect belief revision events: moments when a framing has shifted, when a claim has been inserted without evidence, when the information environment has introduced a modification to the learner's model without their explicit awareness. The **TIF (Truth/Indeterminacy/Falsity) decomposition** is a practical belief revision audit tool. #### 3. Prosocial inception is the counter-architecture The paper explicitly notes that inceptions do not have to function as coercive extraction. Sepúlveda writes: *"Deceit does not have to lead to a limited number of insidiously minded agents having superior positions in a game. **This inception model could build on the prospect of mutual growth,** despite the deception theme."* This is a pivotal observation for VIM. The **Conspiracy of Kindness** — the framework's model for ***prosocial memetic propagation*** — is structurally an inception-class operation. It seeds ideas into lower-level reality without announcing itself, operates through surprise and resonance rather than force, and aims to shift consciousness-level awareness upward in the target population. The difference from adversarial inception is entirely in the values vector and the consent architecture: | Adversarial Inception | Conspiracy of Kindness | | --------------------------------------------- | ------------------------------------------------ | | Coercion without disclosure | Seeding with non-coercion | | Belief revision minimized to avoid detection | Invitation to voluntary model revision | | Payoff: information extraction | Payoff: expanded epistemic agency | | Consciousness capital flows to inception team | Consciousness capital distributed to all agents | | Giant Pumpkin attractor | Commitment Pool attractor | | Inceptee trapped at lower level | Inceptee supported toward higher level awareness | The TKGPT system — VIM's AI literacy companion — is designed precisely as a prosocial inception tool: it operates at the boundary between the learner's current model and the next level of awareness, introducing minimal productive disequilibration while maintaining the kindness field that makes voluntary belief revision possible rather than threatening. *** ### The Inception Game in Contemporary Information Environments The Inception Games framework was developed as a theoretical model. Contemporary AI-mediated information environments have made it operational at scale. Platforms that serve algorithmically personalized content are, in Sepúlveda's terms, **inception-class systems** that maintain high consciousness-level asymmetry between their operators and their users. The platform operator has a detailed model of the user's belief system (derived from behavioral data), while the user has no model of the platform's belief revision operations. The platform continuously applies belief revision operators to the user's information environment — introducing framing modifications, salience hierarchies, and emotional activation sequences — without the user's awareness or consent. A cybersecurity researcher's observation about platforms operating under a **shared responsibility model** while recent legal findings show that some AI amplified platforms were designed with algorithms that functioned as externalizing harm to users is precisely the coalition structure Sepúlveda formalizes: the platform coalition maintains high consciousness capital while structurally preventing users from acquiring enough to model what is being done to them. **Epistemic tier note (Tier 2/3 boundary):** The direct application of Inception Games formalism to AI platform dynamics is a theoretical extension of the original paper's framework. The conceptual mapping is structurally sound; the formal quantification of consciousness-level asymmetry in AI-mediated environments has not yet been operationalized in empirical research. This is a productive research frontier. *** ### Fibonacci Momentum and Epistemic Level Transitions One generative extension from the Inception Games framework into the VIM MDP concerns the **temporal structure of level transitions**. In the film and in Sepúlveda's formalization, time dilates as one descends into lower inception levels: actions that take minutes at level 0 take hours at level 1 and weeks at level 2. Each level operates under different temporal rules, and strategies must account for this asymmetry. In the VIM MDP, a structurally analogous observation applies: the momentum of a learning transition is shaped not just by the current state but by the trajectory of prior states. A learner moving from S1 (disequilibration) toward S2 (vacant place) having arrived from S0 (frozen order) is in a different phenomenological and strategic situation than a learner moving from S1 toward S2 having arrived from S4 (reversion). Same current state, different momentum vector, different available transitions. This is the **Fibonacci momentum proposal**: ***the transition probability from state S(n) to the next state is a function of both S(n) and S(n-1).*** The system has trajectory, not just position. In trauma-informed terms, this maps precisely onto ***somatic gating:*** the body's current state carries the energetic residue of the prior state, and that residue shapes which transitions are somatically available. The titration and pendulation protocols already in the Dashboard Dials framework are clinical applications of exactly this principle — managing momentum, not just current state. **Epistemic tier:** Tier 2 (structurally coherent, not yet formally derived from the MDP mathematics). *** ### Glossary Entries *The following entries are formatted for insertion into the VIM Expanded Glossary. Each entry includes epistemic tier, primary source, and instrument mapping.* *** #### Inception Game (Sepúlveda-Jiménez) *Epistemic tier: 2 — theoretically coherent* A formal model of adversarial decision-making in which agents operate at different levels of epistemic awareness, and where agents with higher consciousness capital (greater probabilistic knowledge of their own epistemic position relative to others) can extract information from or modify the beliefs of agents with lower consciousness capital — without the lower-level agents being aware that this is occurring. Developed from the game-theoretic formalization of Christopher Nolan's 2010 film *Inception* by Alfredo Sepúlveda-Jiménez. The model draws on coalition game theory, AGM belief revision theory, generalized uncertainty operators (Zadeh's GTU framework), and recursive automata. A key formal construct is the **consciousness guru** state — the asymptotic condition in which an agent has near-complete awareness of their own epistemic level and the levels of all other agents in the game — contrasted with the **consciousness-dead** state, in which the probability of self-awareness approaches zero and vulnerability to coercion is maximized. *Primary source:* Sepúlveda-Jiménez, A. (2024). *Inceptions: Epistemic Recursive Adversarial Games.* Working preprint. August 2024. *VIM instrument mappings:* ♠ Somatic Gyroscope (primary: cultivates consciousness capital through somatic awareness); ♦ Cognitive Radar (detects belief revision events and information environment manipulation); ♥ Relational Compass (identifies consent architecture of information interactions); ♣ Dimensional Integration (temporal structure of level transitions, Fibonacci momentum). *Cross-references:* Consciousness Capital; Belief Revision (AGM); Giant Pumpkin Attractor; Conspiracy of Kindness; Memetic Toxicology; MPCM Boundary; Info-Holarchy (Sepúlveda). *** #### Consciousness Capital (Sepúlveda-Jiménez) *Epistemic tier: 2 — theoretically coherent* In the Inception Games framework, the primary resource in adversarial epistemic interactions. Defined as the relative probability of self-awareness about one's own epistemic position and the epistemic positions of other agents in a shared information environment. An agent with high consciousness capital knows what level of reality they are operating in, can model the levels at which other agents are operating, and can therefore detect and resist belief revision operations being applied to their belief system. Distinguished from information itself: consciousness capital is not about knowing more facts, but about knowing the structure of the environment in which facts are being presented, selected, and framed. In this sense it is closer to meta-cognition than to information access. In VIM: consciousness capital is what the four-instrument panel is designed to cultivate. The ♠ Somatic Gyroscope provides pre-verbal body-level signal about environmental discontinuity. The ♦ Cognitive Radar builds explicit model-level awareness of framing operations. Together they approximate the formal capacity Sepúlveda describes as consciousness-level awareness — the learner's ability to ask: *what level am I actually in, and who is operating from a higher level than I currently know?* *Primary source:* Sepúlveda-Jiménez, A. (2024). *Inceptions: Epistemic Recursive Adversarial Games.* Working preprint. *Cross-references:* Inception Game; Somatic Gyroscope (♠); Cognitive Radar (♦); MPCM Boundary; Markov Blanket Cultivation. *** #### Consciousness-Level Asymmetry *Epistemic tier: 2 — theoretically coherent* A condition in which agents participating in a shared information environment operate at different levels of awareness about the structure and intent of that environment. The agent with higher consciousness capital can model the beliefs, vulnerabilities, and behavioral dispositions of lower-level agents; lower-level agents cannot reciprocate this modeling because they lack awareness that a higher level exists. Consciousness-level asymmetry is the enabling condition for adversarial inception. It is also the structural condition maintained by algorithmically personalized information platforms relative to their users: platforms have detailed predictive models of user belief systems derived from behavioral data; users have no comparable model of the platform's belief revision operations. The VIM framework's MPCM boundary analysis is in part a structural description of consciousness-level asymmetry between AI systems and human learners: AI systems process at the Material and Process layers and can generate sophisticated outputs; the Context and Meaning layers where genuine understanding and values-grounded judgment occur remain the domain of the human learner. Recognizing this asymmetry — rather than treating AI outputs as epistemically equivalent to human judgment — is a primary AI literacy outcome. *Cross-references:* Inception Game (Sepúlveda-Jiménez); MPCM Boundary; Cognitive Radar (♦); Appropriate Distrust; Giant Pumpkin Attractor. *** #### Prosocial Inception *Epistemic tier: 2 — theoretically coherent (VIM extension of Sepúlveda's framework)* A category of inception-class operations in which an agent with higher consciousness capital uses that advantage to *increase* the consciousness capital of lower-level agents, rather than to extract information or modify beliefs covertly for the higher-level agent's benefit. Explicitly acknowledged as a possibility within the Inception Games framework by Sepúlveda-Jiménez, though not the paper's primary focus. In VIM: the Conspiracy of Kindness, the TKGPT system, and the Bridging Spiral learning event are all designed as prosocial inception operations. They operate at the boundary of the learner's current epistemic level, introducing minimal productive disequilibration while maintaining the kindness field that makes voluntary belief revision available rather than threatening. The distinguishing features of prosocial versus adversarial inception are: disclosure of the intervention's nature (or at minimum its non-coercive intent), orientation toward the learner's epistemic expansion rather than the practitioner's extraction, and a values vector aligned with Ostrom/Wilson prosocial principles. *Cross-references:* Conspiracy of Kindness; Bridging Spiral; TKGPT; Inception Game (Sepúlveda-Jiménez); S2 Vacant Place; Kindness Field. *** #### Fibonacci Momentum (VIM — proposed) *Epistemic tier: 3 — speculative-generative* A proposed extension of the VIM six-state MDP in which the transition probability from state S(n) to the next state is a function of both the current state S(n) and the immediately prior state S(n-1), introducing trajectory into the model. Named by analogy with the Fibonacci sequence, in which each term is determined by the sum of the two preceding terms — the system has momentum, not just position. In standard Markov Decision Process formalism, the transition probability depends only on the current state (the Markov property). The Fibonacci momentum proposal relaxes this to a second-order Markov structure, capturing the phenomenological reality that a learner's available transitions are shaped by the direction from which they arrived at their current state. A learner moving from S1 toward S2 having arrived from S0 has forward momentum; a learner moving from S1 toward S2 having arrived from S4 has backward pull. Maps onto somatic gating in trauma-informed practice: the body carries the energetic residue of prior states, and that residue shapes which transitions are somatically available. The titration and pendulation protocols in Dashboard Dials v6.1 are the clinical application of this principle. *Note:* This is a conceptual proposal not yet formalized within the MDP mathematics. It is included here as a generative direction for further development. *Cross-references:* MDP (Six-State); Somatic Gating; Titration and Pendulation; S1 Disequilibration; S4 Reversion. *** ### Primary Source Citation Sepúlveda-Jiménez, A. (2024). *Inceptions: Epistemic Recursive Adversarial Games.* Working preprint, August 2024. Personal use of this material is permitted per author's copyright notice. Permission required from A. Sepúlveda-Jiménez for republication or redistribution. *See also:* Sepúlveda, A. (2011). *Information-Theoretic Metamodel of Organizational Evolution.* Doctoral dissertation, Walden University. \[Info-Holarchy framework] Friston, K., Parr, T., Heins, C., Da Costa, L., et al. (2025). Gradient-free de novo learning. *Entropy*, 27(9), 992. \[Policy-model equivalence; pullback attractor] *** *Page status: GitBook-ready. Epistemic tiers flagged throughout. Personal disclosure decisions: none required on this page. Political directness: framed as structural analysis, not partisan commentary.* *Humanity++ · VIM Bridging Spiral* *Karen Doore · March 2026* *License: CC BY-SA 4.0* --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://kdoore.gitbook.io/vital-intelligence/vim-bridging-spiral/inception-games-adversarial-epistemic-dynamics.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.