Many recent neuroimaging studies claim to be able to read minds from brain activity, giving the impression that reductionist solutions to the mind-brain problem have been confirmed. I examine the challenges of a scientific approach to that problem and explore whether recent technological advances in neuroimaging offer any hope for solving it. I come to the conclusion that although some results obtained using artificial intelligence tools may look impressive, they are just imitations of mind reading. Moreover, all neuroimaging studies rely heavily on contrasting a state of interest with a control state and averaging this contrast information across repeated measurements within and between subjects, thereby abandoning any hope of discovering brain correlates of momentary subjective states of mind. I argue that attempts of mind reading from brain activity are based on a priori assumptions about the correctness of the mind-brain identity theory. An open-minded approach to the problem of the mind-brain relationship might begin with an empirical study of how brain activity itself is organized in a given behavior, and only then attempt to link this organization to the corresponding mental constructs.
Gennady G. Knyazev (Thu,) studied this question.