The Charter and the Boundary
Self-determination requires a self. Defining who is in is the first act of power, and the first place it can go wrong.
TAM-CMN.14 · The Common Mind · The Approximate Mind
A developer building a health ecosystem for the Latino community in the United States sits with a question she cannot answer by writing code. She has to decide what Latino means in the context of her system. Does it include Brazilian Americans, who share a hemisphere but not a colonial language. Does it include Afro-Latinos, whose experience within the community is shaped by a Blackness the community sometimes acknowledges and sometimes does not. Does it include undocumented residents, whose health data carries legal risk the system must either protect against or pretend does not exist. Does it include the second-generation teenager in Phoenix who speaks more English than Spanish and feels the label as an inheritance she did not choose rather than an identity she lives in.
The developer is Latina herself. She built the ecosystem because the mainstream health AI served her community badly, with English-default interfaces, cultural assumptions that did not fit, and a blindness to the specific patterns of disease and treatment and access that shape Latino health outcomes. She built something better, and the something better now requires her to draw a line around the community it serves, and the line is the first act of power, before a single model is trained or a single query is answered.
Every community-defined ecosystem faces the boundary question, and there is no clean answer because communities are not clean. Blackness is plural and contested: it crosses national borders, class lines, and continents, and the Black American experience and the Black British experience and the Nigerian experience share a word and diverge in almost everything the word touches. A community ecosystem built for one of these will not serve the others, and the decision about which one it is built for is a decision about who is inside and who is outside, made by whoever holds the configuration power.
The boundary is not a technical problem. It is a political one, and pretending it is technical is how the politics get smuggled in without anyone voting on them.
This is the first of three tensions the essay holds open rather than resolves.
The second tension is internal. A community is not a monolith, and an ecosystem built by the community’s current leadership will reflect the community’s current power structure, including its internal exclusions. The queer member of a conservative religious community. The woman in a community whose public-facing culture celebrates her and whose private institutions constrain her. The dissenter, the wrong class, the person whose identity sits at the intersection of the community and something the community has not yet decided to accept. An ecosystem that gives the community configuration power over its own AI gives the people who hold that power the ability to encode their preferences into a system that touches everyone the ecosystem serves.
The epistemic layer described in CMN-09 finds its hardest job here. It is one thing to monitor whether a model serves the population it claims to serve. It is another to monitor whether it serves the populations within that population that the builders might not have centered. Internal exclusion is harder to catch than external exclusion, because the data that would reveal it is the data the community may not be collecting about its own minorities, and the people being excluded are the people least likely to have the standing to raise the alarm.
The third tension is the relationship between the community ecosystem and the universal floor. Are they complements or competitors for the same scarce will and capital. A government building a public floor has finite resources. A philanthropic foundation funding community ecosystems has finite attention. If the floor and the community layer are understood as rivals, the likely outcome is that the floor gets funded because it serves everyone and the community layer does not because it serves a subset, and the subset, which is the population the floor serves worst, is once again the last priority.
They are not rivals. They are layers, and the relationship runs in one direction: the floor is the seed, the community ecosystem is the growth.
The open public floor provides the baseline that everyone stands on. It is universal, unassuming, and free. It handles the standardized interactions, the forms, the common knowledge, the capability that does not require community-specific understanding. On top of that floor, community capital, Black private equity among other forms, builds the layer that meets a specific population in its own idiom. The surplus from that layer stays inside the community rather than flowing out as rent. The floor did not create the growth. But without the floor, the growth has no foundation, and the community is building on rented ground that someone else can take back.
The economics of this arrangement are the economics CMN-10 described, applied at community scale rather than national scale. When a community ecosystem runs on the open floor, the community’s capital funds only the community-specific layer, the domain models, the cultural knowledge base, the idiom that makes the system feel like theirs. It does not have to fund the infrastructure underneath, because the infrastructure is public. This is the credit-union principle again: the credit union does not build its own payment rail. It uses the rail that exists and builds the relationship layer on top. The AI ecosystem does not build its own inference stack. It uses the floor and builds the community layer above it. The capital required shrinks, the barrier to entry drops, and communities that could never afford to build the whole thing can afford to build the part that makes it theirs.
The inequality that remains is in the growth above the floor, not in the floor itself. Communities with more capital, more talent, more institutional depth will build richer ecosystems than communities without. This is not fully resolved, and the essay does not pretend it is. But the presence of a floor beneath all of them is categorically better than what exists now, which is no floor at all and community capability entirely dependent on whether the community can afford to build from the ground up. The floor does not eliminate stratification. It moves the question of stratification from whether you have access at all to how much is built above the access you have. That is a large move. It is not the last one.
This is where the intersectional argument arrives at its institutional form. The populations the broad market erases most completely, the people who sit at the intersection of multiple marginalizations, are precisely whom a community ecosystem can reach and the universal floor alone cannot. The Afro-Latino patient whose health outcomes are shaped by both Blackness and Latinidad in ways neither a Black health ecosystem nor a Latino health ecosystem fully captures. The queer elder in a South Asian community whose isolation is compounded by a cultural silence the community’s public institutions have not broken. These are the people who live at the intersection, and the intersection is where the averaged model is most dangerously wrong, because it averages across the dimensions that, for this person, interact.
A community ecosystem that takes its own boundary and internal-power tensions seriously, that monitors its own exclusions through the epistemic layer rather than assuming its own goodness, is the institution best positioned to serve the intersection. Not because the community is inherently virtuous. Because the community is close enough to see the specific person, and the specific person is who the intersection is about.
I wonder whether the hardest act of self-determination is not building the institution but deciding, inside the institution, to protect the member it would be easiest to overlook.
The developer in the health ecosystem will not resolve the boundary question in a way that satisfies everyone. The teenager in Phoenix who does not feel the label will use the system or not, and the system will serve her well or not, and the developer’s decision about whom the ecosystem is for will shape outcomes for people she has never met. She knows this. She holds the tension because the alternative, pretending the boundary does not exist, is how communities build institutions that replicate the exclusions they were created to escape.
The ecosystem she builds will be imperfect. It will serve some members better than others, and the others will include people she did not intend to exclude. The epistemic layer will catch some of that and miss some. The boundary will be drawn and redrawn and argued over and never finally settled, because a living community is not a settled thing. The value of the ecosystem is not that it solves these problems. It is that it makes the problems the community’s own to solve, in its own terms, with its own power, rather than leaving them to be solved by a distant system that does not know the community exists.
How this essay connects to others across The Approximate Mind.