What is self-deceit?

Edit 2026-01-10: Looking back now, the subagents model could be more fleshed out, but it’s a solid startng point. This piece was helpful (and fun) to write at the time, and I hope you find it useful too. I’ve found some interesting/more readily practical descriptions since then, notably the frameworks used in The Mind Illuminated and Multiagent Models of Mind, as well as The Hostile Telepaths Problem, which directly addresses self-deceit.

To deceive ourselves is a logical contradiction we nonetheless manage to accomplish every day.

When we say, “I’ll start that big project tomorrow for better quality. I’m not just seeking temporary comfort; I work better under pressure,” we hold a state of wrong belief and hide or rationalize the opposing evidence for comfort. This is the conventional self-deceit. However, deceit, by definition, involves one party hiding the truth from another. If the deceiver and the one being deceived are the same person, where is the truth hidden? Such a clear impossibility forces us to question our most basic model of a single, unified self. If “I” is not one, what is it?

Evolutionary Origins

Robert Trivers (2011) prominently resolves the paradox by defining “self” to be a conscious thinker who uses input passed from an unconscious. Deceit happens when the unconscious hides information to maximize perceived reward, originally for evolutionary gain (Trivers, 2011). In a tribe of our forefathers, a good liar can take credit for a hunt they only stumbled upon, project false strength to scare off rivals for mates, and acquire more resources by faking greater need or loyalty to the chief. With such high advantages, our brains, which did not care about high-minded ideals like Truth, readily began fighting to become the best deceiver. It is in the throes of this arms race towards deceiving others that humans evolved to start deceiving themselves. The most convincing liar is the one who believes their own lies. By hiding uncomfortable truths like a lack of faith or a rival’s true strength from the conscious mind, the unconscious eliminates the tells of deception, thus making interpersonal deceit far more efficient and evolutionarily advantageous.

The Issue

At first glance, this evolutionary explanation seems sufficient; that we rationalize away uncomfortable truths for social benefit explains some samples. However, while Trivers’s model is useful and gives us the origin (reproductive advantage), it leaves us with a black box that falters in other instances.

Take, for instance, a chronic smoker who sincerely vows to quit on Sunday, believing “I am now a non-smoker,” only to buy a new pack on Monday, rationalizing “I need a break today; just one pack won’t hurt. I’m still a non-smoker.” In the Triversian view, which part is the deception? Is the unconscious, which supposedly knows the truth (that quitting is best and the one pack signifies defeat), lying to the conscious self by endorsing the delay? If so, why would a reward-maximizing unconscious suddenly switch sides and endorse a self-destructive belief? Is the unconscious being deceived too? Simply partitioning the mind into “conscious” and “unconscious” kicks the can down the road.

This problem of internal division has been tackled philosophically, for instance by Sartre’s concept of “bad faith,” which asserts a consciousness that actively flees from uncomfortable truths (Detmer, 2013). While phenomenologically useful, these accounts provide no causal mechanism, which is what a psychological definition requires. To understand how this fleeing is possible and truly define self-deceit, we must turn to the brain’s architecture.

Agency, All The Way Down

Drawing from Minsky’s “Society of Mind,” we can model the brain as an ecosystem of competitive, specialized agents: Modern neuroscience confirms the brain is a network of specialized modules, and the agentic nature of these modules is inherited from the very cells they are composed of (Bertolero et al., n.d.; Dennett, 2018). Our neurons, we must remember, are descendants of the selfish, unicellular organisms that fought to survive for billions of years (American Museum of Natural History, n.d.). Neurons compete for resources, forming small coalitions that scale up into functional units that carry out tasks like edge detection in vision (O’Reilly et al., n.d.). These then group into larger networks that carry out more complex tasks (e.g. recognize faces), which in turn are marshalled by even greater systems that represent fundamental drives like hunger, fear, and social affiliation (Yu et al., 2024; Althubeati et al., 2022). Since unused neurons will be starved of resources and die, each neuron and module, can thus be described as having a desire to be useful. While these modules do not desire or cognize in the human sense, modelling them this way is powerful. It precisely predicts, for example, how the vision system in the blinded quickly repurposes itself for auditory and sensory input (Fauth and Norton, n.d.; Castaldi et al., 2020).

All these neuron modules cannot just compete, however; they need a continual source of sustenance, a way to interface with the external world. The mind fulfills this task with a high-level agentic module, a press secretary or CEO, which we know as the conscious self. This outward- and inward-facing executive is the one who carries out the commands of other agentic neuron bundles. Since it must interact in a highly social environment, it is also tasked with creating a coherent narrative of “I,” but contrasting with Trivers’s “conscious,” it only reasons posthoc, not to independently arrive at the most logical decision.

One might object that the conscious self can impose its will; it certainly feels, at first glance, like willpower is something a unified “I” must conjure. However, evidence suggests willpower is itself the output of a competing agentic system, centered in the anterior mid-cingulate cortex (Touroutoglou et al., 2019). The outcome, discipline, is simply the result of a dominant tenacity module. The same can be said for the illusion of an “I” thinking. Thought is actually agents bargaining, streamlined into a narrative by the conscious. This aligns with the experience that all thoughts and desires, like sensations and sounds, simply arise in awareness unprompted. The initial feeling of a unified self is indeed inconsistent, as we also see with voices.

The model of bargaining agentic modules explains the “voices” often mentioned in accounts of experience, such as the common “inner critic,” “inner child,” and even Schizophernia (Sreenivas, 2024). While this negotiation does not operate with explicit thoughts (the words are added during posthoc interpretation), the fact that we interpret their behavior as external just exemplifies their agentic nature.

A purely interpretative consciousness driven by competing actors is not just a useful model but a biological reality, revealed from a study of split-brain patients. In one experiment, as documented by Arakaki et al. (2024), the command “walk” was shown only to the patient’s right hemisphere. The patient complied and got up to walk. Then, the subject’s left brain (disconnected from the right), which houses the primary language module, was asked, “Why did you walk?” Instead of admitting ignorance, it instantly rationalized a story: “Because I was thirsty and wanted to get a Coke” (Arakaki et al., 2024). We very clearly see how the conscious self simply listens to and justifies the information passed by inner neuron modules, regardless of its veracity.

Paradox Dispelled

With this agentic model in place, the smoker’s dilemma is no longer a mystery. It is not a single unconscious deceiving a single conscious, but a power struggle between internal factions, which we will call Health, Nicotine and Stress-Response. Health (coalitions of neurons representing future planning and fear of death) takes control of reporting Sunday night and sends the message: “We must quit.” Unfortunately, Nicotine (a module wired for immediate gratification from cigarettes) has been gaining strength with every use. By Monday noon, its neurons are starved of resources and flood the conscious with stress data. Colluding with Stress-Response, they overwhelm Health with a new, contradictory report to the conscious: “Buying another pack is the right thing to do.” The conscious then rationalizes: “I am too stressed today, and stress is bad for my health. A cigarette will calm me down, making this a temporary, strategic health decision.”

This model is not just a metaphor; it reflects lived experience. In fact, addiction recovery involves recognizing and labeling the “voice of the addiction” (Grover, 2021). As the creator of one such technique recounts, clients describe this voice as a distinct entity, a “little devil on their shoulder” trying to trick them: “‘Take the edge off,’ the Beast urged. ‘Have just one…You need it’” (Trimpey, 1996, pp. 158, 278). This familiar feeling of uncontrollable voices is a testament to the model of agentic modules.

Self-deceit therefore cannot be a stable state of delusion; no logical thinker is being fed false information like in Trivers’s model because there is no thinker. Instead, every act is the visible outcome of an ongoing civil war in the mind. The conscious simply does and justifies what its told by the current victor, which was Health while Nicotine was satiated and dormant. “Self” was deceived when Nicotine passed on a message that is inaccurate (“A pack is necessary”). However, this suggests true veridicality only occurs when the winning agent perfectly reflects reality. Is this possible?

The Eternal Lie

The internal politics that led to Nicotine, in the smoker’s case, deceiving the conscious were innate to a system of competitive agents like the mind. Thus, deceit, to some degree, is necessary for healthy function.

Think about the conscious as a CEO. A CEO needing to act and rationalize cannot function if they know every single detail of each employee’s work and all the small factory malfunctions. They would be paralyzed. They need abstracted, high-level summaries.

The process of abstraction required for the high-level agent to function operates on the same underlying mechanism as deception: sub-agents hide information to be useful. The agent managing physical sensation does not tell the conscious, “The pressure on your left foot is 102 kPa, the temperature is 94.7°F, and 7000 nerve endings are firing.” It says, “Things are fine.” This intended hiding of truth is not a moral flaw.

Now apply this to something more fraught. Imagine you mildly embarrass yourself at a party. The Social-Anxiety Faction might want to replay this event on a loop forever. A healthy Productivity Faction will suppress this information, telling the CEO, “The party went well,” not because it is lying in the Triversian sense (to aid interpersonal deception), but because obsessing over the micro failure is counter-productive to the organism’s larger goals (like sleeping that night). Internal deception is not always malicious; it is sometimes required.

One might argue that deception implies intent to mislead, whereas abstraction is simply efficient data compression. But self-deceit is a spectrum. At one end is functional, necessary information-gating that makes a coherent “I” that functions in society possible. At the other end is malignant delusion. The agentic model reveals there is no clear line between them; they operate on the same fundamental principle of sub-agents hiding information from the executive self, motivated ultimately by a self-interest for resources. True veridicality is pragmatically impossible.

Self-deceit is hence not a unique, pathological process; it is the application of this necessary, everyday mental function of abstraction that, in the extreme, becomes pathological. Deceit always occurs but is detrimental when a mental faction’s “summary” is not for the good of the whole organism, but solely for its own, short-term survival.

Conclusion

The paradox of self-deceit is resolved when we discard the view of a single self. By modelling the mind as an ecosystem of competing agents, the conscious self emerges as simply one of such agents tasked to interface reality by fulfilling commands of the rest and creating a coherent “I.” This conscious self operates only with summaries that result from a continuous mental civil war. Self-deception is then a natural result of healthy abstraction carried out by self-preserving agents.

Understanding self-deceit this way does not leave us as helpless puppets. On the contrary, we now have a new strategy for growth: treat the mind as the ecosystem it is. Ally with the healthier factions, especially when harmful ones are dormant. Nourish them through reinforcement and defund the parasites with abstinence. The goal is no longer to achieve an impossible unity with brute rationality, but to moderate a functional, if healthily deceptive, coalition we can live with.



References