Ethical gradualism is the idea that whether an entity possesses morality is not a yes-or-no question, but rather has answers on a gradual scale. Our paper contributes to the field of machine ethics by replacing the often used, and variously construed, concept of computer "morality"with the more specific concept of "moral relevance". This we define as "the characteristic of having some connection to the moral domain". Our definition requires that an entity's perceived moral relevance can be obtained as an aggregate of multi-axial, continuous-variable moral characteristics such as, but not limited to, patiency, responsibility, and autonomy. These characteristics can furthermore be broken down into concrete sub- (and sub-sub-, etc.) characteristics which are easier to measure in the real world. This gradualist model of perceived moral relevance can then be practically implemented through translation into Web Ontology Language. We depict moral relevance both graphically as a class hierarchy, and in computer code. Our implementation allows computers to recognize perceived moral relevance in other computers. This provides a basic architecture by which computers can learn perceived ethical behavior only by acting with one another. The implementation also makes possible a new kind of experimental moral psychology, in which researchers can compare gradual perceived moral relevance directly between humans and computers.