The field of agent foundations (and AI alignment more generally) is often described as pre-paradigmatic. This refers to scientific paradigms in the sense coined by Thomas Kuhn. It means that there is no generally agreed-upon set of concepts, theories, methods etc that researchers use to make progress.
Despite this, there is a generally agreed-upon cluster of phenomena and problems in AI safety that the field of agent foundations is trying to find a paradigm for. This is why many of Dovetail’s research problems are described in terms of deconfusion or finding definitions.
