Semantic Probing: Feasibility of using sequential probes to decode what is on a user’s mind

In this paper, we investigate the feasibility of using multiple sequential probe words to decode their relatedness to an active semantic concept on a users mind from the respective electrophysiological brain responses. If feasible, this relatedness information could be used by a Brain Computer Interface to infer that semantic concept, by integrating the knowledge of the relationship between the multiple probe words and the unknown target. Such a BCI can take advantage of the N400: an event related potential that is sensitive to semantic content of a stimulus in relation to an established semantic context. However, it is unknown whether the N400 is suited for the multiple probing paradigm we propose, as other intervening words might distract from the established context (i.e., the target word). We perform an experiment in which we present up to ten words after an initial target word, and find no attenuation of the strength of the N400 in grand average ERPs and no decrease in classification accuracy for probes occurring later in the sequences. These results lay the groundwork for the development of a BCI that infers the concept on a users mind through repeated probing.