Formal Ethics Seed Ontology: Project Motivations
My main motivation in doing the formal ethics seed ontology is to facilitate clear discussions of ethics, values, and the choices involved. My experience attempting to discuss ethics with intelligent, educated people is that, often, they believe ethics is too vague, murky, and ill-defined to even begin discussing the interesting, tricky parts! Philosophers and ethicists cannot even agree on what ethics is, let alone on how we should approach ethics. Many approaches, such as preference utilitarianism obviously lead to contradictions, too. Nor is there agreement between realists, quasi-realists, and relativists as to what the nature of ethical statements is1See the discussion in my first worklog for further meta-ethical considerations.. Yet, when poring over encyclopedia pages on ethics in curiosity, there appears to be much agreement on the basic structures involved. So I wondered to what extent the core structures underlying ethical theory can be distilled in such a manner as to allow us to clearly discuss what we value and what to do about it.
In simple English, “ethics is the study of behavior and its value in the context of a society of beings for the purpose of guiding them.”
One ontological simplification is to abstract the normative hypothesis out from the formalisms of ethical statements. For example, “do not kill”, is a clear expression of a prohibition regardless of one’s meta-ethical stance. Is this a divine commandment? Or the result of a mathematical theorem about what agents in a society should do, thus providing a normative obligation that all rational agents should adhere to? Is this an emotivist “booh 👎”? One which is culturally shared? Is this a threat-backed command in a “might makes right” pragmatic moral nihilism? The ethical language and theories can be explored separately from explorations of why one should follow the ethical theories. If I can show that it’s possible to clearly disentangle these concerns, then ethical discussions and decisions may be easier to undertake.
An additional hypothesis is that the languages of deontology, virtue ethics, and utilitarianism can all embed each other. Thus debates between the various ethical paradigms are actually between different ethical theories. It’s important to debate on the appropriate level of discourse. Parfit argues in On What Matters that consequentialist utility optimization is precisely the kind of maxim that satisfies Kant’s categorical imperative. Further, the languages of the paradigms can easily be mixed up: a teleological definition of utilitarianism could hold that one has an obligation to do that which will maximize benefit for all and minimize pain, that is, the “paradigm” actually consists of a single top-level deontic duty. I hope that the greater formal clarity supports this hypothesis on the equivalence of ethical paradigms.
Given that perfecting the definitions of ethical paradigms and fleshing out elaborate theories is beyond what I can do in the time allocated for this project, I’ll be happy to provide a compelling proof of concept that clearly and precisely working with ethical questions is feasible!
I aim to keep the core ontology theory-neutral so that the same definitions can be used for multiple meta-ethical views and multiple theories. Those who believe in ethical normativity can present this as an additional conjecture to be proved (by examples, such as Fuenmayor and Benzmüller‘s proof of Gewirth’s Principle of Generic Consistency). Otherwise, the theories will be “free-floating” without a ground in some meta-ethical stance. One interesting consequence of this is that specifying problems such as, “the truthful LLM desideratum”, quickly suggests the challenge is technical.
Seeing Fuenmayor and Benzmüller’s work, I believe continuations of the formal ethics seed ontology project should probably be done in a well-developed ITP ecosystem such as Isabelle/HOL rather than in SUMO. However, the ontology and knowledge base of SUMO have provided many lessons and guidance on how to set up the core ontology.