A while ago, I thought I had found something enormous. I had been looking at how people and organizations remain stable while operating under uncertainty, and four properties kept appearing: honesty, balance, faithfulness and courage. Honest information seemed necessary for responding to reality. Competing pressures had to remain in some workable balance. Useful action often had to continue before feedback arrived, and sometimes preserving something important required accepting an immediate cost.
The pattern seemed remarkably portable. In behavioral science, I could describe honesty as reduced self-deception, balance as regulation among competing drives, faithfulness as persistence through delayed reinforcement, and courage as accepting personal cost to maintain value-consistent action. Then I tried engineering. Signal fidelity looked like honesty. Interacting control processes looked like balance. Stability across delayed feedback looked like faithfulness. Accepting an efficiency penalty to preserve a safety margin looked remarkably like courage.
Medicine worked too. So did journalism, finance, criminal justice, music and cooking. Religion worked almost embarrassingly well. By the time I had translated the same four properties through Catholicism, Taoism, Sikhism, Sufism, Buddhism, Kabbalah and Advaita Vedanta, alongside physics, chemistry, ecosystems, economics, architecture and artificial intelligence, the conclusion seemed difficult to avoid.
I called it Justice-in-Becoming: The Universal Grammar of Stability Under Uncertainty.
The problem was that I had designed an experiment the pattern could hardly lose.
The pattern really was there
That doesn't mean the mappings were nonsense. Consider an engineered control system. Reliable measurements matter. Interacting control processes have to remain within workable ranges. Delayed feedback can create instability, and a designer may accept an immediate performance penalty in exchange for greater robustness. Now consider journalism. Accurate information matters there too. Speed competes with verification. Consequences can arrive long after publication, and a newspaper may sacrifice short-term attention by refusing to publish something it can't adequately verify.
Those situations genuinely rhyme. The mistake begins when the existence of a rhyme becomes evidence that both domains instantiate the same deeper thing. Signal fidelity isn't literally honesty. An electrical circuit can't deceive us in the moral sense. A stability margin isn't courageous, and a bridge accepting load isn't faithfully persisting through adversity. Those were translations I supplied because I had already decided which relationships I wanted to preserve across domains.
Once that is visible, the original experiment changes character. Instead of twenty-five domains independently revealing the same four-part grammar, one observer carried a four-part grammar through twenty-five domains and repeatedly found ways to express what he encountered in its vocabulary. Every successful mapping then made the next one easier to see. The examples felt independent because they came from different disciplines, while the process producing the comparison remained largely the same.
That's a wonderfully efficient way to manufacture confirmation bias without intending to.
Start with a pattern in one domain and abstract it until many of the surface details disappear. Carry the abstraction somewhere else and look for something that can occupy each position. If a reasonable mapping can be constructed, count the second domain as evidence that the abstraction generalizes. Repeat this enough times and the same pattern begins appearing everywhere, with each new example making the accumulated case feel stronger.
The framework is now participating in the production of the evidence that appears to confirm it. No dishonesty is required. Sincere excitement may actually make the process easier because every successful translation creates another reason to keep looking.
Courage enters the kitchen
Cooking gives us a wonderfully ordinary example. The original framework described successful cooking under uncertainty as requiring accurate ingredient selection and preparation, balanced control of heat, timing and seasoning, persistence while results develop, and a willingness to make short-term adjustments or exercise restraint to preserve the final dish.
From there, the translation was easy. Accurate ingredients and preparation became honesty. Heat, timing and seasoning became balance. Staying with the process became faithfulness. Adjusting or exercising restraint for the sake of the finished dish became courage.
Nothing about that mapping is particularly absurd. It might even help someone think about cooking. The trouble begins when we reverse the direction of inference and treat the successful translation as evidence that cooking has independently revealed the same underlying grammar as moral courage, electrical engineering or contemplative practice.
Perhaps we've learned something about cooking. Perhaps we've learned something about the abstraction. Usually we've learned something about both, and the interesting question is how much of each.
Some concepts travel precisely because we've stripped away enough detail to let them travel. Feedback is an easy example. A thermostat receives information about temperature and changes its behavior. A newspaper receives a correction from a reader and changes an article. A person experiences pain and changes how they move. A business sees falling sales and changes a product. Calling all four feedback can be useful because each contains some relationship between an earlier state, information about consequences and subsequent behavior. The abstraction works because we're temporarily ignoring enormous differences among the things being compared.
That isn't a flaw in abstraction. It's what abstraction does. The mistake is forgetting what we discarded while making it.
The same thing happened with my original grammar. Accurate information, interacting constraints, persistence through delay and acceptance of immediate cost are broad enough to describe many situations. Their portability may tell us that we've found a useful level of description. It doesn't, by itself, tell us that we've found the hidden grammar underneath everything being described.
Not every rhyme is doing the same work
Some cross-domain similarities really do involve shared mechanisms. Feedback delay in one engineered control system can be meaningfully compared with feedback delay in another because the mathematics governing their behavior may share structure. An equation developed for one physical situation can sometimes predict another because both systems instantiate relationships captured by the mathematics. In cases like that, the abstraction does more than provide a metaphor. It constrains what should happen.
Other comparisons are analogical. Signal distortion in electronics and dishonesty between people aren't the same mechanism, yet comparing them may still help us notice that both involve decisions being made from representations that differ from the conditions they are supposed to represent. The useful product of the analogy isn't the claim that distortion is dishonesty. It's the question the comparison gives us: what happens downstream when the incoming representation becomes unreliable?
Then there are similarities created mostly by the flexibility of our categories. If courage can mean accepting social disapproval, tolerating medical discomfort, building engineering safety margins, restraining a chemical reaction, introducing musical tension and adjusting the seasoning in dinner, we should at least wonder whether courage has become so abstract that almost anything involving short-term cost can satisfy it. A concept that fits everything may have discovered something universal. It may also have become difficult to prove wrong.
The distinction matters because cross-domain reasoning itself isn't the problem. We use it constantly, and sometimes it produces extraordinary results. The question is what kind of work the similarity is doing in each case. Is there a shared mechanism? A shared mathematical relationship? A useful analogy? A broad descriptive category? Or have we simply found a vocabulary flexible enough to absorb whatever we encounter?
Those possibilities can overlap. We don't need to choose one explanation for every comparison in advance.
Mathematics doesn't rescue the pattern
At one point I even translated the framework into symbolic mathematics. If an internal state was represented by x and an expressed state by y, honesty could become congruence between them. Delayed feedback could become τ > 0. Short-term cost could become Δc < 0. Long-term coherence could then be represented as a trajectory in which the relevant states converged rather than diverged.
It looked impressive, which was part of the problem.
The notation wasn't necessarily wrong. What mattered was what I thought the notation demonstrated. Mathematics can reveal consequences that aren't obvious from ordinary language when its variables, relationships and assumptions correspond to things we can independently observe or measure. Writing an existing conceptual claim in mathematical notation doesn't provide that validation by itself.
I hadn't derived Justice-in-Becoming from mathematics. I had translated Justice-in-Becoming into mathematics. Once I see that distinction, the notation becomes potentially useful again because I can ask whether any of those variables can actually be defined, measured and related in ways that survive contact with observations I didn't choose to fit the framework.
That suggests a better test for the rest of the pattern too.
If we identify a structure in one domain and use it to predict something previously unknown in another, the abstraction has done more than redescribe what we already knew. If the structure makes some mappings difficult or impossible, we gain boundaries around where it applies. We can remove one supposedly necessary property and see what happens, reverse a proposed relationship, compare cases where the pattern appears with cases where it doesn't, or ask different observers to describe the same material before giving them our preferred vocabulary.
Most importantly, we can establish the observations before translating them. Instead of asking where courage appears in a system, we can ask what immediate and delayed tradeoffs actually occur. Instead of looking for balance, we can identify the interacting variables and watch what happens as their relationships change. Instead of searching for honesty, we can examine what information is available, how accurately it represents the relevant conditions, and what follows when it doesn't.
The old words might return after that.
If they do, we'll know more about why.
The observer belongs in the experiment
This has changed how I think about pattern recognition more generally. When I see the same shape appearing in several places, I now have at least three things to examine: the objects being compared, the abstraction through which I'm comparing them, and the observer doing the comparison.
I decide which features seem important. My previous experiences influence which similarities become salient, and once I've named a pattern, the name changes what I'm likely to notice next. None of that makes observation useless. It means the process producing the observation belongs somewhere in the analysis.
Artificial intelligence makes this especially interesting because a language model is extremely good at constructing plausible relationships among concepts. Give one a framework and ask how that framework appears in chemistry, Buddhism, music and economics, and it can probably produce convincing mappings for all four. Those mappings may help us explore what follows from an abstraction, expose relationships we hadn't considered, or generate questions worth investigating. Their fluency doesn't make them independent confirmation that the abstraction exists outside the conversation.
That distinction becomes harder to notice during a long interaction because the model can also help refine the framework itself. I notice a pattern, the model develops it, the developed version becomes easier to apply elsewhere, and the new applications make the original pattern appear increasingly coherent. Something useful may genuinely be happening in that loop. Internal coherence is information. It just isn't the same thing as independent validation.
Going in search of other people helps, although not automatically. If someone working independently in psychology, systems science or artificial intelligence describes something resembling a pattern I've been exploring, that convergence deserves attention. Yet I still choose which researchers to read, which passages seem relevant and which level of abstraction makes their vocabulary resemble mine. Another person's work can become another surface onto which I project the framework if I only ask how well it matches what I already believe.
The more useful question is whether we're actually observing the same thing. Do our measurements correspond? Do our explanations make different predictions? Would their result have looked different if my interpretation were wrong? Can their framework explain something mine can't? Independent observers become particularly valuable when their observations can resist my interpretation rather than merely supplying another vocabulary for it.
Giving the pattern somewhere to be wrong
None of this has made me less interested in patterns. A pattern becomes more interesting when it survives conditions that could reasonably make it disappear.
If the same mathematical relationship predicts behavior in several unrelated physical systems, that matters. If different observers independently derive similar categories from the same data before seeing one another's work, that matters too. If an abstraction tells us where something will fail rather than explaining the failure after we already know the outcome, we've learned something different from what a good analogy alone can provide.
Disagreement becomes useful under this approach. Suppose two people examine the same examples and compress them differently. We don't immediately have to decide which person discovered the true underlying grammar. We can compare what each compression preserved, what each discarded and what each lets us anticipate. Perhaps one predicts something the other misses. Perhaps each works under different conditions. Perhaps their disagreement reveals an assumption neither observer had noticed.
Reality gets another turn.
That doesn't require us to become suspicious of every resemblance or embarrassed whenever an analogy helped us think. It means the analogy has done its first job when it gives us somewhere useful to look. What happens after we look determines how much more weight it can carry.
What remains of Justice-in-Becoming?
Quite a lot, actually, although it looks different once I stop asking the rest of reality to validate it.
I still think honesty matters in human relationships and institutions. Wisdom and compassion can pull against one another in ways worth holding rather than prematurely resolving. Will without reason and reason without the capacity to act can each fail differently. Humility can protect us from certainty while magnanimity can prevent humility from becoming paralysis. Faithfulness and courage can describe commitments I value when consequences are uncertain and immediate incentives point somewhere else.
Those are ethical commitments and useful tensions. They don't need chemistry to validate them.
An electrical engineer doesn't have to discover courage in a control loop for courage to matter. Taoism and Catholicism don't have to instantiate the same universal grammar for either tradition to contain useful insight. A bridge doesn't need moral integrity for structural integrity to be worth understanding. Once those things are allowed to remain different, the similarities among them become more interesting because I no longer need every comparison to carry the weight of proving the original theory.
The old four-part grammar can still work as a questioning instrument. When something is becoming unstable, I can ask whether the information reaching us is reliable. I can look for competing pressures that have fallen out of workable relation. I can ask whether delayed feedback is causing us to react too quickly or persist too long. I can examine what immediate costs we're unwilling to accept and whether avoiding them is creating larger costs later.
Those questions travel well because they're deliberately abstract. Sometimes they'll reveal something useful, and sometimes they won't. A failure doesn't have to be translated until the framework survives. It can simply show us where the framework stops being useful, where another description works better, or where two things that looked similar from a distance turn out to behave differently when examined closely.
That gives me a different answer to the question I thought I had already answered.
Why can the same pattern appear everywhere?
Sometimes the same mechanism really does recur. Sometimes different mechanisms share a structure at the level we're examining. Sometimes analogy allows one domain to illuminate another. Sometimes we've compressed away the differences that would have made the resemblance disappear. Sometimes we went looking for the pattern, and sometimes, after finding it once, we became extraordinarily good at finding it again.
Several of those things can be happening at the same time. The interesting work begins when we try to separate them.
I once looked across engineering, psychology, religion, economics, art, medicine and everyday life and saw the same grammar appearing everywhere. At the time, I treated the recurrence as evidence that I had found something underneath them all. I don't need to make that claim anymore, and I don't need to discard the observation either.
The recurrence is enough to make me curious.
Now I can ask what produced it, what the comparison preserved, what it discarded, and what would happen if I approached the same material without knowing the pattern I expected to find. Another observer might compress it differently. Another experiment might break the resemblance entirely. Something I thought was analogy might turn out to share a real mechanism, while something I thought was universal might turn out to have traveled only because I carried it.
That's a better place for the pattern to live.
Not underneath everything.
Somewhere we can keep testing it.
