What does it mean to have an experience? Two kinds of narrative world models

Agents construct world models-internal representations of their environment and their place in it-which they use to plan actions. A key subset is narrative world models, which organize actions and events into temporal episodes with defined beginnings and endings, linking these endings to emotional outcomes. While most narrative world models derive from completed past episodes, this article proposes a distinct subset oriented toward future and potentially transformative experiences. Such experiences have particular features: (i) Before the experience, the radical uncertainty about what is to come can trigger an 'all-hands-on-deck' call for our mental and cognitive apparatus, redirecting our attention to the unknown 'gap' that can operate as a black hole. Agents may mentally produce multiple, contradictory imaginary versions of future events. (ii) After the experience, agents may form preference-based, aesthetic or moral judgments about the experience that can reshape their worldview. In such a case, the experience can lead to self-updating. The paper discusses the challenges of these narrative world models, including their implications for artificial intelligence (AI). This article is part of the theme issue 'World models in natural and artificial intelligence'.