Whether pondering a hypothesis, imagining alternative worlds, or even just thinking of a specific person, our thoughts give our mental life its rich, secret quality. Part of the promise of psychology is that this treasure trove might be opened by behavioural analysis, neuroimaging, or therapeutic introspection. Of course, any student will eventually learn to limit their expectations, as psychology cannot even define what thoughts are, let alone uncover the contents of consciousness. Or can it? In my talk, I will discuss how a novel methodology I call neuroadaptive interfacing uses modern neural network models (NNMs) in interaction with evoked brain activity to approximate an individual’s thoughts and subjective perception. Images, typically faces, generated by the NNM are shown to participants who are instructed to concentrate on visual features (hair colour, age, gender) or subjective features (emotion, attractiveness). Machine learning is used to classify whether images match the task’s target, which is then used to optimise the NNM towards representing the mental category. Then, by using the NNM’s generative capacity, completely new images are produced that can be directly tested to match individual perception. I will show the promising results of this methodology with regards to objective, semi-subjective, and finally truly individual aspects of perception. Furthermore, I will discuss how the work relates to cognitive neuroscience, with regards to theories of recognition and mental imagery. Finally, I will show how the work may impact the field of social cognition by providing a new way of estimating social perception and bias.
因線上會議室人數限制，線上演講僅開放系上師生參加。