LAUSR

Brain-computer interface (BCI) research and development continues to grow. In particular, BCI patent applications have been increasing exponentially in a few recent years (Greenberg et al., 2021). The situation is, however, different for different kinds of BCI: invasive and non-invasive, active and passive, especially regarding possible use by healthy users. Invasive BCIs provide best performance, and even may provide access to early stages of motor decision formation, enabling faster interaction compared to usual input devices (Mirabella and Lebedev, 2017), but they are associated with high risk and cost, and will unlikely be available for healthy users in near future. Existing non-invasive BCIs have low bandwidth, speed, and accuracy, and this is why only passive, not active BCIs have been considered as a prospective technology for healthy users in the roadmap of brain/neural-computer interaction (BNCI Horizon 2020, 2015; Brunner et al., 2015). Passive BCIs are those that use “brain activity arising without the purpose of voluntary control” (Zander and Kothe, 2011). As they do not claim the user’s attention, their low speed of interaction can be acceptable (Current Research in Neuroadaptive Technology, 2021). In contrast, a user of an active BCI controls an application explicitly, via conscious control of his or her brain activity (Zander and Kothe, 2011)1. These BCIs have to compete with themanual input devices (keyboard, mouse, touchscreen) and emerging touchless alternatives (voice-, gestureand gaze-based), as playing the same role in human-computer interaction (HCI) (Lance et al., 2012; van Erp et al., 2012). Although some attempts were announced to dramatically improve performance of the non-invasive BCIs by advancing brain sensor technology (most noticeably, Facebook’s plans to enable fast text input “directly from your brain”—Constine, 2017), the electroencephalography (EEG) remains the only widely used technology and performance is still below from what is provided by electromechanical input devices. For example, the best reported average time of activation of a non-invasive asynchronous “brain switch” (a BCI requiring low false positive rate but enabling detection of only one discrete command) is about 1.5 s (Zheng et al., 2022). Moreover, while some non-medical active BCIs use well-established non-invasive BCI paradigms— the motor imagery BCI, the P300 BCI, the steady-state visual evoked potential (SSVEP) BCI and the code-modulated visual evoked potential (c-VEP) BCI—many projects rely on even less precise control based on learned changing EEG rhythms (Nijholt, 2019; Prpa and Pasquier, 2019; Vasiljevic and de Miranda, 2020). Due to low performance, active BCIs are still affordable mainly for people who cannot use other input, such as paralyzed individuals. Nevertheless, attempts to develop active BCIs for healthy people continue. In this Opinion, I briefly overview the application areas for which they are currently developed, then try to figure out what motivates these attempts, and what is the near perspective.