We stopped judging and handed agency to the machine

We trust the pattern processing of LLMs, but not our own reasoning faculties. The reason is structural.

This is the second in a series of arguments about our epistemic condition, though it can be read on its own. If you’re new here or came for my work on gender, the following is an extrapolation of a pattern from all that work about the conditions of knowledge production that actually suppress our innate faculties. These observations have come from the experience of writing publicly, from implementing LLMs for businesses, and from my own social context in the professional class in Austin.

---

I began this series of arguments with the assertion that listening to a book and reading are distinct acts that shouldn’t be conflated, and that conflating them is a laundering of prestige. The effort spent reading is the point. It matters not whether your recall is the same; your engagement with the ideas will be of lesser quality because the audio medium nearly eliminates friction. From this, we move to the structural mechanism that leads people to claim that there is no difference in the effect of two distinct methods of information intake. This condition also explains phenomena such as the attractiveness of the Blank Slate theory in psychology and why you can’t tell your friends they might want to change course for a better life.

The professional-knowledge-managerial class has decided effortless existence is the good life, leaving no room to make it meaningful — meaning requires effort to cultivate. And to seed one’s garden with the correct ideas requires discernment, or judgment. I prefer to use the latter term precisely because it’s loaded (the PMC loves its euphemisms that lessen the weight of — what else? — judgment).

However, the professional-knowledge class has made judgment socially radioactive. This is a structural condition born of several factors, chief among which is the moralization of observation and pattern recognition.

Observation of behavioral patterns, in particular, forms the basis of our judgment about people and our social context. These are core human faculties on which reasoning is based — the ability to extract generalizations from seemingly unrelated phenomena and to build mental models that can be updated with experience. This means human pattern recognition ability is crucial to designing a good life, which is the accumulated result of small decisions with compounding effects.

Patterns are how we understand the world, yet seeing them is disallowed

Given unfamiliar situations daily, we use prior experience to infer the answer even if we haven’t encountered that exact problem. This is a practice of inductive reasoning. We will never encounter identical social situations, but discerning patterns of behavior is crucial for social survival in unfamiliar contexts. This is judgment — facing incomplete information but acting anyway.

Patterns we see with our own eyes are more descriptive of reality than an institutionally blessed study, and should be the basis of our judgments. We will never have all the data, and that’s why our brains evolved to make decisions without all the evidence. Yet, this process is moralized by the professional class such that people have stopped trusting their own reasoning faculties.

We are willing to judge a situation based on the probability of a given result, but we don’t see it that way consciously. We do this daily, multiple times. Yet, in practice at work and among ourselves, we think exercising judgment is potentially harmful and biased. We judge probabilities using heuristics, and some lead to cognitive biases. But we give people a pass on the availability bias and several others, just not their judgment about behavior.

The concept of bias refers to factors that appear irrelevant to the merit of the choices. This is the clinical definition, which has ballooned to include the observation of group-level behaviors. First, stereotypes have been largely proven true at the population level, which people in this class don’t generally know and wouldn’t accept.¹ The problem arises when population-level statistics determine how one treats an individual in that group, even though the brain has evolved to update mental models in response to new information. We just don’t trust ourselves to actually do so, and especially not the ‘uneducated’, because the danger of bias is located in a lack of credentials.

We are biased some of the time, but that doesn’t mean the actions resulting from the bias are necessarily irrational. What if they’re self-protective and discerning? No one would call a woman problematic for assuming a physical threat when walking alone at night, for example. Bias is generally fine for people in protected groups, just not when they’re unflattering to those groups.

The professional class is obsessed with eliminating cognitive biases — they see humans both as blank slates and as fundamentally prone to bias.² We would rather strip judgment from humans and assign it to LLMs, which we imagine we can control bias within.

Judgment was already eroded by social media because it requires people to step outside their emotions and assess the situation with reason. With the obsession over eliminating one kind of bias, the arbitrary judgment of someone based on race or sex stereotypes, they’ve introduced another bias - one against those who exercise judgment without a concern for its social implications. Truth is often socially inconvenient.

If pattern recognition is moralized, judgment becomes philosophically impossible. And without judgment, there is no human agency.

Judgment in institutions & the consequences of moralizing it

Individual judgment has never been more valuable in institutions, yet never less socially valued. We moralize human judgment and outsource it to LLMs, which cannot judge or make truth claims. This is entirely the opposite of how we should comport ourselves. We’re so afraid of human bias that we trust a model trained on text about behavior that’s actually downstream of the experience, making the text only as good as the producer’s interpretation of others.

A researcher is driven by incentives and trade-offs no differently from a layperson, meaning a study of behavior is no more trustworthy than a human gathering data on that phenomenon, because we can’t know what trade-offs were made or who funded the research.

A model created on vast amounts of human-generated text downstream of human observation cannot be more correct about how people behave than a human. Therefore, an attentive person without credentials or institutional incentives may be observing reality more accurately than a credentialed person. This is an epistemic position about the process of knowledge production, and the demand for citations in such cases illustrates the problem I diagnose.

Academia and related knowledge-producing bodies claim a monopoly on expertise, thus on judgment. At the same time, activism has become the chief concern of much if not most scholarship outside the hard sciences. This simultaneous claim of a monopoly on judgment and its evacuation in the name of progressivism becomes dangerous in institutions that produce knowledge and, therefore, make truth claims.

The people in these institutions actually resist exercising judgment – look at any academic paper or book, and claims will be buried in hedging to ensure none is stated strongly enough to invite falsification. This is applicable to the social sciences and even to biology.

Gender ideology is only the most recent example of this. We were told for a decade plus that gender was both mutable and immutable, depending on the situation. The enduring social function of the Blank Slate theory, which dominates the professional class’s understanding of behavior, has no empirical backing, yet abandoning that framework, even among academics, would invite sanction from the institution and from students who don’t want to hear the uncomfortable but true.

The demand for high-fidelity truth is the death of reasoning

The PMC has come to require high-fidelity truth rather than provisional conclusions as the basis for reasoning and discernment. This requirement stops truth-seeking in its tracks because no one can evaluate every possibility or member of the group before concluding x.

For example, I was standing in line at some conference with a bunch of male engineers, including one from Google.³ They were waxing poetic about how, through the Objectives & Key Results framework, YouTube increased users’ total watch time, as though this were an impressive feat rather than a depressing statistic.⁴

I chimed in, saying that what we choose to measure as corporate workers often has adverse social effects – the increase in YouTube watch time has had negative effects on society because consuming short videos fragments attention. They instead questioned how I could know what’s bad for society, collapsing judgment into relativism. They have forbidden intellectual independence because they have moralized pattern recognition and, therefore, judgment. The Google engineer was affronted.

Therefore, judgment becomes cruel unless based on high-fidelity truth, which is unreachable because we cannot know about every instance of a phenomenon. This means our muscle atrophies. According to the PMC, if a judgment is based on anything other than high-fidelity truth, it is harmful. Thus, judgment can never be exercised for the risk of appearing biased.

Judgment invites accountability

We’ve collectively chosen comfort over growth, so moralized judgment requires creating or naming a hierarchy of ideas. Some have better life outcomes than others because of their choices, and we all know this. We also know that openly saying some actions lead to worse outcomes is deemed socially cruel, so we don’t. People remain small because social feedback is cut off.

Judgment means accountability for how your life turned out, so perhaps when a therapist told me not to judge myself, she wasn’t helping me see how to solve the problem. Accountability becomes impossible when judgment is moralized; if you can’t change course based on bad outcomes, growth is impossible. Growth comes when we see that x is a better way of living than y. You have to, frankly, judge your own behavior to change course.

For example, I’ve written that the feminist script for success emphasizes compliance at the expense of cultivating rare and valuable skills. Time has proven me correct, because LLM-driven layoffs have disproportionately affected women concentrated in easily automated functions.⁵ This is a result of career choices women tend to make, rather than targeting them, but the outcome is identical. Women’s judgment, in particular, has been moralized away by other women to keep us collectively small, so no one may rise above any other, a comfortable mediocrity.

Judgment also implies that some see more clearly than others, which is often interpreted as a status threat because it suggests that the person deems everyone else morally inferior. But without judgment, responsibility is diffuse, and nothing is built because people are too afraid of the social cost.

The outsourcing of judgment

LLMs are consensus-production machines — they hide the exercise of human agency, and then prevent it.

The PMC wants to outlaw thinking and hand it over to a pattern-generation machine, which they seem to think is a pattern-recognition-and-judgment machine. That we don’t know what LLMs were trained on is a feature. Think about how an LLM talks to you. It tries to soften and hedge everything the moment you get too close to certitude, especially if that thing is socially dangerous to acknowledge and requires pattern processing about human behavior.

Institutions thus devalue judgment through LLMs, even as they rely on human judgment to evaluate LLM output. The problem is, judgment can’t be measured. It prevents disasters, so we don’t reward people for it, even if they’re responsible for a better outcome through their judgment rather than their credentials.

If agency requires the exercise of judgment, and we’ve made its exercise socially risky, then we’ve necessarily let our agency erode. Life is now more about feeling than reasoning, even though humans have built civilization through reason. Emotion enriches life, but doesn’t build it.

We’ve deemed judgment suspect when humans exercise it, because it can make someone feel inferior for being wrong. The cost is a weakening of that muscle as we hand agency itself over to LLMs

Delegitimizing judgment via technology and credentials is a seizure of agency from individuals, which, paradoxically, undermines those very institutions when they can’t see the truth any longer.

1. Steven Pinker, The Blank Slate — a comprehensive introduction to the issue.

2. Pinker, citing others.

3. I wrote about my experience working there in the DEI industrial complex.

4. Google invented the OKR framework, and then it suddenly became corporations’ favorite vehicle for obscuring both individual judgment and achievement within the group, thereby preventing individual accountability for outcomes.

5. A few places aggregated data, like Women In Tech Network, the International Center for Law and Economics, and Fortune, and these stats support the argument that women’s career choices for communication-heavy functions are at the root of their disproportionate exposure to LLMs.