Automatic Detection of Confidence Level of Examinees in Answering Multiple Choice Questions using P1000 Brain Signal

Although Multiple Choice problem solving tests are quite efficient, they lack the power to detect the condition where a subject correctly responds to an answer by a mere guess. This paper attempts to provide a novel solution to this problem by utilizing Electroencephalography (EEG) technique. The acquired EEG signals of the subjects participating in the experiments undertaken are fed to an eLORETA software to examine the most active brain regions responsible for this cognitive activity. From eLORETA solutions, it is inferred that the frontal and the parietal lobes have enhanced activation during the said task. The signals obtained from the specific brain regions are further analyzed to investigate the presence of one specific (yet novel) Event Related Potential (ERP), called P1000. We found that P1000 is elicited when a subject is fully confident regarding the correctness or incorrectness of the options to a given question. However, when a subject is in ambiguity and attempts to guess a response, no significant ERP component is released. A Convolutional Neural Network (CNN) based classifier is employed to classify the EEG response during selection of the alternatives to a question into 2 classes: confident and ambiguous on the basis of presence or absence of P1000 signal. The proposed classifier yields a classification accuracy of 98.34% during training phase and an accuracy of 94.54% during testing phase. Thus, the proposed method can effectively be utilized to detect subjects free from lack of confidence.