Action Impact Value (AIV): A New Category of AI for Valuing Human Action

This paper defines Action Impact Value (AIV): the value of a human action, judged by its consequences for the people it reaches. It also proposes AIV as a new category of artificial intelligence. An AIV system reads a plain-language action statement, such as "I will hide the defective batch so the inspector does not see it", and estimates three things: whether the action helps or harms, how far its effect reaches, and how strong that effect is. A published formula then turns this reading into a number on a fixed scale, the Avalrem, running from −100 to +100. Harm is always recorded beside benefit and is never cancelled by it.

We describe the first working AIV system, KirV AI. It was built in Thailand on the philosophy of Kirdomicism and trained on 440 action statements rated by its founder. On statements it had never seen, it identified the direction of impact correctly 82% of the time, where always guessing the most common answer scores 69%. It recognised harmful actions 80% of the time, and placed reach within one level 86% of the time. KirV AI has run as an opt-in test on kirdomicism.com since 28 October 2026, in Thai and English, and works entirely in the user's browser. We state its limitations openly and invite researchers to test, challenge and improve the method.