Introduction
Most progress in AI has come from engineering. More data, bigger models, faster chips. The hard questions now are different. What should a system care about? When does it actually know something, and when is it just predicting? Who is it working for?
Philosophers have argued about these questions for a long time, and they left behind some useful tools. Our bet is that those tools belong inside the system, as parts that do real work.
Philosophy should be a working part of the system.
Ethics as a Design Constraint
Today most AI ethics happens after the model is built. Teams add filters, write policies and run audits on something that's already finished. The values sit outside the system, like a fence.
We think they belong inside it. A consequentialist weighs outcomes. A deontologist keeps certain promises no matter what. A virtue ethicist asks what a decent person would do. Put those questions into the objectives and the tests, and good behaviour starts coming from the system itself.
Epistemology and Machine Knowledge
A language model can give the right answer without understanding anything. Epistemology, the study of knowledge, helps us say what's missing. Believing something is one thing. Having good reason to believe it is another.
We want systems that remember why they believe a claim, how strong the support is, and what would change their mind. People trust a system like that more, and it's easier to fix when it's wrong. Ideas like calibration and provenance came from philosophy long before anyone put them in software.
Pragmatism in Practice
Pragmatists judge an idea by what happens when you use it. For AI, that means asking whether the system helps real people get real things done where it's deployed. A good benchmark score is a start, and that's all it is.
A pragmatic system expects to be wrong sometimes. It treats every interaction as new evidence and changes course when an assumption fails. That's how a tool keeps getting better after launch day.
Toward Integration
No single framework covers everything. Ethics can't tell you what's true. Epistemology can't tell you what matters. So the job is to combine them, in systems that reason about values, knowledge and usefulness at the same time.
Philosophers can't do that alone, and neither can engineers. It takes a team that treats philosophy as building material, something you can test like any architecture or dataset.