Light World, Dark World
Mar. 10th, 2026 02:37 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
nemorathwaldThe following is a ~2,400-word AI re-write (lightly edited) of Duncan Sabien's 10,000+ word essay "Truth Or Dare". It was originally written as intentionally meandering, so if this synopsis interests you, you might wish to read the whole thing for the numerous examples and metaphors. I'd like more people to have access to these concepts without committing to something novelette-length.
Light World, Dark World: How People End Up in Incompatible Realities Of Danger
The Core Observation
People who share roughly the same socioeconomic circumstances, geography, and educational background often experience the social world in fundamentally incompatible ways. Some people move through life with a baseline expectation that others are basically trustworthy, that things will generally work out, and that openness and vulnerability are reasonable bets to take. Others move through the same world with persistent wariness, a strong prior that people have hidden malicious agendas, and a felt sense that extending trust is naive at best and dangerous at worst.
What makes this interesting is that neither group is wrong about their own experience. When you investigate the specific history of someone who believes the world is cold and threatening, you find a world that really was cold and threatening. Their worldview is well-calibrated to the data they've actually collected. The same is true in reverse.
What makes it stranger is that the distribution doesn't look like a bell curve. You'd expect most people to be in the middle, with a few outliers at each end. Instead, the distribution appears bimodal — people cluster toward the poles. Something is driving people out of the middle and into one camp or the other.
(c.f. the excellent Scott Alexander essay Different Worlds.)
The Mechanism
The explanation lies in a set of interlocking feedback loops.
Perceptual filtering. The brain cannot process everything it receives. It operates by building predictive models of the environment and then checking selectively for deviations from those models, rather than taking in data neutrally. This means that what you expect to see, you will tend to see. That's not because you're hallucinating, but because perception is a top-down process as much as a bottom-up one. A person who has learned to expect manipulation will genuinely notice potential manipulation in ambiguous signals, where someone else would not notice it at all. Attention is a finite resource allocated according to existing models of what matters. Whatever your brain has been trained to flag, you will find in abundance.
Behavioral shaping. People don't just filter information differently. They behave differently as a result of their filtered experience, and those behaviors elicit different responses. Someone who carries a default posture of wariness and defensiveness tends to generate wariness and defensiveness in others. Someone who radiates relaxed openness tends to elicit relaxed openness. This is not magical thinking; it's a straightforward consequence of the fact that social interaction is bidirectional. The signals you send shape what comes back. The person who enters every room braced for hostility will, with regularity, find hostility. That's partly because hostile people are attracted to easy targets and partly because their own bracing posture provokes friction.
Selection effects. Beyond direct behavioral feedback, there are structural sorting processes. High-trust social environments tend to eject people who repeatedly break the norms of trust — who fail to reciprocate, who treat gifts as commodities, who introduce low-trust behaviors into gift-economy relationships. This means people with dark-world orientations accumulate fewer and fewer access points to the very social environments that might offer them contradictory evidence. Meanwhile, when they do attempt to experiment with trust, they tend to do so in environments where the other participants are also relatively inexperienced with genuine trust, or where actual predators have positioned themselves after being ejected from high-trust spaces. The experiment fails. The worldview hardens.
Cluster formation. Over time, these sorting processes create geographically overlapping but experientially non-overlapping social worlds. Two people on the same street can have radically different networks, exposures, and baseline experiences, and each will have abundant evidence for their own worldview precisely because their evidence is drawn from a non-random sample of social reality.
The net result is a self-reinforcing system. Entry into either attractor makes exit progressively harder. The feedback loops tighten. The perceptual filters strengthen. The behavioral habits calcify.
---
The Prion Problem
A useful concept here is the prion. A prion is a misfolded protein that propagates by causing neighboring correctly-folded proteins to adopt the same malformation. Dark-world orientations behave similarly in social contexts.
Someone operating from a low-trust frame doesn't just protect themselves, they alter the social environment around them. When one member of a group starts tracking debts precisely, insisting on exact repayment, treating favors as transactions rather than bonds, the gift economy logic that held the group together begins to corrode. Other members start reciprocating in kind, not because they wanted to, but because continuing to operate in gift-economy mode in the presence of someone running market logic feels asymmetric and exploitable. The low-trust behavior propagates.
The same dynamic holds for anxiety and paternalism. A highly anxious person in a group doesn't just experience their own anxiety. Their visible distress activates threat-detection systems in others. A parent radiating fear about a child's safety creates a social pressure toward risk aversion that affects even parents who were themselves comfortable with the risk. You can feel the anxiety of others as information about whether danger is present, even when it isn't.
This is why even well-intentioned people from dark-world contexts can be destructive to high-trust social environments. The issue isn't malice. The issue is that the behaviors adaptive for one environment are disruptive in the other, and they propagate.
---
Why Dark-World Views Are Hard to Exit
The dark-world orientation is not just a persistent personality trait. It's something closer to a self-sealing epistemic system.
Consider what it would take to convince someone with a strong prior on manipulation that a given person's friendly behavior is genuine. Every piece of friendly evidence can be reinterpreted as a tactic. The friendlier and more consistent the behavior, the more sophisticated the manipulation presumably is. There's no observation that definitively disconfirms the hypothesis, because the hypothesis has built-in explanations for all apparent counterevidence. The belief system is so stable precisely because of unfalsifiability, similar to a theological worldview that attributes disconfirming events to divine mystery or satanic deception.
This is why well-meaning exhortation fails. Telling someone in the dark world to "just trust people more" is like telling someone to update a Bayesian prior by willpower. It isn't how it works. The prior is maintained not by a conscious choice but by a filter that processes incoming data before it ever reaches conscious evaluation. The person isn't refusing to see evidence of trustworthiness, they're genuinely not registering it as trustworthiness evidence.
This is also why positive thinking frameworks (such as "The Secret") are counterproductive for people actually embedded in difficult circumstances. The framing implies that the person's problem is insufficient belief rather than an actual pattern of accumulated bad experience. That framing is both epistemically wrong and morally offensive. It blames the victim, and it offers a mechanism that cannot work. You cannot override a body of real experience by consciously asserting the opposite conclusion. The data that built the worldview is still there. The filters trained on that data are still running. The assertion of positivity doesn't reach them.
---
If the dark-world orientation is built from a body of real experience, the only real remedy is a body of real experience that goes in the other direction, delivered in conditions where the threat-detection system can actually register the situation as not threatening.
Both conditions are necessary and both are difficult to achieve.
On the first condition: the countervailing evidence has to be proportionate in weight to the original evidence. Someone whose dark-world orientation was built over a decade of suffering needs something like a decade of genuine trustworthiness. A handful of good experiences could be reinterpreted as exceptions to the rule. It must be a sustained pattern that is thick enough and consistent enough to actually shift the underlying models. This takes time and sustained attention, like terraforming. You cannot plant a sequoia in dead sand. You have to build the preconditions for the preconditions.
On the second condition: trustworthiness has to be legible, not just present. A person running high-sensitivity threat detection will not automatically register a trustworthy environment as trustworthy. The trustworthy environment has to contradict the dark-world hypotheses in ways that are visible and resistant to reinterpretation. Merely not harming someone is not enough, because "not harming" can be filed as "hasn't harmed me yet." The evidence has to be sufficiently clear and sufficiently unusual relative to the person's expected baseline to actually penetrate the filtering system.
This means that the most dangerous thing is a partial or careless effort. An environment that presents itself as high-trust but then delivers untrustworthiness, or an environment that is genuinely trustworthy but whose trustworthiness isn't legible to someone with strong threat-detection, doesn't just fail to help. It actively makes things worse. It becomes more evidence for the hypothesis that high-trust environments are facades. It depletes a budget that was already limited. The person's filters tighten further, and subsequent genuine efforts face an even steeper climb.
---
The Asymmetry of Advice
One important corollary of the above: people from dark-world contexts who offer advice about social environments are systematically poorly positioned to advise about high-trust dynamics, because high-trust dynamics are outside the range of their experience and therefore outside the range of their models.
This isn't a character failing. It's an epistemic limitation produced by a history. Someone who has never experienced a social environment where people are genuinely non-manipulative and willing to extend trust without instrumentalizing it cannot be expected to have good models of what that environment requires or enables. Their advice, however sincere, will be calibrated for the wrong environment. It will import low-trust solutions into high-trust contexts, and those solutions will tend to corrode exactly what they're intended to protect.
This applies symmetrically: someone from a reliably high-trust environment who has never experienced systematic betrayal will give bad advice to someone who is genuinely embedded in a dangerous social context. "Just trust people" is exactly as bad as "trust no one" when applied to the wrong environment.
The practical upshot is that advice about social trust is only reliable from someone who can accurately identify which type of environment they're in and adjust accordingly. Most people cannot, because the identifying is itself done by the same filters that were shaped by their prior environment.
---
What Good Conditions for Recovery Look Like
Certain conditions consistently seem necessary, even if not sufficient, for someone to move from a dark-world orientation toward something more open.
The process requires consent. You cannot engineer someone's perceptual update against their will. Attempts to do so will be correctly identified as manipulation and will confirm the dark-world hypothesis. The person has to be, at minimum, willing to participate in whatever the process of building new experience looks like.
The environment has to be genuinely different from what the person is accustomed to, not merely asserted to be different. This is one reason why helping someone move from a dark-world orientation cannot happen in the same space where the dark-world behaviors are actively occurring. A rehabilitation context and an "everything is already great here" context require different conditions and can't fully share space.
The incremental steps have to be survivable. Someone with very few resources (emotional, social, material) cannot afford to take big risks. A failed trust experiment when you have no reserve is devastating in a way that a failed trust experiment when you have substantial reserve is not. You have to start with the smallest possible risk and build from there, in sequence, in the same way that terraforming a desert starts with the hardiest possible plants and works up through successive generations to more complex ones.
And the expectations of the person helping have to be realistic. The old patterns decay slowly and can spring back rapidly. A single bad experience can recontextualize a long run of good ones. Progress is not linear, and treating it as if it should be creates conditions for a particularly damaging failure.
---
The Value of High-Trust Environments
None of the above is to say that dark-world orientations are simply pathological. They are adaptive, calibrated responses to real environments. The person who grew up surrounded by unreliable people made a correct inference. The problem is not the inference. It is that the inference gets applied past the context that justified it, in part because the filter system doesn't automatically update, and in part because the selection and sorting dynamics mean that dark-world contexts are genuinely hard to escape.
What is worth naming explicitly is that high-trust environments enable things that are categorically unavailable in low-trust ones. The quality of interaction that's possible when people genuinely aren't tracking debts, aren't monitoring for manipulation, and aren't braced against each other - the depth of vulnerability, the quality of play, the willingness to be seen without editing - these are not small goods. They represent a substantial portion of what makes human social life worth having. People who've only ever experienced low-trust dynamics often can't believe they're real, because they've never encountered them and their models have no category for them. But they are real, and the difference between experiencing them and not experiencing them is significant.
The point is not to shame people for being in the wrong place on the distribution. The point is that the distribution is not fixed, that the mechanisms are understandable, and that moving from one attractor to the other, while genuinely hard, is possible, and worth taking seriously as a goal.
The explanation lies in a set of interlocking feedback loops.
Perceptual filtering. The brain cannot process everything it receives. It operates by building predictive models of the environment and then checking selectively for deviations from those models, rather than taking in data neutrally. This means that what you expect to see, you will tend to see. That's not because you're hallucinating, but because perception is a top-down process as much as a bottom-up one. A person who has learned to expect manipulation will genuinely notice potential manipulation in ambiguous signals, where someone else would not notice it at all. Attention is a finite resource allocated according to existing models of what matters. Whatever your brain has been trained to flag, you will find in abundance.
Behavioral shaping. People don't just filter information differently. They behave differently as a result of their filtered experience, and those behaviors elicit different responses. Someone who carries a default posture of wariness and defensiveness tends to generate wariness and defensiveness in others. Someone who radiates relaxed openness tends to elicit relaxed openness. This is not magical thinking; it's a straightforward consequence of the fact that social interaction is bidirectional. The signals you send shape what comes back. The person who enters every room braced for hostility will, with regularity, find hostility. That's partly because hostile people are attracted to easy targets and partly because their own bracing posture provokes friction.
Selection effects. Beyond direct behavioral feedback, there are structural sorting processes. High-trust social environments tend to eject people who repeatedly break the norms of trust — who fail to reciprocate, who treat gifts as commodities, who introduce low-trust behaviors into gift-economy relationships. This means people with dark-world orientations accumulate fewer and fewer access points to the very social environments that might offer them contradictory evidence. Meanwhile, when they do attempt to experiment with trust, they tend to do so in environments where the other participants are also relatively inexperienced with genuine trust, or where actual predators have positioned themselves after being ejected from high-trust spaces. The experiment fails. The worldview hardens.
Cluster formation. Over time, these sorting processes create geographically overlapping but experientially non-overlapping social worlds. Two people on the same street can have radically different networks, exposures, and baseline experiences, and each will have abundant evidence for their own worldview precisely because their evidence is drawn from a non-random sample of social reality.
The net result is a self-reinforcing system. Entry into either attractor makes exit progressively harder. The feedback loops tighten. The perceptual filters strengthen. The behavioral habits calcify.
---
The Prion Problem
A useful concept here is the prion. A prion is a misfolded protein that propagates by causing neighboring correctly-folded proteins to adopt the same malformation. Dark-world orientations behave similarly in social contexts.
Someone operating from a low-trust frame doesn't just protect themselves, they alter the social environment around them. When one member of a group starts tracking debts precisely, insisting on exact repayment, treating favors as transactions rather than bonds, the gift economy logic that held the group together begins to corrode. Other members start reciprocating in kind, not because they wanted to, but because continuing to operate in gift-economy mode in the presence of someone running market logic feels asymmetric and exploitable. The low-trust behavior propagates.
The same dynamic holds for anxiety and paternalism. A highly anxious person in a group doesn't just experience their own anxiety. Their visible distress activates threat-detection systems in others. A parent radiating fear about a child's safety creates a social pressure toward risk aversion that affects even parents who were themselves comfortable with the risk. You can feel the anxiety of others as information about whether danger is present, even when it isn't.
This is why even well-intentioned people from dark-world contexts can be destructive to high-trust social environments. The issue isn't malice. The issue is that the behaviors adaptive for one environment are disruptive in the other, and they propagate.
---
Why Dark-World Views Are Hard to Exit
The dark-world orientation is not just a persistent personality trait. It's something closer to a self-sealing epistemic system.
Consider what it would take to convince someone with a strong prior on manipulation that a given person's friendly behavior is genuine. Every piece of friendly evidence can be reinterpreted as a tactic. The friendlier and more consistent the behavior, the more sophisticated the manipulation presumably is. There's no observation that definitively disconfirms the hypothesis, because the hypothesis has built-in explanations for all apparent counterevidence. The belief system is so stable precisely because of unfalsifiability, similar to a theological worldview that attributes disconfirming events to divine mystery or satanic deception.
This is why well-meaning exhortation fails. Telling someone in the dark world to "just trust people more" is like telling someone to update a Bayesian prior by willpower. It isn't how it works. The prior is maintained not by a conscious choice but by a filter that processes incoming data before it ever reaches conscious evaluation. The person isn't refusing to see evidence of trustworthiness, they're genuinely not registering it as trustworthiness evidence.
This is also why positive thinking frameworks (such as "The Secret") are counterproductive for people actually embedded in difficult circumstances. The framing implies that the person's problem is insufficient belief rather than an actual pattern of accumulated bad experience. That framing is both epistemically wrong and morally offensive. It blames the victim, and it offers a mechanism that cannot work. You cannot override a body of real experience by consciously asserting the opposite conclusion. The data that built the worldview is still there. The filters trained on that data are still running. The assertion of positivity doesn't reach them.
---
What Actually Works, and Why It's Hard
If the dark-world orientation is built from a body of real experience, the only real remedy is a body of real experience that goes in the other direction, delivered in conditions where the threat-detection system can actually register the situation as not threatening.
Both conditions are necessary and both are difficult to achieve.
On the first condition: the countervailing evidence has to be proportionate in weight to the original evidence. Someone whose dark-world orientation was built over a decade of suffering needs something like a decade of genuine trustworthiness. A handful of good experiences could be reinterpreted as exceptions to the rule. It must be a sustained pattern that is thick enough and consistent enough to actually shift the underlying models. This takes time and sustained attention, like terraforming. You cannot plant a sequoia in dead sand. You have to build the preconditions for the preconditions.
On the second condition: trustworthiness has to be legible, not just present. A person running high-sensitivity threat detection will not automatically register a trustworthy environment as trustworthy. The trustworthy environment has to contradict the dark-world hypotheses in ways that are visible and resistant to reinterpretation. Merely not harming someone is not enough, because "not harming" can be filed as "hasn't harmed me yet." The evidence has to be sufficiently clear and sufficiently unusual relative to the person's expected baseline to actually penetrate the filtering system.
This means that the most dangerous thing is a partial or careless effort. An environment that presents itself as high-trust but then delivers untrustworthiness, or an environment that is genuinely trustworthy but whose trustworthiness isn't legible to someone with strong threat-detection, doesn't just fail to help. It actively makes things worse. It becomes more evidence for the hypothesis that high-trust environments are facades. It depletes a budget that was already limited. The person's filters tighten further, and subsequent genuine efforts face an even steeper climb.
---
The Asymmetry of Advice
One important corollary of the above: people from dark-world contexts who offer advice about social environments are systematically poorly positioned to advise about high-trust dynamics, because high-trust dynamics are outside the range of their experience and therefore outside the range of their models.
This isn't a character failing. It's an epistemic limitation produced by a history. Someone who has never experienced a social environment where people are genuinely non-manipulative and willing to extend trust without instrumentalizing it cannot be expected to have good models of what that environment requires or enables. Their advice, however sincere, will be calibrated for the wrong environment. It will import low-trust solutions into high-trust contexts, and those solutions will tend to corrode exactly what they're intended to protect.
This applies symmetrically: someone from a reliably high-trust environment who has never experienced systematic betrayal will give bad advice to someone who is genuinely embedded in a dangerous social context. "Just trust people" is exactly as bad as "trust no one" when applied to the wrong environment.
The practical upshot is that advice about social trust is only reliable from someone who can accurately identify which type of environment they're in and adjust accordingly. Most people cannot, because the identifying is itself done by the same filters that were shaped by their prior environment.
---
What Good Conditions for Recovery Look Like
Certain conditions consistently seem necessary, even if not sufficient, for someone to move from a dark-world orientation toward something more open.
The process requires consent. You cannot engineer someone's perceptual update against their will. Attempts to do so will be correctly identified as manipulation and will confirm the dark-world hypothesis. The person has to be, at minimum, willing to participate in whatever the process of building new experience looks like.
The environment has to be genuinely different from what the person is accustomed to, not merely asserted to be different. This is one reason why helping someone move from a dark-world orientation cannot happen in the same space where the dark-world behaviors are actively occurring. A rehabilitation context and an "everything is already great here" context require different conditions and can't fully share space.
The incremental steps have to be survivable. Someone with very few resources (emotional, social, material) cannot afford to take big risks. A failed trust experiment when you have no reserve is devastating in a way that a failed trust experiment when you have substantial reserve is not. You have to start with the smallest possible risk and build from there, in sequence, in the same way that terraforming a desert starts with the hardiest possible plants and works up through successive generations to more complex ones.
And the expectations of the person helping have to be realistic. The old patterns decay slowly and can spring back rapidly. A single bad experience can recontextualize a long run of good ones. Progress is not linear, and treating it as if it should be creates conditions for a particularly damaging failure.
---
The Value of High-Trust Environments
None of the above is to say that dark-world orientations are simply pathological. They are adaptive, calibrated responses to real environments. The person who grew up surrounded by unreliable people made a correct inference. The problem is not the inference. It is that the inference gets applied past the context that justified it, in part because the filter system doesn't automatically update, and in part because the selection and sorting dynamics mean that dark-world contexts are genuinely hard to escape.
What is worth naming explicitly is that high-trust environments enable things that are categorically unavailable in low-trust ones. The quality of interaction that's possible when people genuinely aren't tracking debts, aren't monitoring for manipulation, and aren't braced against each other - the depth of vulnerability, the quality of play, the willingness to be seen without editing - these are not small goods. They represent a substantial portion of what makes human social life worth having. People who've only ever experienced low-trust dynamics often can't believe they're real, because they've never encountered them and their models have no category for them. But they are real, and the difference between experiencing them and not experiencing them is significant.
The point is not to shame people for being in the wrong place on the distribution. The point is that the distribution is not fixed, that the mechanisms are understandable, and that moving from one attractor to the other, while genuinely hard, is possible, and worth taking seriously as a goal.
