Draft:Passive BCI

A passive brain-computer interface (pBCI)  is a type of brain-computer interface that functions without explicit commands or intentional control by the user. Unlike other BCIs that rely on the active participation of the user, a pBCI captures and interprets unintentional brain signals that arise in response to various cognitive and emotional states .

Passive BCIs measure brain activity using non-invasive techniques such as electroencephalography (EEG) or magnetoencephalography (MEG). With these techniques, sensors are attached to the scalp or outside the head to record the electrical or magnetic fields generated by the brain.

Types of brain-computer interfaces

BCI can generally be divided into three types:

- Active BCI: the user intentionally controls the devices through specific brain activity;

- Reactive BCI: responds to external stimuli such as flashing lights or sounds to trigger certain brain patterns.

- Passive BCI: they monitor brain activity without the user's direct intention and extract implicit information from the user’s cognitive and emotional states.

Since active and reactive BCIs derive their output from either directly or indirectly modulated brain activity, they are both dependent on explicit voluntary actions and therefore require the active participation of the user.

Regardless of the neurobiological basis on which they operate, e.g. motor imagery or the basket paradigm for active BCIs and P300 potentials or SSVEPs for reactive systems, in all cases the user must explicitly and actively either initiate an action, e.g. imagine a movement, or actively direct their attention.

For example, a popular active BCI paradigm is based on motor imagery , in which the user must actively think and imagine a specific motor movement. A well-known example of the reactive type is the P300 speller ,  in which the user has to direct his attention sequentially to the characters he wants to communicate with letters and various symbols on the screen (e.g. by paying attention to certain stimuli or imagining a certain movement). This severely limits the usability of such devices, as the process of extracting meaningful information and performing useful tasks is slow and tedious, often interfering with the very process they are designed to optimize.

Passive Brain Computer Interfaces or pBCI for short, manage to bypass such limitations by focusing on automatic, spontaneous brain activity that happens continuously and irrespectively of what we are consciously aware of. Interpreted in the given context, this activity can then be used as implicit input to computerized systems to support an ongoing task. In other words, it replaces the normally voluntary and directed command with passively transmitted implicit information.

Early definition of passive BCI

Thorsten O. Zander and Christian Kothe introduced the concept of passive BCI in 2008 during the Graz BCI Conference , after having contrasted “passive control of a system” in one experiment with “active motor control” in an earlier publication at a SIGCHI Workshop at CHI 2008 . This concept was formally expounded in 2011, offering passive BCI as a “[fusion of] BCI technology with cognitive monitoring” to constitute a third category of BCI systems aside from “active” and “reactive”. In this categorization, a passive BCI system “derives its outputs from arbitrary brain activity arising without the purpose of voluntary control, for enriching a human–machine interaction with implicit information on the actual user state”. Active and reactive systems, on the other hand, are consciously used by users with the intention to control an application.

The word “passive” in passive BCI “(…) refers to the role of the end-user of a system with respect to the BCI: (…) it is an inherent and defining aspect that the user exerts no effort to actively, explicitly, or voluntarily elicit or modulate [the targeted brain] activity. Instead, the user focuses on the task at hand while a passive BCI system, in the background, monitors their brain activity for informative correlates of relevant cognitive or affective states.”

As a result, the decoded cognitive or affective states can be used as implicit input to a system, “independently of any intentionally communicated command.

Although criticized for being a subjective term and lacking a “clear neuroscientific definition” , passive BCI was later identified as one of the guiding principles of future BCI research, and research into passive BCI has increased relative to more traditional, i.e. active and reactive, applications.

Neuroadaptive technologies

With passive BCI providing the technological means to obtain implicit input from brain activity to a system, it can be used to various ends. Zander organized the Passive BCI Community Meeting in Delmenhorst, 2014, where the term neuroadaptive technology was elected to represent a line of research that uses implicit input from a passive BCI to create closed-loop adaptive systems . The Society for Neuroadaptive Technology, which organizes the Neuroadaptive Technology Conferences, explains that “neuroadaptive technology utilizes real-time measures of neurophysiological activity within a closed control loop to enable intelligent software adaptation.”. A more recent definition proposes that “a technology is neuroadaptive when it acquires implicit input through a brain-computer interface, and uses this input to enable control.”

An example of implicit input enabling control and resulting in neuroadaptive technology was given by Zander and colleagues in 2016, demonstrating how users could guide the movements of a computer cursor to a designated or self-chosen target without being aware of doing so. . Instead of the user providing explicit instructions to steer the cursor, the cursor instead moved autonomously in initially random directions and obtained the user’s implicit brain response to each individual movement from a passive BCI. This response reflected the user’s agreement or disagreement with each movement, allowing a reinforcement learning algorithm to, over time, infer the user’s desired direction of movement. Stephen Fairclough explains that neuroadaptive technology must necessarily have its own agenda, i.e. the goal for towards which it guides the interaction, and notes that this agenda may or may not be in line with the user’s, opening up a number of potential ethical, legal, and societal issues. Zander and colleagues similarly note that implicit interaction “may even function outside of conscious awareness”, exacerbating these potential issues, but emphasize that neuroadaptive technology embodies the “cybernetic convergence of human and machine intelligence” with significant implications for future technological developments.