Fig 1.
Architecture of the BBI and “20 Questions” Experiment.
In the experiment, two participants (an “inquirer” and a “respondent”) played a question-answering game similar to “20 Questions.” The respondent is given an object (e.g., “dog”) that is unknown to the inquirer and that the inquirer has to guess. The inquirer asks a question about the object by selecting a question (using a mouse) from questions displayed on a screen. The question is then presented visually to the respondent through a web interface. The respondent answers “Yes” or “No” directly through their brain signals by paying attention to one of two flashing LEDs (“Yes” = 13 Hz; “No” = 12 Hz). The BBI uses EEG to decode the respondent’s answer, and a TMS apparatus to convey the answer to the inquirer by generating a visual percept through stimulation for “Yes” and the absence of a percept for “No.” In the figure, the BBI system is highlighted in red.
Table 1.
Age, gender, and role of all the participants who completed the experiment.
Table 2.
An example of the sets of objects (first column) and associated questions used during the experiment.
Fig 2.
EEG Components of the BBI for the Respondent.
During the experiment, the respondent sat in front of a computer screen (A) on a comfortable chair, while EEG signals were recorded from the scalp using an active electrode cap (D). SSVEPs in the EEG signal from the occipital lobe were driven by the frequencies of two flashing LEDs, positioned on the left (“Yes” answer, 13 Hz: B) and the right (“No” answer, 12 Hz: C) side of the screen.
Table 3.
TMS stimulation intensities and sites for the five inquirers in the experiment. Intensities are expressed as percentage of the stimulator’s maximum output.
Fig 3.
TMS Components of the BBI for the Inquirer.
During the experiment, the inquirer sat in front of a computer screen on a BrainSight chair, with his or her head kept in place by a two-pronged head rest (A). A figure-of-8 TMS coil (B), connected to a MagStim Super Rapid2 stimulator (C) was positioned over the inquirer’s occipital lobe to deliver visual stimulation in accordance with the respondent’s answer. The head position and the coil position were carefully checked with the aid of a laser pointer (D).
Fig 4.
Different Measures of BBI Performance Across Conditions and Subjects.
(A) Mean number of objects correctly guessed by the inquirer over 10 experimental (red) and 10 control trials (black) across 5 pairs of subjects; chance performance is 0.125 (dotted line); (B) Mean area under the ROC curves (see Fig 5, below); chance performance is 0.5 (dotted line); (C) Mean number of bits transferred during the experimental and control conditions, using the mutual information criterion; chance performance corresponds to 0 bits (dotted line). In all figures, the grey lines represent the five pairs of participants. Note that in (A), two pairs near the center of the plot had identical performance, giving the appearance of only four lines.
Fig 5.
Performance of Each Pair of Subjects.
Each plot illustrates the performance of pairs of subjects in terms of Receiver Operating Characteristic (ROC) for the experimental (red dots) and control (gray dots) games. The first plot (A) depicts the performance of each pair as whole while the other two plots represent the individual performances of respondents (B) and inquirers (C) in each pair. Overlapping points are indicated by darker shades of red or gray.