#frAIday: Reflexivity in AI Systems

#frAIday hybrid
Professor Susanna Schellenberg will give the lecture in Galaxen, Umeå University. You can participate via link or join us at Umeå University. Welcome! 

Abstract
Reflexivity is a central element of human intelligence. It allows us to introspect on our thoughts and preferences, reflect on our plans, values, and actions, and engage in metacognition. Our capacity for reflexivity enables us to self-represent, have first-person thoughts, and evaluate our progress on projects while reconfiguring our approach as needed. The capacity for reflexivity is key to bringing AI to the next level. AI systems have already been developed here and there that manifest reflexivity. Indeed, the last few years have seen an explosion of work on self-location in robotics, metacognition in LLMs, developing the functional artificial equivalent of our human prefrontal cortex, and self-assessment and adaptation mechanisms in AI systems. Reflexivity is the critical capacity in each of these mechanisms. To date, however, no explicit attention has been directed at artificially replicating this central element of human intelligence. The manifestation of reflexivity in current AI systems is a side-product of other goals. I argue that, for the field to evolve, reflexivity needs to be put center stage.