Brain Computer Interfaces
By Rolando Masis-Obando
Summary
Brain-computer interfaces (BCIs) are technologies that enable direct communication between the neural signals in the body and external devices. They can “read” biometric signals from the body and decode them into some action and can also “write” by stimulating nerves via electrical stimulation. When combined with AR and VR, BCIs offer exciting possibilities for enhancing user experiences and interaction. While they offer exciting possibilities and have been used in a plethora of situations to help disadvantaged people, BCIs pose many potential risks to both their clinical use and future widespread consumer adoption.
Existing Tech
Brain-computer interfaces come into two categories: invasive and non-invasive. Invasive BCIs require performing surgeries that give sensors access to the tissues and neurons in the brain. Non-invasive BCIs usually involve capturing neural signals via electrodes (EEG), magnets (fMRI, MEG) or optical devices (fNIRs). To date, invasive BCI is reserved for serious medical conditions with many years to follow before safe everyday consumer use. It is still an open question whether invasive BCI will be necessary for every-day use given that statistical and technological advancements in non-invasive BCI make noninvasive devices more affordable, accessible and consumer friendly. For the purposes of clarity, BCI here, will refer to any device that measures biometric signals and has the potential to perform an action from them.
Non-invasive BCIs include devices that can read neural signals from the brain, but also devices that can infer intention from muscles such as surface EMGs (sEMG). A notable example of sEMG used to control software is Meta’s EMG wristband (formerly called CTRL-Labs) which can detect and classify incoming signals from the brain to the hands and fingers.
Historically, the majority of BCIs have been dedicated to act as neuro-prostheses to restore motor function for people with severe motor disabilities including paraplegics and those with Amyotrophic Lateral Sclerosis (ALS). However, BCIs have also been used for a wide range of tasks, including meditation training, phobia and PTSD therapy, controlling computers, cell phones, and robots, among others.
Definitions
Brain-computer interface (BCI): also known as brain-machine interfaces (BMI), are devices that create a link between the neural activity in the brain with a computer that can process and interpret that neural activity into action. Most neuroimaging devices can be used as BCIs.
Electroencephalography (EEG): EEG devices can be used as noninvasive BCIs that measure the electrical activity in the brain using small electrodes attached to the scalp. They have high temporal resolution, often times collected over 1000 samples per second, but low spatial resolution (given the limited number of electrodes that fit around the head).
Magnetoencephalography (MEG): MEG can also be used as a noninvasive BCI that measures the magnetic fields produced by the brain’s electrical currents. The temporal resolution is high and has better spatial resolution than EEG.
Functional Magnetic Resonance Imaging (fMRI): Uses massive magnets to measure the concentration of blood in the brain to infer neural activity. It does this by detecting subtle changes in the magnetic field in active regions of the brain where oxygenated blood tends to flow. Spatial resolution tends to be high, while temporal resolution tends to be low.
Functional Near-Infrared Spectroscopy (fNIRS): Uses near infrared light to measure the concentration of hemoglobin in the blood vessels of the brain surface to infer neural activity. Spatial resolution is higher than EEG but lower than fMRI due to the penetration depth of near infrared light in brain tissue. Temporal resolution is higher than fMRI.
Electromyography (EMG): EMG measures the electrical activity of skeletal muscles via either injected needles or surface electrodes. EMG can act as a BCI by detecting the neural signals traveling down muscle fibers and interpreting them as intentions for movement.
Concerns
While there are BCIs that can both detect neural signals and stimulate regions of the brain, in general, most invasive and non-invasive BCIs are ‘read-only’. A “read-only” device only measures neural signals and is not capable of stimulating (i.e., sending a current) to alter the electric potential in a neuron or cluster of neurons. While the risks associated with “write” capable devices is much higher than “read-only” devices, due to the self-evident possibility of permanently damaging neural ensembles, there are still concerns that need to be addressed for “read-only” BCI.
Concerns associated with BCIs, whether they only read-out signals or are capable of writing them, range from ethical considerations associated with personal autonomy, legal ambiguities, privacy, security, health, and socio-cultural ramifications (King et al., 2021).
Personal autonomy risks involve BCIs that may alter the attributable agency to the user. In other words, there are potential risks that a BCI may reduce their free-will or ability to self-govern. Personal autonomy risks begin at informed consent, where a BCI representative may not fully disclose risks or may use tactics towards explicit or implicit coercion. Personal autonomy risks also include those once a BCI has been installed. For example, a BCI used for communication may verbalize insulting or unintended comments (King et al., 2021). Similarly, a BCI that allows for control of a robotic arm to slice ingredients with a knife may spontaneously initiate a stabbing action (whether through a spontaneous intrusive thought or malfunction) such that a user murders their partner. A situation like this leads us to legal ambiguity considerations.
Who is responsible for the final actions of a BCI? In the previous example, the BCI processed an action before the user was able to veto it. Criminal guilt in a situation like this is ambiguous. Who is held legally liable for BCI-facilitated actions? The users? The developers? The company that provided the BCI? Legal ambiguity considerations also include the risk of regulatory oversight to govern safe development and deployment into society (King et al., 2021). An example of this includes laws that may favor BCI-enhanced individuals for insurance purposes vs those that lack them.
Issues of privacy and security often come to mind when thinking about BCIs. There are many risks associated with personal neural data that is collected through BCIs that could be manipulated, controlled, broadcast or even sold without consent or legal protection. By nature of how neural signals in the brain are distributed across multiple regions, more information than what was intended to be read-out by a BCI can be collected. Additionally, due to the electronic nature of BCIs, the possibility that security breaches and/or malicious actors hacking devices becomes very real.
BCIs also pose risks to health from both a psychological and a physical standpoint. For example, irreparable physical damage could occur during or after installation. Malfunctions or lack of training that limit the ability to control the BCI could result in mental fatigue or distress.
BCIs also pose risks to socio-cultural norms and customs. For example, if invasive BCIs reach general consumer adoption, BCI enhancement may only be affordable by certain segments of the population leading to increasing socio-economic, and possibly, sociophysiological inequality.
It is important to note, that while a lot of these risks sound like they center around invasive BCI, they also apply to non-invasive BCI.
Mitigations
For any of the aforementioned issues, there is a growing field of study, namely the neurorights or neuroethics fields, which try to identify and create formulations to ensure the safe adoption of BCIs for individuals and society. However, in tandem with developing ethical frameworks, most mitigations will require legal regulation to ensure accountability and liability when systems go awry. Concerns of privacy are in the front and center of all BCI applications, especially given that these issues remain unresolved in non-BCI applications, such as those of social media, wearable technology, and AI. Because BCIs will and often do use machine learning and deep neural networks (i.e., AI) to decode neural signals the legal regulation that is currently being developed for safe deployment of AI should include language that extends its protection towards AI used with BCIs.
Papers:
Last updated