The Violence of Abstraction: What We Lose When We Model

Every data model is an act of violence against the complexity it represents. A database schema with 47 tables reduces the messy, contradictory, context-dependent reality of a business to 47 clean structures with defined relationships. The reduction is necessary. The loss is real. Alfred North Whitehead called this “the fallacy of misplaced concreteness”: treating an abstraction as if it were the thing itself. In data modeling, this fallacy is committed approximately 47 times per schema.

What is lost when we abstract the world into models?

Context, nuance, contradiction, and the lived experience of the people whose reality the model claims to represent. Every category boundary is a decision about what matters and what does not. Those decisions carry ethical weight.

I designed a customer data model that included a field for “gender” with three options: male, female, other. The model was clean. The reality was not. “Other” compressed an enormous range of human experience into a single category that communicated, through its label, that everything not male or female was residual, an afterthought. The model was not neutral. It was a claim about what matters, embedded in a database schema that 200,000 people would interact with.

Whitehead’s fallacy of misplaced concreteness is the mistake of treating abstractions as if they were concrete reality. Every data model commits this fallacy. The model says “a customer has a name, an email, and an address.” The reality is that a customer has a life, a context, and a relationship with the business that no schema can capture. The model is useful. The model is also a lie, a simplification so radical that it distorts what it represents. The ethical obligation is to remember the distortion, to remember that the map is not the territory and that the people reduced to rows in your database are more than your schema can express.

The violence of abstraction is not optional. You cannot model without abstracting. But you can abstract with awareness, naming what you have excluded, documenting the simplifications, and remembering that every data contract is a contract about what will be forgotten as much as what will be preserved.