Goals, abstractions and value In neuroscience and machine learning, the term “value” is often used interchangeably with “reward”, focussing on its hedonic dimension while overlooking its functional, concept-like nature. Yet in everyday life we seldom receive clear numerical rewards at the end of our actions, and must instead construct our own values. In this talk, I will present work from my lab that challenges the canonical view equating value with reward, demonstrating how the brain can construct the value of an option or action on the fly by forming abstractions that flexibly adapt to changing goals. I will also share recent unpublished data showing how this mechanism shapes confidence, facilitates learning, and promotes generalisation. I will argue that rethinking the nature of value is both feasible and necessary. ON SITE & Zoom : https://unige.zoom.us/j/66873810837?pwd=TSEnpNm3PaAJoYLU3Mx9THb0LnzCO2.1 Meeting ID: 668 7381 0837 Passcode: 124349