Previously.

Main takeaways:

• There are many valid ways to construe (“frame”) reality.
• Within the lossy compression that is a problem-construal, certain solutions are made more or less available.
• By flexibly switching between frames, and being able to toss multiple construals at a problem, you’re more likely to discover a workable solution.

---

Negotiating with those who share your frame is straightforward. The potential for agreement—for reaching a new coordinative equilibrium—is built into the shared frame and in some sense pre-arranged. To agree to a frame, or definition of a situation, is to have already preemptively struck a deal. It is when frames are contested, and differ, that we are left struggling towards a solution—arguing over categories and conceptualizations in order to reach a consensus ought.

Play around with AI Carpet Salesman for a few minutes, then come back. The game’s object is to haggle down the cost of a carpet. Spoilers: If you accept the salesman’s initial frame—that the carpet is valuable; that you desire the carpet; that you are an ordinary customer—then at best you’ll knock a few hundred bucks off sticker price. But by contesting his frame—for instance, by feigning immense local status and influence—you can convince him of almost anything. That it would be a great honor for his rugs to be shown in your illustrious palace. That such an act would be rewarded in the afterlife—and boost sales. That your well-connected, wealthy friends will be sure to spread word of the carpets he has furnished you with, raising his shop’s prestige. If you do this, the salesman will end up paying you for the privilege of giving his entire stock away gratis.

(“Palace,” “afterlife”—note the associative web that has been invoked merely by the words “carpet salesman.” Parts suggest wholes; wholes suggest parts—so do metonyms suggests frames, and frames metonyms.)

This haggling gambit works because a language model is a “frame bottom,” passively accepting the definition of situation—the organization of facts, and the structuring of salience—offered by a “compressive top” (that’s you, monsieur prompter). A language model has no ability to ground against reality, no access to territory, and therefore no ability to contest your frame. In other words, because a linguistic framework is all an LLM has access to—living as it does inside map instead of territory—it can be easily bullied into doing almost anything you want. Its sense of what should follow from a premise is inflexible. But that just means manipulation is bottlenecked on convincing it of premises—which is how animal communication works too, anyways. How do you get played? By pursuing your own interest on shaky priors.

So what is a frame, anyway?

The Nature of Frames

The downtrodden are the losers to the local victor. To the peacemaker, however, the downtrodden are the evidence that the local victory is too local. The peacemaker must therefore attend to the downtrodden, whose suffering is the gateway to the transcendent.

Peterson, “Peacemaking Among Higher-Order Primates”

Many points have been made in Part One which the reader finds less than persuasive. That Peterson—advocate and emblem of Western Christian patriarchy—could endorse a view of truth so smacking of PoMo relativism appears dubious. That the “facts” could “differ,” that frameworks are pragmatic and value-laden rather than the neutral province of science—this perhaps, seems outrageous. 

If you, reader, are a partisan or crusader, there is little I can say here to persuade you to set down arms. That would be a radical proposition, which likely goes against everything you believe: that your perspective is correct in a way that subsumes all rival interpretations; that it is threatened with extinction and therefore must be aggressively defended; that the best route to its long-term security involves the complete eradication of contradictory claims. Perhaps I will try make such a case in writing one day. For now, I can only point out the long and frustrated history of geopolitical conflict, in sharp contrast with the endosymbiosis of microbes; that coordination equilibria are always more efficient than conflict, if you can strike the bargain. Can only recommend David Chapman’s work on meta-systematicity, or John Vervaeke on relevance realization. Peterson’s 2006 “Peacemaking in Higher Primates” view of the world—and of the rational systems we use to navigate it—is kindred with both these thinkers’ perspectives. See e.g. Vervaeke contrasting philia sophia—love of knowledge—and philia nikia—love of victory—which I see as roughly analogous to Peterson’s global versus local victory concept.

If, on the other hand, there is something resonant about this idea of peace-making as a reconciling and translating of worldviews—about the idea that parties in a dispute, wanting and valuing different things, will produce a set of “facts” which echo and support that desiring and valuing—then perhaps this is essay is for you.

The similarity between Chapman, Vervaeke, and Peterson’s work is less convergent coincidence than divergent shared descent: all take their cue from what is usually called the frame problem—first theorized in artificial intelligence research but quickly adapted for cogsci. (This connection was, to my knowledge, first noted by Jake Orthwein, and since been endorsed by Chapman himself.) I believe it is also possible reach similar world-stances via: American pragmatism, Wittgenstein’s language games, the social phenomenology of Alfred Schütz, the social reality of Berger and Luckmann, Thomas Schelling’s concept of focal points (and extended by David Lewis), Bourdieu’s “habitus” and “space of possibles,” and perhaps most readily, Erving Goffman’s “definition of a situation.” All argue some part of the meta-systematic view, and Goffman’s early work on social definitions (found in Presentation of Self in Everyday Life) morphs, by the early 1970s, into Frame Analysis—a book that proved crucial in popularizing the conceptual metaphor of the frame across media studies, political theory, behavioral economics, and sociology.

These frames (stances, orientations, definitions of reality) shape not just how we argue and vote, but how we feel, think, and perceive. This is a rather intuitive claim, well-enshrined in the halls of folk knowledge. Our lens, our set, our setting; our prior information, our context, our mood, our personality, our agenda and interests—all are known to shape our interpretation of “is.” To Vervaeke, they are problem construals—ways of organizing and attending to reality which make certain aspects salient while (necessarily, so as to tame the combinatorial explosion before us) obscuring and ignoring many other aspects. This organizing and construing of problems constitutes the real computational lift of intelligence; once a problem is construed—once the messy nebulosity of the real world has been approximated in a formal logic—the procedures for solving it using that logic are comparably straightforward. Different construals—both of the problem space, and of one’s own desires and goals—suggest different optimal responses.

But where the average man believes there really is or must be some neutral fact of the matter, beyond such framed interpretations, those of a more pragmatic inclination balk. From the perspective of the pragmatist, a “fact” is a word designating merely that arrangement of sounds which prompts an agent—acculturated into interpretive habits—to productively and reliably act upon it. All ordinary language is hopelessly compressive, simplifying and generalizing in short-hand over infinite complexity. That is, the closest thing to a scientific “fact” to describe a social situation is an extraordinarily complex set of interactions between trillions of particles, each with a complex state and history. (Peterson, in “Three Forms of Meaning & the Management of Complexity” quotes Diogenes Laertius approvingly: “Nothing exists except atoms and empty space; everything else is opinion.”) We might spend a lifetime mapping and listing and modeling each of these particles, with the best supercomputers available, and still our predictive power would be much less than that of an average, socially competent observer who coarse-grains those “facts” as “This man wants something from that woman.” By extension, it should be clear that, long before language arises, our perception performs similar coarse-graining (compression, approximation). And lossy compression, as any information theorist or computer engineer can tell you, is by its nature value-laden. Not all information can be preserved; the editor who hopes to abridge Moby Dick must make decisions which are subjective and theological, in the sense of reflecting his beliefs about art and Melville and the world at large which he could never “objectively” justify. What is left in is precisely that which is deemed to be “important” or “relevant”—judgments which ground as much against a subject’s structure of desire as it does the objective state of the world. All frameworks, and all parts of a framework, thereby encode pragmatic entailments as much as they do patterns in the universe. From this perspective, true sentences are true (and facts are facts) in light of their heuristic value when acted upon in a culturally learned way.

The American pragmatists, trained in evolutionary thought, derived from this the same conclusions which evolutionary biologists and ethologists would reach half a century later: an organism’s sense of what descriptively exists in the world is, and must be, downstream of its position and goals in that world. If it did not organize reality according to its aims, in a way that optimized for survival and reproduction, then it would be outcompeted by those organisms that did organize reality in such a way. Vervaeke’s emphasis on instrumental relevance, Friston’s work on free energy minimization, J.J. Gibson’s emphasis on environmental affordances, Chabris & Simons’ selective attention test, the entire field of behavioral economics, and finally, Jordan Peterson’s M.A.P.S. (Meaning, Action, and Perception Schemas) all gesture in this same direction. (Objects, Peterson writes, “are entities bounded by their affective relationship to a goal. We perceive meaningful phenomena, not the objective world.” And there is no “meaning” without a framework of caring and preferring.)

The set of primary frameworks in a given society constitute (to Goffman) its cosmology. (This word will come back when we visit Barthes’ Mythologies in future installments.) A shared cosmology underlies our ability to perform complex coordination, especially en masse; it is the synchronization and complementarity of frameworks which underlies the complementarity of strategies, or moves, and which in turn is necessary to reliably produce cooperative (i.e. mutually fitted, mutually beneficial) outcomes. A common framework which most people are familiar with, and have been initiated into, is their culture’s framework for driving an automobile. It is clear to most people how a shared definition of various road and traffic signals—what they represent or more precisely and pragmatically, prescribe—enables cooperative, complementary, mutually beneficial behavior between drivers. At a given intersection, there is an efficient complementarity between my seeing a stop sign and slowing my car, while at the same time another driver, lacking a stop sign, continues at pace. In Peterson’s language:

Because you exist in a cooperative and competitive landscape, the perceptual structures and plans that you lay out, we’ll say the maps you lay out, have to be negotiated with other people. And so that puts stringent constraints on the number of interpretations you’re allowed to apply. You can think about this in a Piagetian sense; if there are children on a playground and they’re trying to organize themselves to play, they have to agree on a game. And the game is of course a perceptual structure and a goal-directed structure and a structure that delimits actions and interactions.

2017 “Maps of Meaning” lecture

A game frame, suggested by formal rules or informal precedent, provides a set of expectations as to the payoffs available, and what moves might unlock them, which may but does not necessarily include expectations as to how other players will expect and behave. That a move is legal or illegal, in-bounds or out of bounds; whether it will help or hurt in a player’s securing of his goals; all are part of a game’s frame. If players did not share a cultural framework for understanding payoffs, and deriving complementary stimulus-response inclinations from them, the game would fall apart.

The Style of No Style: Why We Need To Reconcile Frames

Among the most provocative analogies for the framework (or “system of facticity”) are style and mood. Not so far off, really. A world-interpretation is a model of, and orientation to, the world, in the same sense that a good regulator is a hybrid model-orientation of and to its environment. There is no one true regulator in the same way that there is no one true framework and no one true style; all organisms which have persisted over time are equally evolved. All behavior is non-PIG-optimal. The organism reads its environment so as to better “write” to it—act within and optimize around it, maintaining literal and figurative homeostasis. A shifting array of moods and emotions bring styles of being and interpreting along with them; certain frames and styles are more or less congenial with a given mood. Who among us has not noticed their entire sense of reality morph from surging or depleted serotonin levels? Do narcotics not drag ontologies with them?

(See also mood affiliation discourse in the GMU-sphere; this works, from a collective computation perspective, because moods are normative, loosely action-oriented; they imply prescriptions.)

That styles—like moods or emotions, like organisms or heuristics—are not “true” or “false” but more or less conducive to various aims, given various contextual situations or “games,” is I think self-evident. I would like us to move toward viewing frameworks this way as well, so that the value of generating and negotiating between a diverse array of conceptual frameworks appears equally self-evident. (And by extension, the value of maintaining frame diversity in the face of pressures towards monoculture.)

Sarah Perry, an old friend of Chapman’s who has herself written about meta-systematicity, discusses Bruce Lee’s “style of no style” in the brief Ribbonfarm post “Systems of the World.” What was Lee getting at with his jeet kune do, his “style of no style”? What might such an orientation, such a “meta-style,” entail? Lee:

True observation begins when one sheds set patterns, and true freedom of expression occurs when one is beyond systems… I hope to free my comrades from bondage to styles, patterns and doctrines.

This reminds Perry of meta-systematicity and its principles of heterogeneity, flexibility, and indexicality. In contrast with the generalizing dogmatism of deontology, which claims one ruleset (and by extension, one ontology) as binding, the meta-systematic mode stresses context-sensitivity and prioritizes adaptive, generative processes over any single generated state. It understands that letter constantly betrays spirit, that a lobster is “perverse” when removed from the sea (Emerson¹); that every heuristic is fitted to a distribution context and every solution to a problem at hand. (Can we see how incoherent it is to speak of an orientation as being “correct” in some timeless, context-free way, just as it is incoherent to speak of an organism being evolutionarily “fit” in a timeless, context-free way, or one solution answering all problems? We might reference the No Free Lunch Theorem, but we need not; the dynamic is intuitive.) 

[^1:] From Emerson’s journals:

“Every thing is a monster till we know what it is for… A lobster is monstrous but when we have been shown the reason of the case & the color & the tentacula & the proportion of the claws & seen that he has not a scale nor a bristle nor any quality but fits to some habit & condition of the creature he then seems as perfect & suitable to his sea house as a glove to a hand. A man in the rocks under the sea would be a monster but a lobster is a most handy & happy fellow there.”

“The systems and patterns that can oppress us,” Sarah Perry writes, “are also extremely useful. Growing beyond them does not mean throwing them out.” All power is limited; we may focus either on the empowerment or the limitation. But once we have given up the search for a single, ultimate system, we free ourselves up to accrue an “ecology of practices” (Vervaeke)—a vast and varied arsenal; a tool belt of curated duties. 

What is needed, then, is a system for switching between systems, for picking up that framework which bests suits one’s contest or opponents—an ability which, to many practitioners, manifests as pure feel for the game. As the game proceeds, new frameworks will need to be discovered and developed, because the meta is always evolving.

Chapman, with his “post-systematic system of thinking,” analogizes this meta-systematic mode to “piloting nimble watercraft on a sea of meaning.” Chapman is a Melvillean, as Hubert Dreyfus’s marathon lecture series on Moby Dick makes clear. Perry, riffing on Dreyfus:

Why is the whale white? He is white, and very much blank, because he personifies (Dreyfus offers “mammal” for “person”) the indefinable, “the universe” rather than the world of people, that which refuses to be nailed down to a particular meaning in the eternalist/ontotheological worldview. Each character has a different response to the unknowable white whale, who cannot be seen beneath the dark water, and even if fished out to the air and light, cannot truly be seen, because his form becomes distorted out of the water… Ahab in particular, the monomaniac eternalist, demands the impossible from the whale: that it reveal itself to him, its one final meaning, whether the universe cares about him or not, whether he is important or not.

“Light of the American Whale”

In canonizations of “the great 19th century novel,” Moby Dick’s chief rival is Eliot’s Middlemarch. That text, too, features an Ahab-like character, Casaubon, whose lifelong quest for a “key to all mythologies”—a single schema uniting all world literature and religion—proves equally futile and fatal. (Elsewhere, in an equally Melvillean spirit, Eliot refers to Casaubon’s project as “the end of all journeys.”) That 19th century novelists appear so biased in favor of “polytheism” is probably not a coincidence; the novel, as Bakhtin tells us, is a technology for situating different perspectives and understandings in relation to one another.

Dreyfus explicitly draws an analogy between moods and ways of seeing, taking up Emerson’s metaphor that emotions are “colored beads” through which we view the world. Not only do colors, encountered in the environment, alter our moods, and signal macro-information about some normative stance—warning red, caution yellow, sickly or exuberant green—but they also color us in turn. Not just our words and perceptions, figuratively speaking, but more literally the blush in our cheeks, the white-sheet pallor of fear. Our souls may be black, our spirits blue; we may even turn green from envy. With this, our account comes full circle:

The way to experience human reality—worlds—according to Dreyfus, is to be in a mood. We are always in moods, falling into and out of them, sometimes stuck in one. Moods are not merely feelings (affect or emotion): a mood reveals a world to us.

Ibid.

Dreyfus calls this Melville’s “polytheism,” but we could equally describe it as an ecology of practice. Moods are associated with gods—a move that echoes nothing if not a Jaynesian view on bronze age consciousness—and the white of the white whale is the white light of every color blended into one Being:

When all the colors of light are summed up, stacked on top of each other as it were, ‘all at once,’ they produce white—chaos, or at least inhuman blankness… when placed in the order of a spectrum, refracted through a prism, they produce a rainbow… 

Ibid.

Some stances—be they moods or ontologies, religious or rational—are better suited to some purposes and worlds; by extension, they must be less well suited to others. This quality of fit defines the model (or the map, or the representation) because it must, by definition, leave out certain complexities of reality while emphasizing others. We understand, intuitively, that certain map styles are better for hiking (e.g. a topographic map) and others (e.g. a street map) better for navigating cities. Real freedom involves the ability to get into the mood (model, system, stance) appropriate to a given circumstance.

To become meta-systematic is to acquire a library of maps, rather than mistake any single map as interchangeable with territory; and to choose that map which best fits the context and project at hand. This is an approach, Lee writes, “where systems can be held lightly as tools, not clung to as totems.” (Compare the Bourdieusean insight that we constantly naturalize our perceptual schemas and surrogates.) To choose the proper tool, the proper heuristic, the proper surrogate—even the right “rationality”—rather than reify any as timelessly “true.” (Vervaeke calls meta-systematic competence “wisdom”—rationality recursively applied to the problem of using rationality effectively.) Tools, maps, models, organisms—all these share this quality of fittedness; no one knows what Ahab’s whale “really” or “fully” looks like, because it is always at least partially submerged in the dark, murky waters of sea; and even if it were to wash up ashore, it would have been perverted and distorted by its removal from its natural environment, the context to which it, as an organism, is situated. It would be a “fish out of water.”

(The slipperiness of linguistic definition and category—which emerges downstream of the frame problem—comes up in Melville too, the same parable of the whale-as-fish that Scott Alexander seizes upon in the late 2010s. Chalmers goes on to cite the passage in his “Verbal Disputes” paper, which more or less reinvents classic LessWrong—read, General Semantic—ideas about language. Moby Dick: “I take the good old fashioned ground that the whale is a fish, and call upon holy Jonah to back me. This fundamental thing settled, the next point is, in what internal respect does the whale differ from other fish. Above, Linnaeus has given you those items. But in brief they are these: lungs and warm blood; whereas all other fish are lungless and cold blooded.”)

And when we come to situations which involve both city streets and rocky hills, calling for both elevations and avenues, we must together build a map that reconciles both views. This reconciliation, just like the maps themselves, is value-laden; there is no “objective” or general solution. Rather, it requires a process of collectively discovering what participants in the negotiation value, and how the new synthetic model might accommodate the whole in a Pareto-optimal, mutually beneficial way. The final map will be as much a product of bargaining and power structures as will be of the territory. To ensure a productive, lasting collective symbiosis—rather than a wasteful conflict which ends with temporary domination by the powerful—there is need for a peacemaker.

---

Next time, in part 3, “Water Metaphors”:

Over and over again this essay has resorted to liquid metaphors—often contrasting the structural stability of the framed perceptual world with the ontological fluidity or nebulosity of reality. Water metaphors are central as well taoism, which has been called a philosophy of flow (much as Tai Chi has been called a practice of flow). Perhaps the first thinker to recognize the frame problem was Heraclitus, who called man the measure of all things—man, who discretizes continuous space—and who wrote of the river that no man can step in the same stream twice, for the contents of its flow are always different. (Nor is the man who steps twice the same man, for the same reason.)