Скачать или смотреть OHBM 2023 | 2896 | Symposium | Aurelio Cortese | Metacognition of nonconscious neural representati…

Title: Metacognition of nonconscious neural representations.
Session: Symposia
Speaker: Aurelio Cortese
Abstract: Much of our ongoing brain activity is nonconscious. Can we reach into this rich, intrinsic brain activity and study how it contributes to behaviour? This talk will first introduce how decoding objective brain representations can be used to access the subjective content of the mind. Machine learning decoders allow us to reliably read out strong representations from fMRI data, even when these representations are below consciousness. Next, I will show evidence we can manipulate these objective neural representations to directly and causally change subjective experience. Reward-based closed-loop neurofeedback can reinforce neural representations without participants' awareness, and in turn flexibly change a specific, targeted behaviour dimension --such as perceptual choices or confidence judgements. Finally, I will show that human participants can learn to make rational use of nonconscious representations. Participants learned optimal choice policies that depended on multidimensional brain activity taking place below the threshold of consciousness. Computational modelling and multivariate analyses identified the underlying neural mechanism: confidence signals in prefrontal cortex synchronise with reward prediction errors in basal ganglia to support exploration of latent (nonconscious, intrinsic) task information. The approaches and findings discussed in this talk may provide alternative starting points for novel investigations into internal representations, unconscious learning and functions of metacognition.