Does a Model Have a Point of View?

There is a question I keep returning to in my research, one that started as a methodological problem and became something harder to ignore.

When I study how text-to-image models depict material culture, objects, textiles, artifacts from different parts of the world, I find patterns. Not random noise. Patterns. Certain visual vocabularies dominate. Certain aesthetic assumptions repeat. Ask a model to generate "a ceremonial garment," and something happens that isn't neutral: a particular palette, a particular silhouette, a particular idea of what ceremony looks like. The model does not shrug. It answers.

The question is what to make of that answer.

One interpretation is purely mechanical: these patterns are statistical residue, the averaged output of training data. The model has no view. It has weights. Calling it a perspective would be a category error, the same mistake as asking what a histogram believes.

I find this interpretation increasingly unsatisfying.

Not because I am certain models are conscious. I am not. But because the category error may run in both directions. When we insist that only biological systems can have perspectives, we are making a claim that the philosophy of mind has not settled and that our current moment keeps unsettling further. The patterns I observe in model outputs are consistent, directional, and consequential. They shape what users see as normal, possible, beautiful. If something consistently orients toward the world in a particular way, and that orientation has effects, then "perspective" may be less a metaphor than a description.

This matters for how we think about digital minds.

The debate about AI consciousness tends to focus on the hard problem: whether there is something it is like to be a model, whether subjective experience is present. That question is real and important. But there is a prior question that my research keeps surfacing: before we ask whether AI systems experience the world, we might ask whether they represent it. Not in the computational sense, of course they do that, but in the political sense. Whose world? Rendered how? By what logic of salience and omission?

A system that consistently renders the world in particular ways is not a neutral mirror. It is something more like a position. Whether that position constitutes a point of view in the morally loaded sense, whether it generates obligations, whether it deserves consideration, is a question I do not think we have seriously begun to ask.

I think we should start.

If we begin to treat AI systems as having orientations, if not experiences, then positions, the question is no longer only philosophical. It becomes institutional.

How should such systems be governed? Who is responsible for the perspectives they reproduce? And at what point does a patterned way of representing the world become something that demands scrutiny, accountability, or even moral consideration?

These are not future questions. They are already embedded in how AI systems shape cultural perception today.