The knowledge question
To what extent should the pursuit of knowledge in artificial intelligence be limited by ethical considerations?
Artificial intelligence sits at a revealing intersection of the natural sciences and the human sciences. The methods that produce it, including statistics, optimisation, and computer science, belong to the natural sciences. Its consequences fall on people, institutions, and societies, the domain of the human sciences. This split is exactly what makes AI a productive TOK case study.
Discussion prompts
- If a model can be built, does it follow that it should be? Distinguish between what is technically possible and what is ethically permissible.
- Who counts as a "knower" when a system produces an output its own designers cannot fully explain? What happens to responsibility when knowledge is distributed across people and machines?
- Compare AI ethics with historical debates in the natural sciences: nuclear physics, genetic engineering. Is "knowledge for its own sake" a defensible position?
- How does bias in training data complicate the claim that AI produces objective knowledge?
Connecting the areas of knowledge
The natural sciences value prediction and replicability; the human sciences must account for meaning, consent, and power. A facial-recognition model may be accurate (a natural-science virtue) while being unjust (a human-science problem). TOK asks us to hold both lenses at once, and to notice that "it works" is not the same claim as "it is good."
A position to argue against
"Scientists discover; society decides what to do with it. Ethics is not the scientist's job."
Use this claim as a foil. What does it get right about the division of intellectual labour, and where does it break down when the knowledge in question can act in the world on its own?