Social interactions are a crucial part of human life. Understanding the neural underpinnings of social interactions is a challenging task that the hyperscanning method has been trying to tackle over the last two decades. Here, we review the existing literature and evaluate the current state of the hyperscanning method. We review the type of methods (fMRI, M/EEG, and fNIRS) that are used to measure brain activity from more than one participant simultaneously and weigh their pros and cons for hyperscanning. Further, we discuss different types of analyses that are used to estimate brain networks and synchronization. Lastly, we present results of hyperscanning studies in the context of different cognitive functions and their relations to social interactions. All in all, we aim to comprehensively present methods, analyses, and results from the last 20 years of hyperscanning research.

Social interactions are a crucial part of human life. Understanding the neural underpinnings of social interactions is a challenging task that the hyperscanning method is trying to tackle in the last two decades. Here, we review the existing literature and evaluate the current state of the hyperscanning method. We review the type of methods (fMRI, M/EEG, fNIRS) that are used to measure brain activity from more than one participant simultaneously and their pros and cons for hyperscanning. Further, we discuss different types of analyses that are used to estimate between brain networks and synchronization. Lastly, we present results of hypercanning studies in the context of different cognitive functions and their relations to social interactions. All in all, we aim to comprehensively present methods, analyses, and results of the last twenty years of hyperscanning research.

Frozen transient imbibition states in arrays of straight cylindrical pores 400 nm in diameter were imaged by phase-contrast X-ray computed tomography with single-pore resolution. A semiautomatic algorithm yielding brightness profiles along all pores identified within the probed sample volume is described. Imbibition front positions are determined by descriptive statistics. A first approach involves the evaluation of frequency densities of single-pore imbibition lengths, and a second one involves the evaluation of the statistical brightness dispersion within the probed volume as a function of the distance from the pore mouths. We plotted average imbibition front positions against systematically varied powers of the imbibition time and determined the optimal exponent of the imbibition time by considering the correlation coefficients of the corresponding linear fits. Thus, slight deviations from the proportionality of the average imbibition front position to the square root of the imbibition time predicted by the Lucas–Washburn theory were found. A meaningful pre-exponential factor in the power law relating imbibition front position and imbibition time may only be determined after ambiguities regarding the exponent of the imbibition time are resolved. The dispersion of peaks representing the imbibition front in frequency densities of single-pore imbibition lengths and in brightness dispersion profiles plotted against the pore depth is suggested as measure of the imbibition front width. Phase-contrast X-ray computed tomography allows the evaluation of a large number of infiltrated submicron pores taking advantage of phase-contrast imaging; artifacts related to sample damage by tomography requiring physical ablation of sample material are avoided.

Despite significant advances in machine learning, decisionmaking of artificial agents is still not perfect and often requires post-hoc human interventions. If the prediction of a model relies on unreasonable factors it is desirable to remove their effect. Deep interactive prototype adjustment enables the user to give hints and correct the model's reasoning. In this paper, we demonstrate that prototypical-part models are well suited for this task as their prediction is based on prototypical image patches that can be interpreted semantically by the user. It shows that even correct classifications can rely on unreasonable prototypes that result from confounding variables in a dataset. Hence, we propose simple yet effective interaction schemes for inference adjustment: The user is consulted interactively to identify faulty prototypes. Non-object prototypes can be removed by prototype masking or a custom mode of deselection training. Interactive prototype rejection allows machine learning naïve users to adjust the logic of reasoning without compromising the accuracy.