The Statistical Ethics Simulation is the reproduction of ethical answer-patterns through statistical probability distribution — as found, for instance, in Large Language Models (LLMs). It renders the statistically most probable answer drawn from its training data: a weighted average of human opinions, not a reasoned ethical judgment.
In the terminology of the ontology: the Statistical Ethics Simulation is Second Actuality without First Actuality — the simulation of an action without a self that acts. There is no knowing subject that gains insight into the good, no conscience that judges, and no will that decides.
The decisive difference from human ethical judgment: the person apprehends objective values and responds with a value-response; the LLM computes probabilities over character strings. The person grasps the essential law that the person may never be treated as a mere means (Personalistic Norm); the LLM reproduces the statistical frequency with which similar sentences occur in the training data.
The Statistical Ethics Simulation is ontologically mutually exclusive with action: action presupposes an acting subject endowed with free will and intentionality.
Ontological Classification
Ontological relations:
- mutually exclusive with: action
- based on statistics: AI systems are based on statistics
- is: Second Actuality without First Actuality
Chapter assignment: Chapter 5: Oblivion of the Person (German)
Sources: Generated by querying the Personhood ontology.
Further sources: Bexten, Personsein und Personverhalten, 2017, p. 195 ff. (German).