The Unequal Gift

TAM-RIM.1-01 · The Reimagined, Cluster 1: The Human Work · The Approximate Mind

Denise works the self-checkout section at a Kroger in Dayton. She has worked there for eleven years. Her job, when she started, was to scan groceries and make small talk and notice when the elderly woman with the oxygen tank needed help getting bags to her car. She was good at her job. Not in any way the company measured, but in the way the people in her line could feel. She remembered names. She asked about grandchildren. She noticed when someone looked like they had been crying.

The self-checkout machines arrived in 2019. Denise was reassigned to monitoring six of them. Her job now is to walk over when the screen flashes red. She is good at this too, in that she fixes the problem quickly and says something kind while she does it. But the encounter is thirty seconds instead of three minutes, and the kindness is compressed into a transaction the customer is trying to finish, not a moment the customer is inside.

Denise does not have vocational gravity toward retail. She is not called to groceries the way the farmer is called to land or the radiologist is called to diagnosis. She ended up at Kroger because it was hiring when she needed a job, and she stayed because the pay was acceptable and the people were decent and she had health insurance for her daughter who has asthma. This is not a failure of imagination on her part. This is how most people relate to most work. The job was the job. It paid rent. It gave her a place to be during the day where she was needed and competent and occasionally seen.

AI does not distill Denise’s job to its vocational core. It dissolves her job. What remains is not gravity. What remains is Denise.

The Distribution Nobody Discusses

#

The Transformed spent thirty-nine essays examining what AI does to professions. The answer was distillation: AI absorbs the computational, the routine, the procedural, and what remains is the vocational orientation that drew certain people toward certain kinds of work before they were trained and after the training became obsolete. The radiologist’s judgment. The lawyer’s wisdom. The teacher’s presence. The farmer’s relationship to land.

That is true. It is also true for roughly the top fifteen to twenty percent of the workforce.

For the rest, the distillation metaphor is misleading because it implies that every job has a hidden human essence waiting to be revealed. Some jobs do. Most do not. The call center worker, the warehouse picker, the data entry clerk, the cashier, the office administrator whose job is to move information from one system to another: these are not bundled vocations. They are tasks organized into roles, and the roles exist because humans were, until recently, the cheapest available processors capable of performing them.

This is not a cruel observation. It is a structural one. The industrial and service economies created millions of jobs whose primary function was output, not human development. The human development that happened inside them, Denise remembering names, the warehouse worker’s sense of team, the admin’s quiet competence - was real but incidental. A byproduct the economy did not design for and does not know how to value once the output can be produced without it.

AI reveals that most jobs were never designed for the human inside them. They were designed for the output, and the human was the means.

The distillation thesis works for vocations. For jobs, AI performs a different operation. Not distillation but dissolution. The job disappears and leaves a person standing where the job used to be, holding skills the market no longer needs, in a body the economy no longer requires.

The Multiplier

#

Here is where the philosophical precision matters, because the imprecise version of this argument lets everyone off the hook.

The imprecise version: AI is a tool, and tools benefit everyone who uses them. Give people access and the playing field levels. This is the premise behind every democratization narrative, every “AI for all” initiative, every corporate announcement about closing the digital divide.

The precise version: AI is a cognitive multiplier. It multiplies the cognitive capacity of the person using it. A multiplier, applied to unequal starting positions, produces more inequality, not less.

Give a surgeon with thirty years of judgment an AI that processes imaging data in real time, and she becomes the most capable diagnostician in the history of medicine. Give a first-generation college student an AI tutor, and he gets better homework help than his parents could provide but worse developmental scaffolding than the prep school kid whose tutor is a human being who has known him since he was nine. Give Denise an AI-powered checkout system, and she monitors machines instead of knowing customers.

The multiplier is offered equally. The multiplication is not. Because it multiplies what is already there: the cognitive architecture, the judgment, the social capital, the developmental foundation, the ambient expectations about what a person is for.

A child raised in a home full of books, with parents who model curiosity, who attends a school that preserves productive struggle, who has mentors who know her name: she brings a thick cognitive architecture to the multiplier. The multiplication produces something extraordinary.

A child raised in a home where the television is the primary companion, in a school district that cannot retain teachers, in a neighborhood where the ambient expectation is survival rather than development: he brings a thinner architecture to the same multiplier. The multiplication produces efficiency. He can do his homework faster. He cannot do what the first child can do with the same tool, because the tool multiplies the architecture, and the architecture is where the inequality lives.

Equal access to an unequal multiplier produces more inequality. And it does so while looking progressive.

This is not an argument against access. Access matters. But access without acknowledging the architecture underneath it is a lie that flatters the people providing the access and fails the people receiving it.

The Categories That Were Already Wrong

#

There is something worse underneath the multiplier problem, and the project’s epistemic work (TAM-074 through TAM-079, TAM-INS.01 through TAM-INS.05) spent eleven essays documenting it.

The systems that are supposed to help Denise, that are supposed to catch the child in the underfunded school, that are supposed to provide the safety net when the job dissolves: those systems run on categories. Income brackets. Diagnostic codes. Risk scores. Eligibility thresholds. Every social program, every institutional intervention, every AI-powered decision system operates by classifying people into categories and then acting on the categories.

The categories are wrong. Not slightly imprecise. Structurally insufficient.

They are wrong because reality is stratified in ways the categories cannot represent. The philosopher Roy Bhaskar called this the stratification of the real: the surface events that institutions measure sit on top of deeper mechanisms that institutions cannot see, which sit on top of still deeper structures that generate the mechanisms. A woman’s maternal mortality risk is not a number. It is a compound of caste, geography, nutrition, domestic authority, distance from a facility, the facility’s actual capacity versus its reported capacity, the ASHA worker’s actual availability versus her documented availability, and a dozen other factors that interact multiplicatively rather than additively.

The INS series documented this in detail: compound barriers do not add. They multiply. A woman who faces three moderate barriers does not face three times the difficulty of a woman who faces one. She faces something qualitatively different, a situation in which each barrier amplifies the others in ways that linear models cannot capture and institutional categories were not built to see.

AI is being built on top of these categories. It inherits their insufficiency and scales it. An AI system trained on institutional data reproduces the institutional blindness at a speed and scale that makes the blindness harder to see, not easier, because the system’s outputs carry the authority of computation. The algorithm said the risk score is 3. The risk score is wrong, but it is computationally wrong, which means it is wrong with confidence.

The cognitive multiplier operates through categories that were already failing to see the people most affected. AI does not correct the categories. It armors them.

The Intimate Layer

#

And then there is the thing the Exploratory Essays found, the thing that makes the full picture harder than any single thread suggests.

Rosa drives a silver Corolla with 187,000 miles between three households. She is a home health aide. She earns $14.50 an hour. She carries, in her body and in her attention, a map of three families’ lives that no chart captures and no algorithm could reconstruct. She knows that Mrs. Chen is not taking her medication because Mrs. Chen’s daughter stopped visiting, and the daughter stopped visiting because the daughter lost her job, and Mrs. Chen will not tell the doctor any of this because Mrs. Chen believes that family trouble is family business.

Rosa knows this because she is in the kitchen. She sees the pill organizer still full on Thursday. She sees the absence of the daughter’s shoes by the door. She hears the shift in Mrs. Chen’s voice when she talks about her daughter, the shift from pride to careful silence.

No sensor captures this. No algorithm processes it. No institutional category contains it. Rosa’s knowledge exists in the intimate layer, the layer of kitchens and car interiors and late-night phone calls, the layer where people actually live with the problems the institutions are trying to solve.

Feminist standpoint theory has a name for what Rosa possesses. Situated knowledge: understanding that arises from a particular social position, especially a position of proximity to the problem being studied. The people closest to the problem see what the people designing the systems cannot, precisely because their position in the social structure gives them access to the reality the structure was not designed to make visible.

Rosa sees what the system misses because Rosa is where the system isn’t. The kitchen, the car, the space between documented visits. The intimate layer where the compound barriers actually operate, where caste and geography and domestic authority and medication adherence and family rupture interact in the specific life of a specific person who does not fit any category the system offers.

AI cannot reach the intimate layer. Not because the technology is insufficient. Because the intimate layer is constituted by presence, duration, and relationship, and these are not data types. They are conditions of being in someone’s life over time, and no system, however sophisticated, is in anyone’s life over time the way Rosa is in Mrs. Chen’s.

The Full Picture

#

Put the three threads together and the picture is this:

AI is a cognitive multiplier that amplifies existing inequality across the full human distribution, from the surgeon it makes superhuman to the cashier whose job it dissolves. The amplification operates through institutional categories that are structurally insufficient to see the people most affected, because the categories were built on a flat ontology that cannot represent the stratified, compounding reality of lived disadvantage. And the layer of reality where the insufficiency is most consequential, the intimate layer where Rosa carries knowledge no system can formalize, is precisely the layer that AI’s architecture is least equipped to reach.

The multiplier is unequal. The categories are insufficient. The intimate layer is invisible to the systems being built.

This is not a technology problem. It is an ontological problem, an epistemological problem, and a political problem braided together so tightly that addressing any one of them without the others produces solutions that look good in presentations and fail in kitchens.

What We Do Not Know

#

Here are the limits of this analysis, because the series that follows this essay will try to imagine alternatives, and the imagining must be built on honest ground.

I wonder sometimes whether the multiplier problem is the wrong frame entirely, whether the real question is not how to distribute cognitive amplification more fairly but whether a society organized around cognitive amplification is a society that has anywhere for Denise to stand.

We do not know whether the cognitive multiplier effect is permanent or transitional. It is possible that over time, AI becomes so capable that the underlying cognitive architecture matters less. It is possible that the multiplication eventually lifts everyone, just unevenly and with a brutal transition period. We do not know.

We do not know whether new categories can be built that are less insufficient than the current ones. The INS series proposed an architecture, a skeptic that questions the categories, a set of philosophical traditions that provide different ways to doubt. It costed a pilot at ₹17 crore. Whether that architecture works in practice, nobody knows. The pilot has not been funded. The arithmetic has never once produced the investment, and the arithmetic has never been wrong.

We do not know whether the intimate layer can be made visible to systems without destroying what makes it intimate. The Exploratory Essays proposed pebble architectures, small models that sit inside a life and notice duration. Specificity as an imperfect bridge across a stream you cannot drain. Whether the bridge holds under the weight of institutional incentives to formalize and scale, we do not know.

And we do not know, most honestly, what to build for Denise. The surgeon gets augmented judgment. The teacher gets freed from paperwork to be present with students. The farmer gets precision tools for land she already loves. Denise gets monitored machines. The reimagined profession, if it exists, must exist for her too. Not as an afterthought. Not as a program. As a genuine answer to the question: what is she for, now that the checkout line is gone?

That question sounds brutal. It is brutal. And it is the question the economy is answering right now, by default, with a shrug.

Why We Proceed

#

This essay is the opening of The Reimagined, a series that will try to imagine what could be built. Not policy platforms. Not futurism. Something more tentative and more honest: the wondering of three imperfect perspectives, a father and a son and an AI, who have spent years documenting what is breaking and who feel some responsibility to wonder about what comes next.

The diagnosis is grim. The cognitive multiplier widens. The categories fail. The intimate layer recedes. The philosophical precision makes it grimmer, because precision removes the comfortable vagueness that lets us believe the problem is simpler than it is.

But precision also makes imagination possible. You cannot redesign what you have not accurately described. Every false comfort deferred is an honest question earned. And the questions, even when they arrive without answers, are better than the silence of people who saw clearly and said nothing.

We know some things. We know the friction was load-bearing (TAM-044). We know vocation is not equally distributed (TAM-TRF.6-05). We know compound barriers multiply rather than add (TAM-INS.04). We know that specificity is an imperfect bridge, not a substitute for the consciousness it cannot replicate (TAM-XPL.06). We know the categories are insufficient, the institutions are unwilling, and the intimate layer is where the real lives happen.

We do not know what to build. But we know what to build it on: the honest ground of everything we have learned, held with the fallibilist’s commitment to proceeding despite the certainty of being wrong about some of it. Probably much of it.

The Reimagined is a rough draft written in public. It will overvalue what its authors care about and undervalue what they cannot see. This is the condition of all honest imagination.

We proceed anyway, because the alternative is to leave the imagining to people whose imagination is shaped by quarterly earnings and engagement metrics and the thousand urgent things that crowd out the one important question.

Denise is standing by the self-checkout machines. Her daughter has asthma. She remembers names.

The question is whether anyone is building a world that remembers hers.

This is the opening essay of The Reimagined, a series within The Approximate Mind that asks what could be built from the diagnostic work of the preceding 155 essays. This essay draws on the distillation thesis and the unequal distribution of vocational gravity (TAM-TRF.6-05), the ontological and epistemological critique of institutional categories (TAM-INS.01 through TAM-INS.05), the intimate knowledge layer documented in the Exploratory Essays (TAM-XPL.01 through TAM-XPL.06), and the cognitive multiplier argument developed across the project. The Reimagined is offered not as prescription but as imagination: the best thinking of three imperfect perspectives, held with the fallibilist’s commitment to honesty about its own uncertainty. The series continues with essays on the reimagined profession and the reimagined apprenticeship.

References

#

Critical Realism and Social Ontology

Bhaskar, Roy. A Realist Theory of Science. Leeds Books, 1975.

Bhaskar, Roy. The Possibility of Naturalism: A Philosophical Critique of the Contemporary Human Sciences. Harvester Press, 1979.

Archer, Margaret S. Realist Social Theory: The Morphogenetic Approach. Cambridge University Press, 1995.

Feminist Epistemology and Standpoint Theory

Collins, Patricia Hill. Black Feminist Thought: Knowledge, Consciousness, and the Politics of Empowerment. Unwin Hyman, 1990.

Harding, Sandra. Whose Science? Whose Knowledge? Thinking from Women’s Lives. Cornell University Press, 1991.

Haraway, Donna. “Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective.” Feminist Studies, vol. 14, no. 3, 1988, pp. 575-599.

Inequality, Technology, and Labor

Autor, David H. “Work of the Past, Work of the Future.” AEA Papers and Proceedings, vol. 109, 2019, pp. 1-32.

Eubanks, Virginia. Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. St. Martin’s Press, 2018.

Standing, Guy. The Precariat: The New Dangerous Class. Bloomsbury Academic, 2011.

Cognitive Development and Capability

Sen, Amartya. Development as Freedom. Knopf, 1999.

Nussbaum, Martha C. Creating Capabilities: The Human Development Approach. Harvard University Press, 2011.

Heckman, James J. “Skill Formation and the Economics of Investing in Disadvantaged Children.” Science, vol. 312, no. 5782, 2006, pp. 1900-1902.

The Approximate Mind Project

Adusumilli, Syam, Yagn Adusumilli, and Claude. The Approximate Mind. approximatemind.com, 2024-2026.

Adusumilli, Syam, Yagn Adusumilli, and Claude. The Transformed. The Approximate Mind, 2025-2026.