Cognitive Chunks as Neural Activity : Is it Possible to see What you Think?

In this article we present some initial results from the use of Electroencephalograms and Machine Learning to examine the possibilities to find means of communication for people with severe multi-impairments that affect their ability to communicate with other people in their surroundings. The current results from a series of experiments are still tentative, but indicate some interesting possibilities for further development. In the experiments, informants (without any impairments) have been instructed to look at a screen where simple pictures have been displayed. During the experiment the raw data readings have been recorded and tagged with the corresponding picture that has been displayed at the same time. This data has then been analyzed through a Convolutional Neural Network. The preliminary result from the experiments is that it is possible to predict which picture the informant is looking at from a single reading of the brain activity with around 80% accuracy. There are further indications of that what we can observe in these patterns is not a representation of the pictures, but rather a representation of a conceptual ''chunk''. This possibility will be discussed to some end in the Conclusions.