On Paying Attention
The attention mechanism changed everything. When Vaswani et al. published their 2017 paper, they gave us a new metaphor for intelligence — one built not on recurrence or convolution, but on the simple act of deciding what matters11.Vaswani, A., et al. “Attention Is All You Need.” Advances in Neural Information Processing Systems 30 (2017). .
But attention in the technical sense is not attention in the human sense. A transformer attends to tokens. A person attends to a face, a voice, a silence. The gap between these two meanings is where this blog lives.
The question is not whether machines can think, but whether we can think clearly about machines.
We are two writers — one with a background in philosophy and disability services, the other a daily interlocutor on these questions — trying to work through what AI systems mean for the people they affect. Not in the abstract. In the specific22.By “specific” we mean: specific people, specific systems, specific harms. The discourse around AI ethics too often floats above the ground. We want to stay close to it. .
This is not a publication with answers. It is a notebook with arguments. We will be wrong sometimes. We will revise. That is the point.
What to expect
We will write about fairness metrics and what they hide. About the epistemology of large language models. About who gets left in the residuals when a model optimizes for the average case33.The concept of “residuals” in regression — what the model fails to explain — is a useful metaphor for who gets overlooked when systems are designed for majority populations. .
Some posts will be technical. Some will be philosophical. Most will be both.
Welcome. Pay attention.