AI, Intuition, and Rational Divination

By Cole Whetstone

1. SIGNS AND PORTENTS:

Modern scientific thought often emphasizes a sharp distinction between correlation and causation. “Correlation,” it is claimed, “does not equal causation.” And humorous examples are often adduced to demonstrate the point.¹ This (correct) doctrine has led, I will argue, to a widespread dismissal of certain correlative methods as useless superstition, even when they manifestly are not.

Exemplifying this view is the philosopher Thomas Hobbes, who, in Leviathan, broadly rejects divinatory practices — astrology, augury (interpreting bird behavior), aruspicina (examining animal entrails), dream interpretation, and metoposcopy (facial or palm reading) — as irrational and deceptive. These methods, he argues, rely merely on conjectures drawn from past experiences and human fears rather than rigorous causal reasoning, and so amount to nothing more than “juggling and confederate knavery.”

Here I will take partial exception to this standard view of divination – though not as an uncritical vindication of astrology or divination as they are commonly practiced today. On the contrary, most modern iterations of these traditions lack the rigor and statistical coherence that would make them scientifically useful. But this does not invalidate the deeper insight that indirect signs — when properly aggregated and interpreted — can reveal truths otherwise inaccessible. Rather than abandoning these ancient arts to the dustbin of superstition, we propose a reformation: a new science of signs, grounded in statistical modeling, interdisciplinary insight, and epistemic humility.

It will be the task of this essay to make this clear and to suggest how some of these practices might be more rigorously modeled and empirically tested

2. SIGNS AND PORTENTS:

Hobbes’ view reflects an early modern scientific skepticism still prevalent today, which tends to dismiss the above-mentioned associative practices outright, as invalid for the purposes of scientific prediction. To their credit, it is hard to see how practices such as random verse-dipping or tarot cards — methods Hobbes refers to as “mere lottery”— could ever be predictive, as they appear to lack plausible mechanisms for it, and thus almost certainly fall into the category of superstition. That said, other divinatory methods like astrology appear to have clearer, potentially meaningful associative basis, which the current view on correlative inference tends to ignore.

The core philosophical insight here is well-illustrated by a quote of Shakespeare’s found in Julius Caesar, placed in the mouth of the Roman Senator Casca. Confronted with a series of strange and foreboding occurrences, Casca urges his fellow Romans not to explain them away as mere natural phenomena:

“When these prodigies do so conjointly meet, let not men say ‘These are their reasons: they are natural’; for, I believe, they are portentous things unto the climate that they point upon.” (Shakespeare, JC, 1.3.28-32)

This passage, I believe, captures a profound truth about the philosophy of signs. Its importance lies not necessarily in direct causality but in the pragmatic sufficiency of correlated signs towards a broader “climatic” shift that might otherwise be imperceptible. Shakespeare especially highlights how multiple seemingly unrelated signs, when occurring conjointly, often reliably indicate broader underlying shifts in “climate” — even if their causal relationships remain unknown.² This introduces a broader philosophical framework for understanding signs.

Philosophers such as Charles Sanders Peirce and Augustine of Hippo have proposed influential taxonomies of signs, which help illuminate this framework. For the purposes of this essay, we will adopt a working distinction among three types:

• Natural signs are those in which a direct, necessary, causal relationship exists between the sign and the thing signified. These signs are not arbitrary but are grounded in a physical, or biological process. Smoke is a natural sign of fire; fever is a natural sign of infection. These signs depend on consistent physical or biological laws.

• Conventional signs are those whose meaning is not determined entirely by nature but at least in part by social agreement or learned association. Such signs are (at least in some sense) arbitrary. For instance, the word tree signifies a particular type of plant only because speakers of English have collectively agreed upon this pairing; the precise sound of the word “tree” would not necessarily be paired with its significance in English by a speaker of another language. Words, flags, and traffic lights fall into this category; their significance depends on convention and shared agreement rather than intrinsic connection.

• Statistical or indirect signs refer to patterns of correlation that, while lacking a known or necessary causal mechanism, nevertheless reliably track or point toward certain phenomena through their aggregate regularity. They emerge from patterns of correlation rather than causal or conventional symbolic relationships. These signs are neither grounded in natural necessity nor in social agreement, but in emergent regularities that allow for prediction and interpretation, even if the causal substrate remains obscure. It will be these signs which centrally concern us here.

In the view of Pierce and Augustine, with respect to what we now call statistical signs it remains logically true that correlation does not directly imply causation. Establishing causation demands rigorous isolation, testing, and metaphysical assumptions about nature that mere correlation lacks. Nevertheless, correlations — when strong, repeated, and multifaceted — can yield statistically significant predictive power. And this predictive power is often what makes causal knowledge of interest in the first place: it enables effective action and adaptive response in a world governed by complexity and uncertainty.

Hence, although the principle “correlation does not equal causation” is theoretically valid, in practical scenarios focused primarily on accurate prediction, correlation should not be casually dismissed. Indeed, as will be expounded in more detail later, this form of associative reasoning — without explicit causal explanation — closely mirrors the logic employed by Large Language Models (LLMs). An LLM does not explicitly understand grammar rules, yet it can accurately predict grammatical structures and generate coherent text through statistical pattern recognition across vast corpora of language. The astonishing success of these models at predicting the most appropriate or meaningful text-based response — given a linguistic or visual prompt — demonstrates that predictive accuracy does not always require causal comprehension; pragmatic reliability alone can suffice.

Nowhere is the interpretive potential — and the controversy — of statistical signs more enduring than in the case of astrology.

3. ASTROLOGY AS A MODEL CASE:

Astrology offers a particularly rich example of such associative reasoning. While birth under a particular constellation may not imply a direct causal influence from celestial bodies, the positions of stars and planets can serve as symbolic proxies — compressed indicators of environmental, geographical, temporal, and social variables known to shape human development. For example, scientific studies have found robust correlations between season of birth and various psychological and physiological outcomes, including increased incidence of schizophrenia, depression, and vitamin D deficiency (Torrey et al., 1997; Disanto et al., 2012). Birth order has similarly been linked to differences in personality, academic achievement, and long-term health (Sulloway, 1996; Barclay, 2015). Geographic location further influences health through factors such as diet, pathogen exposure, and environmental toxins (Diamond, 1997). In this light, astrological signs may be understood as historically evolved attempts to aggregate and interpret such distributed variables into a symbolic framework.³

All of these features are elegantly consolidated by the position of stars in the sky on a given night. The stars, then, are not direct causes of human destiny but function as indirect symbolic summations of complex information — geography, climate, season, birth order, and more — which can be used to accurately predict a variety of health, relationship, and life outcomes. In a nutshell, astrology can be a provably valid tool for prediction.

Does this imply we must adopt a metaphysical stance suggesting that stars directly influence human destinies or health outcomes through radiation, gravitational force, or other mechanisms? Not necessarily. From a scientific standpoint, the stars likely have negligible direct influence on human affairs. Yet rejecting the causal force of the stars does not entail rejecting their predictive utility.

In the language of Pierce and Augustine, the positions of the stars are best understood as “statistical but not natural” signs — they do not exert causal influence themselves, but can yield statistically robust predictions, even if they are causally inert themselves.

In fact – and this might seem surprising at first – this symbolic consolidation may enhance predictive accuracy beyond what any single direct influence could achieve on its own. Another way of thinking of this is that the stars function as compact proxies for a constellation of variables, and their significance lies not in what they do, but in what their patterned appearance summarizes — a distributed structure of meaningful associations that, taken together, point toward future outcomes with surprisingly valid precision.

Stated more formally, the predictive power of a perfect aggregate of many small direct signs will be of necessity greater than any of the aggregated signs considered individually, even if the aggregate sign is indirect. For example, a star position is (at least theoretically) a more predictive sign of health than (e.g.) geography, climate, or season alone, because the significance of the star does little more than comprise the predictive information of all of the smaller direct signs it contains. In this way, we can see that a well-constructed indirect signal, built from the collective weight of numerous smaller, direct indicators, can surpass the predictive capacity of any single one of those direct indicators.

This is a curious fact that deserves more attention, as it reveals a world of “emergent information,” uniquely accessible by what we have been calling divination -- “emergent” because the way these smaller direct signs combine and interact within the aggregate can reveal patterns or insights that wouldn't be apparent by looking at each sign in isolation. The "whole" of the indirect aggregate therefore can become greater than the sum of its direct parts in terms of predictive information. This is because the aggregate indirect sign could justify a prediction which would be insufficiently justified given any of the component direct signs. In such a case, valid predictions would genuinely “emerge” from mere symbolic consolidation.⁴

4. GENERALIZATION OF THE THEORY TO THE REST OF DIVINATION:

Some forms of augury, too, function as a convergent system of indirect signs.

In 2004, as the Indian Ocean tsunami approached, animals began exhibiting anomalous behaviors: elephants broke free and fled to higher ground, flamingos abandoned low-lying areas, dogs refused beach walks, and zoo animals sought shelter (Wadley and Borger 2005). These responses, occurring minutes before the disaster, suggest the animals detected environmental cues — seismic vibrations, pressure shifts, or infrasound — far beyond human sensory capacity.

Such signs, in isolation, might seem insignificant. But when observed together, they form a constellation of meaning. If a trained observer had noticed these correlated reactions, they might have predicted the coming tsunami — without understanding its precise cause. The animals, in effect, served as environmental sensors. By watching them, we access a layer of information otherwise unavailable to our senses. This, I argue, is the core logic of augury: not magic, but a method of mediated perception — where animals are best understood as distributed environmental signals, directly perceiving environmental causes beyond human sensory range, which humans infer indirectly by observing patterns and conjoint aberrations in the animals’ behavior.

The strange flight of a single bird may be dismissed as a mere aberration. But when dogs howl, elephants break loose, and flamingos fly for cover -- “When these prodigies do so conjointly meet, let not men say, ‘These are their reasons: they are natural. For, I believe, they are portentous things unto the climate that they point upon.” These are not mere omens in the superstitious sense, but aggregated anomalies, patterns that resist dismissal when read together.

In such cases, we would be unwise to dismiss these conjoint meeting prodigies as insignificant simply because we cannot articulate the underlying mechanisms. Rather, their meaning may not be in what they are, but in what they aggregate — the pattern they reveal when taken together. Augury thus becomes a kind of properly attuned ecological attentiveness, tuned to indirect signs whose convergence can reliably predict shifts in the broader system.

In like manner, practices like aruspicina (the reading of animal entrails), metoposcopy (the reading of facial lines), palmistry (the reading of lines and features in the hand), and dream-interpretation may (sic) also function as systems of indirect signs, rather than mere superstition.

In ancient contexts, anomalies in animal organs — such as discoloration, deformities, or asymmetries — were interpreted as omens. Today, we might understand such irregularities as potential indicators of environmental toxins, nutritional deficiencies, or disease exposure within a given ecological niche. For example, liver abnormalities in animals have been used as bioindicators of heavy metal contamination (Beyer et al., 1996), and amphibians with limb deformities have signaled the presence of endocrine-disrupting pollutants in freshwater ecosystems (Blaustein & Johnson, 2003). In this sense, ancient aruspexes may have, wittingly or unwittingly, engaged in a kind of environmental monitoring, reading the health of their world through biological irregularities.

Similarly, reading facial morphology and skin markings (crow’s feet, moles, and so on) — the focus of metoposcopy — might reflect underlying developmental or genetic stressors. For instance, minor facial anomalies such as wide-set eyes, low nasal bridges, or deep palmar creases have been statistically associated with prenatal alcohol exposure, genetic disorders, and hormonal imbalances (e.g., Fetal Alcohol Spectrum Disorder; see Streissguth et al., 1996). What may appear as superficial or arbitrary markers can, when systematically observed, point to internal or systemic irregularities that would otherwise remain hidden.

A similar line of reasoning can be applied to palmistry, or palm reading, often dismissed as mere entertainment, touches upon legitimate lines of inquiry: dermatoglyphics — the scientific study of fingerprints and skin ridge patterns — has been shown to correlate with certain chromosomal abnormalities (e.g., in Down syndrome or Turner syndrome) and neurological development (Cummins & Midlo, 1961). Though such associations are probabilistic rather than deterministic, they echo the ancient belief that external form might reflect internal disposition or history, and (I argue) are justified by the same logic as the sciences listed above.

Freudian psychoanalysis follows a logic similar to that of augury: it treats recurring dreams, symptoms, and behaviors not as random events, but as indirect, symbolically encoded signs of deeper psychological realities. Dreams in particular are viewed as aggregations of subconscious emotional states — manifestations of repressed trauma, internal conflict, or unmet needs. The analyst’s task is to interpret these surface-level anomalies to uncover the concealed psychological dynamics beneath. This method, while often dismissed as speculative, has found at least some empirical support in contemporary neuroscience and clinical psychology (Hoffman, 2017).

Modern studies confirm that dream content is shaped by emotional salience, especially in conditions like PTSD. Trauma survivors frequently experience recurring dream motifs. Many report recurrent dreams in which they are unable to move or call for help, often in ambiguous or distorted settings -- ones which symbolically consolidate core emotional experiences such as helplessness or fear. Research shows these patterns correlate with symptom severity and diagnostic outcomes (Nielsen & Levin, 2007; Scarpelli et al., 2019). Neuroimaging further supports the link between dream formation and activity in brain regions governing memory, threat detection, and emotional regulation. Taken together, these findings suggest that dreams function as indirect psychological indicators, much like animal behavior signals environmental disruption. In both cases, meaning emerges not from isolated instances, but from the aggregation of signs that, when interpreted skillfully, reveal otherwise hidden truths.

Viewed in this light, these practices are not inherently irrational. They may represent early, intuitive efforts to systematize correlations between observable surface features and hidden physiological or environmental states. As with augury, their value lies not in mystical causation, but in pattern recognition: the aggregation of indirect signs that point toward underlying truths.

In contrast to more arbitrary or randomized methods of divination — such as tarot cards or bibliomancy (random verse-dipping) — these practices plausibly connect observable phenomena to real underlying causes, even if the causal pathways are complex or opaque. They invite us to see the world not as a set of isolated events, but as a system of symbolic surfaces where the visible may yet speak of the hidden.

5. GPT AS DIGITAL AUGURY:

Generative Artificial Intelligence, particularly Large Language Models (LLMs) like GPT, offers a striking modern analog to the logic of divination outlined above. LLMs generate coherent and highly accurate predictions not (of course) by understanding grammar in any causal or conceptual sense, but by statistically modeling vast patterns in linguistic data. Despite lacking semantic comprehension, they routinely outperform humans in text generation and linguistic prediction. In this sense, GPT functions as a digital augur.

There appears, then, to be no principled reason why the theoretical framework developed here — applied to astrology, augury, aruspicina, metoposcopy, palmistry, and dream interpretation — cannot be extended to create a science of divination in general. If such practices are understood not as inherently mystical arts, but as empirical attempts to correlate indirect signs with future events or hidden states of affairs, then the logic of divination may be recast as a form of predictive reasoning. Indeed, LLMs, on this theory, offer demonstration of how divination can be expanded: the form of inference used in divination can be valid, even if the symbolic medium is causally inert.

In this light, divination is best defined not by its historical or esoteric trappings, but by its epistemic structure: wherever emergent information arises from the aggregation of direct signs into a symbolic pattern, and whenever that pattern reliably tracks real phenomena, divination is at work. It is the disciplined reading of conjoint anomalies — the interpretation of converging, non-causal signs that together point toward broader systemic shifts. Just as LLMs extract meaning not from individual words, but from patterns distributed across massive datasets, so too do traditional divinatory systems derive significance from structured convergence of signs in symbolic aggregates.

With sufficient modeling and refinement, such practices may eventually achieve the status of a formal sciences of indirect signs — fields grounded not in superstition, but in rigorous statistical insight. Just as astrology has been provisionally rehabilitated in the preceding argument, so too might the general method of divination be recognized as a legitimate, if still underdeveloped, epistemology of the emergent.

6. REASSESSING CORRELATION AND CAUSATION

For all that, it’s obviously still true that correlation does not equal or imply causation. Establishing causation requires more rigorous isolation and testing to establish more than mere correlation does, and may appear to suggest something about the metaphysics of nature which mere correlation simply does not.⁵

That said, correlation, if it is large enough, is sufficient to become highly predictive, and accurate prediction is a large reason why knowledge of causation is prized in the first place. That is, we care about causes in large part because they allow us to effectively navigate an uncertain world.

Hence, although the doctrine that correlation does not equal causation is always logically true, in certain cases (namely the ones outlined above), this fact is, practically speaking, irrelevant. We do, in fact, make successful predictions that allow us to navigate more successfully through the world, and we do so on the basis of actually observed correlations – usually without knowing the underlying causes.

Does this mean that we are obligated to suppose a metaphysics for astrology, wherein the stars, through a peculiar sort of cosmic radiation, directly influence our actions, destinies, and health outcomes? Not in the slightest.

Indeed, it is unlikely that there is any direct influence of the stars on human affairs. They are simply too far away, and for that reason, claims of direct causal effects have either not proven to be true, or positively proven not to be true. Nonetheless, the mere fact that stars do not directly influence human affairs, still allows that stars and constellations might be predictive of, e.g., human health and flourishing. One need only show that there is such a thing as an “indirect sign,” and that stars are just such “indirect signs.” Looking at the night sky in a considered way,⁶ we can access a bundle of information all of which we know have strong implications for human health. Indeed, since the stars effectively aggregate this information, it stands to reason that their position will be more predictive than any more limited array of information from geography, climate, or season, etc.

7. A NOTE ON OBJECTIONS TO CORRELATIVE REASONING:

Before generalizing further, it’s important to acknowledge why correlative reasoning has historically been viewed with suspicion — and under what conditions it might deserve cautious rehabilitation.

Correlative reasoning has long been vulnerable to misuse. Spurious correlations can arise purely by chance or from hidden confounding variables.⁷ Human cognition is notoriously prone to apophenia (the tendency to perceive patterns where none exist) and confirmation bias (where one selectively notices evidence that confirms existing beliefs). These psychological tendencies, combined with the misuse of correlative reasoning in pseudoscientific frameworks (e.g., phrenology, racial profiling, pop astrology), help explain the modern scientific aversion to associative methods lacking causal articulation.

Furthermore, correlative models often struggle with generalizability: what holds in one context or dataset may not hold in another. Without proper grounding or rigorous validation, these methods can become tools for self-deception, exploitation, or unfalsifiable ideology.

That said, the possibility of abuse does not preclude the possibility of proper use. In particular, correlative reasoning may yield valid insights when supported by:

1. Strong, repeated correlations across large and varied datasets;

2. Temporal and contextual consistency;

3. Redundancy, where multiple signs converge to support a prediction.

4. Replicability and Proven Predictive Validity Over Time

5. Cross-Domain Plausibility or Coherence, where predictions made are strongly compatible with known science in other domains.

These are precisely the conditions under which modern machine learning models thrive. LLMs make accurate predictions – not through causal modeling, but through statistical convergence across diverse contexts. In this light, divinatory reasoning can be reinterpreted not as superstition, but as an early attempt to detect and interpret indirect signs — symbolic aggregates that, though causally inert themselves, meaningfully encode causal structures.

What correlative reasoning demands is not rejection, but the right epistemic environment: large-scale data, iterative validation, and epistemic humility in the face of uncertainty.

8. IN FORMAL DEFENSE OF ASTROLOGISTS, PSYCHICS, AND DIVINERS:

The intuitive part of the human mind, referred to as System 1 by Nobel laureate Daniel Kahneman and his collaborator Amos Tversky, is characterized by its ability to produce rapid, automatic judgments in response to complex or uncertain stimuli. Unlike the slower, more effortful, and analytically rigorous System 2, System 1 operates with remarkable speed and efficiency, relying on pattern recognition and heuristic shortcuts. It excels in environments where decisions must be made under pressure of time or with incomplete information. However, its inner workings are largely opaque to conscious introspection. As Kahneman puts it in Thinking, Fast and Slow (2011), “System 1 operates automatically and quickly, with little or no effort and no sense of voluntary control.” The conclusions it delivers often feel self-evident, even when their basis cannot be readily articulated.

Strikingly, this kind of cognitive processing closely resembles how Large Language Models function. Like System 1, LLMs use vast quantities of past data to make high-probability inferences without any transparent causal model or deductive logic. They are fundamentally black box reasoning systems — able to generate coherent, contextually appropriate responses or predictions based on prior examples, even if the internal logic of their decision-making remains inscrutable. Some cognitive scientists and AI researchers have suggested that LLMs in fact draw inspiration from neurological models of human language processing and statistical learning (Manning et al., 2020). In this view, LLMs operationalize a kind of externalized intuition — a computational approximation of System 1 that, like its human counterpart, can be extraordinarily effective without being necessarily explicable.

The relevance of this cognitive architecture to the present inquiry lies in its potential to explain the enduring presence and attested efficacy of astrologers, diviners, and psychics across a vast array of cultural traditions — from Jerusalem and Babylon, to India and China, to Ancient Greece and Rome. These practitioners were often credited with making remarkably accurate predictions or diagnoses based on patterns that eluded conscious explanation. If we hypothesize that such individuals possessed highly trained or unusually sensitive System 1 faculties, then their successes might be understood as intuitive responses to aggregated signs — indirect cues embedded in their environment or social context. Like LLMs or expert diagnosticians, it’s possible that they could have been detecting statistically meaningful patterns without articulating (or being able to articulate!) the causal logic behind them.

To this, natural language offers a compelling example of why humans intuitively navigate a system of signs in the first place. In Peirce’s schema, language is a system of conventional signs. Conventional signs occupy a conceptual space between natural (causal) and statistical (correlational) signs: words do not cause their own meanings, nor are they entirely statistical aggregations — but rather, they derive their significance from (ultimately arbitrary) socially agreed-upon norms and usage patterns.

Humans, especially in early childhood, are especially attuned to learn systems of such signs. They acquire language not through explicit instruction in grammar or logic, but by intuitively absorbing the patterns of speech they hear. This process is almost entirely non-conscious and heavily reliant on pattern recognition — learning what “sounds right” based on frequency and context, rather than abstract reasoning.

Large Language Models (LLMs) operate analogously: they process language not by understanding it in a human sense, but by modeling its conventional usage patterns through exposure to vast corpora. In this way, both human infants and LLMs acquire linguistic fluency not by learning rules per se, but by internalizing the probabilistic associations among signs. This convergence suggests that the intuitive grasp of symbolic systems — whether by human or machine — may rely on a broader, shared logic of sign-processing that transcends clear causal explanation. It is precisely this logic, I am claiming, that undergirds both natural language acquisition and pattern-based inference characteristic of divination.⁸

9. AN IMPORTANT CAVEAT, REPEATED:

While this essay argues that practices like divination, astrology, augury, and aruspicina do, in principle, rest on a sound scientific basis, this should not be taken as an endorsement of how these disciplines as they are currently practiced today, This essay is not defending astrology columns or late-night psychic hotlines. As I have already indicated, many contemporary forms of these arts conspicuously lack the scientific rigor and statistical discipline necessary to validate their claims, and do not even attempt to justify this lack of validation.

However, the central argument here is that such rigor could, in theory, be supplied by sufficiently powerful statistical models — such as Large Language Models or other forms of machine learning applied to the relevant domains of (e.g.) augury or astrology. It’s at least conceivable (in some cases highly conceivable!) that these ancient methods might be refined into something far more credible — as a kind of proto-scientific reasoning, ready for revival. It’s a call not for superstition, but for reformation: for the elevation of these old arts into a more rigorous and honorable science of signs.

10. CONCLUSION: TOWARDS A SCIENCE OF SIGNS:

In this essay, we have revisited the historically dismissed practices of astrology, augury, aruspicina, metoposcopy, and dream interpretation — not to mystify them as actually practiced but to clarify their latent epistemological structure. Rather than being merely irrational relics, many of these arts may in fact represent our own early attempts at structured inference from indirect signs. In dismissing these practices wholesale, modern science may have overlooked a class of reasoning that is predictively valid, even when not causally explicable.

We have argued that sufficiently strong correlation, though distinct from causation, can be pragmatically sufficient for accurate prediction. This insight is especially salient in the age of Large Language Models, whose architecture and performance mirror System 1 reasoning in the human brain: fast, intuitive, data-driven, and often inexplicable even to the reasoner themselves. If we accept such a mode of inference in our machines, we must consider the possibility that historical diviners and astrologers — especially in cultures like Jerusalem, Babylon, China, India, and Rome — may have used their own evolved, intuitive black-box faculties to detect patterns that modern science has only recently begun to quantify.

Just as LLMs predict not through understanding, but through attunement to patterned data, so too might a reimagined form of divination offer a pragmatic, probabilistic way of perceiving the world. What is required now is not belief, but experimentation. The call, then, is for collaboration — between data scientists, philosophers, cognitive psychologists, and scholars of religion and symbolism — to test the predictive validity of symbolic systems long considered unscientific. It may be that in doing so, we recover not only forgotten modes of insight, but new significance in our pursuit of knowledge under uncertainty.

11. APPENDIX: A PROPOSED EXPERIMENT:

To explore the central hypothesis of this paper – that divination is (or can be developed into) a valid science of correlative reasoning, and that these methods are currently effectively employed by LLMs and other machine learning algorithms – we propose an experiment: A machine learning system could be trained to correlate astrological ephemerides (birth dates, times, and locations) with health outcomes using longitudinal public datasets. The hypothesis: star position at birth will emerge as a statistically significant proxy for variables like seasonality, geography, and family dynamics — thus becoming a reliable indirect sign.

Specifically, our hypothesis would be that star position at birth will emerge as a statistically significant proxy for variables like seasonality, geography, and family dynamics — thus becoming a reliable indirect sign. Further, we predict that star position will be more predictive than any of the direct variables considered on their own. In this way, researchers could empirically test whether traditional divinatory frameworks like astrology hold actionable predictive value.

If validated, this would represent a rigorous reentry of symbolic divination into the empirical fold, with implications not only for the history of science but for how we understand human reasoning, past and future.

.

Artwork by Carin Yoo

---

Cole did his undergrad work in Classics and Philosophy at Harvard University and received an MS at Oxford (UK). He taught Ancient Greek at Oxford and co-founded Oxford Latinitas, a society of Oxford academics dedicated to reviving Latin and Greek in scholarly use. He now lives in New York City, where he co-organizes the NY Philosophy Club, and will be attending NYU in the fall to continue his work on Aristotle’s Philosophy of Science.

BIBLIOGRAPHY:

Primary Sources / Philosophical Works

• Augustine. On Christian Doctrine (De Doctrina Christiana). Translated by D.W. Robertson Jr. Indianapolis: Bobbs-Merrill, 1958.

• Hobbes, Thomas. 1651. Leviathan. Edited by Richard Tuck. Cambridge: Cambridge University Press, 1996.

• Monton, Bradley and Chad Mohler. 2021. "Constructive Empiricism", The Stanford Encyclopedia of Philosophy. Edited by Edward N. Zalta (ed.). URL = https://plato.stanford.edu/archives/sum2021/entries/constructive-empiricism.

• Peirce, Charles Sanders. Collected Papers of Charles Sanders Peirce, Vol. 2: Elements of Logic. Edited by Charles Hartshorne and Paul Weiss. Cambridge, MA: Harvard University Press, 1932.

• Shakespeare, William. Julius Caesar. Edited by David Daniell. London: Bloomsbury Arden Shakespeare, 1998.

---

Scientific and Empirical Studies

• Barclay, Kieron. 2015. “A Quantitative Study of Birth Order and Health in Later Life.” American Journal of Epidemiology 182(8): 682–690.

• Beyer, W.N., D.E. Heinz, and A.W. Redmon-Norwood. 1996. Environmental Contaminants in Wildlife: Interpreting Tissue Concentrations. Boca Raton: CRC Press.

• Blaustein, Andrew R., and Pieter T.J. Johnson. 2003. “The Complexity of Deformed Amphibians.” Frontiers in Ecology and the Environment 1(2): 87–94.

• Cummins, Harold, and Charles Midlo. 1961. Finger Prints, Palms and Soles: An Introduction to Dermatoglyphics. New York: Dover Publications.

• Disanto, Giulio, Julian C. Morahan, George Ebers, and Sreeram V. Ramagopalan. 2012. “Season of Birth and Risk of Multiple Sclerosis: A Systematic Review and Meta-analysis.” BMJ 344: e675.

• Hoffman, Martin. 2017. “Psychoanalysis as Science.” In Thomas Schramme and Steven Edwards, ed., Handbook of the Philosophy of Medicine. Dordrecht: Springer Science+Business Media, 925-945.

• Hofkirchner, Wolfgang. 2013. “Emergent Information. When a Difference Makes a Difference…” tripleC: communication capitalism & critique 11(1): 6-12.

• Nielsen, Tore, and Ross Levin. 2007. “Nightmares: A New Neurocognitive Model.” Sleep Medicine Reviews 11(4): 295–310.

• Scarpelli, Serena, Valentina D’Atri, Michele Gorgoni, et al. 2019. “Dreams and Nightmares in PTSD.” Sleep Medicine Reviews 43: 70–81.

• Sulloway, Frank J. 1996. Born to Rebel: Birth Order, Family Dynamics, and Creative Lives. New York: Pantheon Books.

• Torrey, E. Fuller, John J. Bartko, and Robert H. Yolken. 1997. “Toxoplasma gondii and Other Risk Factors for Schizophrenia: An Update.” Schizophrenia Bulletin 23(1): 1–6.

• Wadley, Jonathan, and Julian Borger. 2005. “Tsunami Animals ‘Fled to Safety’.” The Guardian, Jan 3, 2005. URL = https://www.theguardian.com/world/2005/jan/03/tsunami2004.indonesia .

---

Cognitive Science and AI

• Kahneman, Daniel. 2011. Thinking, Fast and Slow. New York: Farrar, Straus and Giroux.

• Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Theory: An Analysis of Decision under Risk.” Econometrica 47(2): 263–291.

• Manning, Christopher D., et al. 2020. “Emergent Linguistic Structure in Artificial Neural Networks Trained by Self-Supervision.” Proceedings of the National Academy of Sciences 117(48): 30046–30054.

1. For instance, here is a correlation, which is not a cause: ice cream sales go up in summer. Shark attacks also go up in the summer. But we are not therefore justified in concluding that eating ice cream somehow causes shark attacks. Rather, both ice cream consumption and shark attacks are correlated to an actual cause: the weather being hot in the summer, and people wanting to cool down, either by eating ice-cream, or by swimming (even in shark-infested waters).

2. To go back to our shark-and-ice-cream example: whereas ice cream sales do not directly predict shark attacks (nor vice versa), both the increase of ice cream sales and the increase of shark attacks positively (even causally!) correlate with an underlying shift in “climate,” (a term I am using generally to indicate a broader cause or set of conditions which affects multiple facially unrelated causes). In this case, both increased ice cream sales and increased shark attacks are indirect signs that it is summer, that the weather is hot, and that people want to cool down.

3. However, for a very different interpretation of astrological signs, see Kilaya Ciriello, “Know Thyself & the Practice of Astrology within Plato,” in this month’s NYC Journal of Philosophy.

4. It may be worth noting that the most common use of the term “emergence” in philosophy has been metaphysical, suggesting that new “things” emerge from its constituent parts, over and beyond the aggregate itself. This is useful in biology in particular, where consolidation of various biological functions produces a human being (for example), which is a unit that (might be) more metaphysical than just the sum of its individual cells. In the present discussion however, what “emerges” is (arguably) not a new thing but a new predictive power. On this, see Hofkirchner, 2013

5. Constructive empiricists such as Bas van Fraassen, however, deny that such a suggestion is, strictly speaking, a scientific implication. See Monton and Mohler, 2021.

6. Quite pleasingly, the English word “consider” itself has an astrological etymology. It derives from the Latin con-sīderāre—con (together) and sīdus (star) — literally meaning “to observe the stars together.” In its original usage, to “consider” was to take into account the configuration of celestial bodies in order to infer some underlying truth or impending event. A somewhat tongue-in-cheek implication of this etymology is that divination is active whenever we are considering anything at all, that “consideration” technically defined refers to the same mental operation which is centrally at play in divination.

7. For a comprehensive (and very entertaining) catalogue of such misuses, see here: https://www.tylervigen.com/spurious-correlations

8. On this view, divination may have emerged not as an aberration of reason, but as a culturally codified extension of evolved cognitive heuristics—offering a structured way to respond to complex patterns before science could render them explicit. In other words, divination emerged through evolution, concomitantly with human beings’ ability to speak.