Global laws & litigation, ?BCI (Brain-Computer Interface) technology 1) Select a country or a region from the privacy laws around the world: Indicate which laws are present in this area and th

NOVEMBER 2021

Understanding the Data Flows and Privacy Risks of Brain-Computer Interfaces

PRIVACY AND THE CONNECTED MIND

Authors

Jeremy Greenberg, Policy Counsel, Future of Privacy Forum Katelyn Ringrose, Policy Fellow, Future of Privacy Forum

Sara Berger, Research Staff Member and Neuroscientist, IBM Research Jamie VanDodick, AI Ethics Leader, Chief Privacy Office, IBM

Francesca Rossi, AI Ethics Global Leader, IBM Joshua New, Technology Policy Executive and Senior Fellow, IBM Policy Lab

Acknowledgments

The Future of Privacy Forum would like to thank the following individuals for their advice and expertise: Dr. Tamara Bonaci, Assistant Teaching Professor at the Khoury College of

Computer Sciences at Northeastern University; Dr. Laura Y. Cabrera, Dorothy Foehr and J. Lloyd Huck Chair in Neuroethics, Associate Professor, Center for Neural Engineering, The

University of Pennsylvania State University; and Dr. Peter Reiner, Professor of Neuroethics at the University of British Columbia.

Thank you to FPF Policy Interns: Samuel Adams, Noah Katz, and Hannah Schaller for their contributions to this paper. An additional thank you to IBM legal counsel, Ron Leviner, and

IBM Racial and Social Justice Scholar, Alex Baria, for their contributions to the paper, and to Guillermo Cecchi and Jeff Rogers from IBM for their suggestions.

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 1

Executive Summary ______________________________________________ 2

Introduction ____________________________________________________ 4

Part I: BCIs are Devices That Can Both Record and Modulate an Individual’s Brain Signals Through the Collection and Processing of Neurodata __________ 5

Part II: BCIs Provide Benefits and Present Risks in a Number of Sectors Including Health, Gaming, Employment, Education, Smart Cities, Neuromarketing, and the Military ____________________________________ 11

Part III: A Mix of Technical and Policy Solutions Can Mitigate Risks __________ 26

Conclusion ____________________________________________________ 32

Endnotes _____________________________________________________ 33

TABLE OF CONTENTS

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 2

EXECUTIVE SUMMARY

This report provides an overview of the tech- nology, benefits, privacy and ethical risks, and proposed recommendations for promot-

ing privacy and mitigating risks associated with brain-computer interfaces (BCIs). BCIs are com- puter-based systems that directly record, process, or analyze brain-specific neurodata and translate these data into outputs that can be used as visu- alizations or aggregates for interpretation and reporting purposes and/or as commands to control external interfaces, influence behaviors, or modu- late neural activity. While neurodata can take many forms, this report discusses “neurodata” as data generated by the nervous system, which consists of electrical activity between neurons or proxies of this activity. Personal neurodata refers to neurodata that is reasonably linkable to an individual.

BCI devices can be either invasive or non-invasive. Invasive BCIs are installed directly into—or on top of—the wearer’s brain through a surgical procedure. Today, invasive BCIs are typically used in the health context. Non-invasive BCIs rely on external elec- trodes and other sensors or equipment connected to or monitoring the body for collecting and modulating neural signals. Consumer-facing BCIs use various non-invasive methods, including headbands.

Some BCI implementations raise few, if any, pri- vacy issues. For example, individuals using BCIs to control computer cursors might not reveal any more personal information than typical mouse us- ers, provided BCI systems promptly discard cursor data. However, some uses of BCI technologies raise important questions about how laws, policies, and technical controls can safeguard inferences about individuals’ brain functions, intentions, moods, or identity. These questions are increasingly urgent in light of the many potential applications expanded use of BCIs in: › Healthcare – where BCIs could monitor

fatigue, diagnose medical conditions, stimulate or modulate brain activity, and control prosthetics and external devices.

› Gaming – where BCIs could augment existing gaming platforms and offer players new ways to play using devices that record and interpret their neural signals.

› Employment and Industry – where BCIs could monitor workers’ engagement to improve safety during high-risk tasks, alert workers or supervi- sors to dangerous situations, modulate workers’ brain activity to improve performance, and provide tools to more efficiently complete tasks.

› Education – where BCIs could track student attention, identify students’ unique needs, and alert teachers and parents of student progress.

› Smart Cities – where BCIs could provide new avenues of communication for construction teams and safety workers and enable potential new methods for connected vehicle control.

› Neuromarketing – where marketers could incorporate the use of BCIs to intuit consumers’ moods and to gauge product and service interest.

› Military – where governments are researching the potential of BCIs to help rehabilitate soldiers’ injuries and enhance communication.

This report focuses on the current privacy impacts of BCIs, as well as the data protection questions raised by realistic, near-future use of BCIs. While the potential uses of BCIs are numerous, BCIs cannot at present or in the near future “read a person’s complete thoughts,” serve as an accurate lie detec- tor, or pump information directly into the brain. It is important for stakeholders in this space to delineate between the current and likely future uses and far- off notions depicted by science fiction creators, so that we can identify urgent concerns and prioritize meaningful policy initiatives. We take seriously the concerns raised by futuristic potential developments and keep them in mind as we make recommenda- tions, but in this paper we focus on the immediately pressing need to address issues already faced and likely to be faced in the upcoming decade.

Although the report primarily focuses on the privacy concerns—including questions about the trans- parency, control, security, and accuracy of data— involving existing and emerging BCI capabilities, these technologies also raise important technical considerations and ethical implications, related to, for example fairness, justice, human rights, and personal dignity.1 These concerns are equally crit- ical and complex, so this report highlights where

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 3

additional ethical and technical concerns emerge in various use cases and applications of BCIs. Greater in-depth discussion of areas beyond privacy war- rant additional research and careful consideration, and we hope to turn to those issues in future efforts.

To promote privacy and responsible use of BCIs, stakeholders should adopt technical guardrails including:

› Providing on/off controls when possible— including hardware switches if practical;

› Providing users with granular controls on devices and in companion apps for managing the collec- tion, use, and sharing of personal neurodata;

› Providing heightened transparency and control for BCIs that specifically send signals to the brain, rather than merely receive neurodata;

› Designing, documenting, and disclosing clear and accurate descriptions regarding the accuracy of BCI-derived inferences;

› Operationalizing industry or research-based best practices for security and privacy when storing, sharing, and processing neurodata;

› Employing appropriate privacy enhancing technologies;

› Encrypting personal neurodata in transit and at rest; and

› Embracing appropriate protective and defensive security measures to combat bad actors.

Stakeholders should also adopt policy safeguards including:

› Ensuring that BCI-derived inferences are not allowed for uses to influence decisions about individuals that have legal effects, livelihood effects, or similar significant impacts—e.g. assessing the truthfulness of statements in legal proceedings, inferring thoughts, emotions or psychological state, or personality attributes as part of hiring or school admissions decisions, or assessing individuals’ eligibility for legal benefits;

› Employing sufficient transparency, notice, terms of use, and consent frameworks to empower users with a baseline of BCI literacy around the collection, use, sharing, and retention of their neurodata;

› Engaging IRBs and other independent review mechanisms to identify and mitigate risks;

› Facilitating participatory and inclusive community input prior to and during BCI system design, development and rollout;

› Creating dynamic technical, policy, and employee training standards to account for the gaps in current regulation;

› Promoting an open and inclusive research ecosystem by encouraging the adoption, where possible, of open standards for neurodata and the sharing of research data under open licenses and with appropriate safeguards in place. A similar open-skills approach could also be considered for a subset of direct-to-consumer BCIs; and

› Evaluating the adequacy of existing policy frameworks for governing the unique risks of neurotechnologies and identifying potential gaps prior to new regulation.

Key Terminology and Definitions

› Neurodata – Data generated by the nervous system,2 which consists of the electrical activities between neurons or proxies of this activity.

› Personal Neurodata – Neurodata that is reasonably linkable to an individual.

› Neurotech/Neurotechnology – Technology that collects, interprets, infers or modifies neurodata.

› Brain-Computer Interface (BCI) – Computer-based systems that directly record, process, or analyze brain-specific neurodata and translate these data into outputs that can be used as visualizations or aggregates for interpretation and reporting purposes and/or as commands to control external interfaces, influence behaviors, or modulate neural activity.

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 4

INTRODUCTION

Brain-computer interfaces (BCIs) are a prime example of an emerging technology that is advancing new areas of human-machine inter-

action. Today, BCIs are primarily used in the health- care context for purposes including: rehabilitation, diagnosis, symptom management, and accessibility. While BCI technologies are not yet widely adopted in the consumer space, there is a recent interest and proliferation of new direct-to-consumer neuro- technologies. The emergence of such technologies across various sectors poses numerous benefits and raises significant questions about user privacy.

When connected to the Internet,3 BCIs can be clas- sified as a type of wearable or implanted instrument within the Internet of Bodies, a network of devices connected to, and generating information from, the human body.4 Such communication has long been supported by various interfaces, from the keyboard and mouse to touchscreens, voice commands, and gesture interactions. As computers become more integrated into daily human experience, new ways of commanding computer systems and experienc- ing digital realities have gained in popularity, with novel uses ranging from gaming to education.

While BCIs offer benefits from improving patient health outcomes to providing more immersive and customizable education, training, and entertain- ment, the technologies raise many of the same risks posed by digital home assistants, medical devices, and wearables. New and heightened risks associ- ated with privacy of thought also emerge, resulting from recording, using, and sharing of a variety of

neural signals.5 According to a recent report, con- sumers list privacy and security as major concerns regarding neural interfaces, second only to product safety.6 Sometimes, BCIs must always be on in order to function properly—particularly in the health and medical context. Always-on tech can collect more information than users expect, particularly when individuals are not provided sufficiently clear and detailed notice prior to consent. This report explores how BCIs fit into the broader scheme of next-gen- eration interfaces, and suggests safeguards to mitigate potential privacy and security risks.

Because of the emerging-nature of BCIs, it is im- portant to consider both current and future-facing privacy and ethical risks based on technical capa- bilities, use cases, and the current understanding of neurodata. Along with identifying what neurodata and personal neurodata are collected by BCIs and what conclusions or inferences are drawn based on this data, it is equally important to specify what BCIs cannot achieve, especially given the current hype cycle surrounding technologies that can easily veer into unrealistic, sci-fi territory. At the moment, BCIs cannot read an individual’s precise thoughts, accurately determine whether someone is telling the truth or lying, or directly pump knowledge or skills into an individual’s brain or make someone “smarter.” While these capabilities could exist in the future and warrant discussion and debate, they are far attenuated from current realities. This report appreciates the importance of such discussions, but seeks to focus on the current—and likely, near- term—capabilities of BCIs discussed in this report.7

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 5

A. BCIs are Computer-Based Systems that Record, Modulate—or Both Record and Modulate—Electrical Brain Signals, Which Can Be Translated Into Outputs

BCIs are computer-based systems that directly re- cord, process, or analyze brain-specific neurodata and translate these data into outputs that can be used as visualizations or aggregates for interpreta- tion and reporting purposes and/or as commands to control external interfaces, influence behaviors, or modulate neural activity. BCIs can be broadly divided into three categories: 1) those that record brain activity; 2) those that modulate brain activity; and 3) those that do both, also called bi-directional BCIs (BBCIs).8 BCIs that record brain activity are more commonly used in the healthcare, gaming, and military contexts. Modulating BCIs are typically found in the healthcare context. For example, mod- ulating BCIs are used to treat Parkinson’s disease and other movement disorders by using deep brain stimulation to treat the rigidity, slowness, and resting tremors common in Parkinson’s patients.9 While BCIs technically refer to devices that directly record or modulate the brain, other related neu- rotechnologies indirectly record and modulate. One of the most successful examples of indirect stimulation is cochlear implants, which help re- store hearing and suppress tinnitus by modifying the information that is provided to a compromised auditory system.10 BBCIs, which both record and modulate, can be an especially useful rehabilita- tion tool for spinal injuries or strokes.11

B. BCIs Can be Invasive or Non-Invasive and Employ a Number of Techniques for Collecting Neurodata and Modulating Neural Signals

BCIs can be invasive or non-invasive.12 Invasive BCIs are installed directly into—or on top of—the wearer’s brain through a surgical procedure. To- day, invasive BCIs are used in the health context. For example, invasive clinical BCI implants have been used to improve patients’ motor skills.13 Inva- sive BCI implants can involve a number of different types of implants. An electrode array called a Utah array is installed into the brain and relies on a se- ries of small metal spikes set within a small square implant to collect or modulate brain signals. New innovations like neural lace and neural dust are meant to drape over or be inserted into multiple areas within the brain.14

Utah array. Image courtesy Wikipedia.

Part I: BCIs are Devices that Can Both Record and Modulate an Individual’s Brain Signals Through the Collection and Processing of Neurodata

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 6

Other prominent examples of invasive BCIs rely on electrocorticography (ECoG), in which electrodes are attached to the exposed surface of the brain to measure electrical activity of the cerebral cortex. ECoG is most widely used for helping medical providers locate the area that is the center of epi- leptic seizures. This detection helps facilitate more targeted medical treatment but does not constitute medical treatment itself.15

In April 2021, Neuralink—Elon Musk’s startup cen- tered around creating a minimally invasive BCI— released a video of a macaque monkey playing a videogame using an invasive BCI.16 Explaining Neuralink’s invasive BCI prototype, “in a lot of ways,” Musk said, “it’s kind of like a Fitbit in your skull, with tiny wires.”17 While the Neuralink de- vice is still in the prototype stage, the technology points to a possible future where invasive BCIs are used for commercial purposes, such as gaming, entertainment, education, and wellness. Today it seems unlikely that consumers would be willing to surgically implant a device into their brain for commercial enjoyment, cognitive monitoring, edu- cation, and other direct-to-consumer uses, but only time will tell whether invasive BCIs for commercial purposes will eventually become mainstream.

Unlike invasive BCIs, non-invasive BCIs do not require surgery. Instead, non-invasive uses of BCI-technolo- gy rely on external electrodes and other sensors for collecting and modulating neural signals.

One of the most prominent examples of a non-in- vasive BCI technology is an electroencephalogram (EEG)—a method for recording electrical activity in the brain, with electrodes placed on the surface of the scalp to measure the activity of neurons in the brain.18 EEG-based BCIs are common in the gam- ing space in which collected brain signals are used to control in-game characters and select in-game items. Another noteworthy non-invasive meth- od is near-infrared spectroscopy (fNIRS), which measures proxies of brain activity via changes in blood flow to certain regions, specifically changes in oxygenated and deoxygenated hemoglobin concentrations using near-infrared light.19 fNIRS is especially prominent in wellness and medical BCIs, such as those used to control prosthetic limbs.20

Other non-invasive techniques go beyond sim- ply recording neurodata by also modulating the brain, which is one reason the term “non-inva- sive” is fairly contentious, with researchers and scientists finding the line between invasive and non-invasive uses of BCIs difficult to draw. For example, can a device that modulates a brain in a closed-loop fashion—meaning that neurodata recorded by the BCI serves as an input in how the BCI stimulates the user’s neural signals—ever truly be non-invasive? What about a device that is not implanted surgically, but still carries the potential for stimulation? For instance, transcranial direct current stimulation (tDCS)21 and transcranial magnetic stimulation (TMS)22 are both used to modulate neuroactivity in various areas, including the frontal lobes. Researchers have proposed that these forms of stimulation may increase memory, and learning abilities; however, such claims are still under review.23 Non-invasive neurotechnolo- gies should not be equated to non-harmful tech- nologies—just because a device is not directly implanted to sit on or within the human brain does not mean that device does not pose unique health and other privacy and data use risks.24

An example of a non-invasive EEG-fitted BCI device.

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 7

BCIs are generally characterized by four components: 25

› Signal Acquisition and Digitization: involves sensors (e.g. EEG, fMRI, etc.) measuring neural signals. The device amplifies signals to levels that enable processing and sometimes filters collected signals to remove unwanted data elements, such as noise and artifacts. These signals are digitized and transferred to a computer.

› Feature Extraction: As part of signal processing, applicable signals are separated from extraneous data elements, including artifacts and other undesirable elements.

› Feature Translation: Signals are transformed into usable outputs.

› Device Output: Translated signals can be used as visualizations for research or care, or they can be used as directed instructions, including feedforward commands utilized to operate external BCI components (e.g. external software or hardware like a robotic arm) or feedback commands which may provide afferent (conducted inward) information to the user or may directly modulate on-going neural signals.

An example of these components can be found in the following figure.

human body. For instance, an electromyography (EMG) sensor is a neurotechnology that can be worn non-invasively as a wristband26 or inserted into the human body to indirectly record motor neurons and their electrical activity in muscles.27 Today this method is typically used to diagnose neuromuscular abnormalities, but future use cas- es point to using EMG for detecting an individual’s intent to move fingers and other appendages for operating virtual keyboards and other devices.28

While the focus of this report is technologies that record or influence neurodata from the brain, neurodata is also found throughout the nervous system (including from the spinal cord and periph- eral nervous system) and thus similar but non-BCI neurotechnologies are being developed that capitalize on these downstream signals. Other invasive and non-invasive techniques include indirectly collecting neurosignals sent from the brain with sensors placed on other parts of the

Brain Signals

Signal Acquisition Digitized Signal Processing Control Signals

Feedback

Device Command

EEG ECoG

Single Unit

Feature Extraction

Translation Algorithm

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 8

A Timeline of Interfaces 29

1924First Human EEG Recorded

1968

1973

2005

1998

1982

1973

1969

2019

2018

2016

2012

1952 First Voice Interface

First Virtual Reality Headset

First Successful Cochlear Implant

The Term “Brain-Computer Interface” is Coined

First Computer Mouse is Commercially Available

First Multi-Touch Touchscreen

First Invasive BCI That Produces High-Quality Signals

First Person to Control an Artificial Hand Using BCI

Paralysis Patients Control Robotic Arms Using BCI

First BCI to Restore Sensation to a Paralyzed Person

Signals from an Invasive BCI are Accurately Decoded Into

Text with an Error Rate as Low as 3% When Tested On

Vocabularies Up to 300 Words

BCI Provides Rudimentary Vision to a Low-Vision Patient

2021 A Paralyzed Man Uses a BCI to Type with His Thoughts

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 9

C. Recorded Neurodata Becomes Personal Neurodata When It is Reasonably Linkable to an Individual

Neurodata is data generated by the nervous system, which consists of the electrical activities between neurons or proxies of this activity. These neurons help carry out tasks, such as comprehen- sion, movement, and communication. Neurodata can be both directly collected from the brain, or indirectly collected from an individual’s spinal cord, muscles, or peripheral nerve in the form of a down- stream signal from brain activity or a preparatory signal prior to brain activity.

At times, neurodata can be personally identifiable when reasonably linkable to an individual or when combined with other identifying data associated with an individual, such as when part of a user profile. Personal neurodata is neurodata that could be reasonably linkable to a particular individual.30 The collection and processing of personal neuro- data can produce information related to an indi- vidual’s biology and cognitive state. Additionally, the processing of personal neurodata can lead to inferences about an individual’s moods, intentions, and various physiological characteristics, such as arousal. Machine learning (ML) sometimes plays a role as a tool for helping determine if a neurodata pattern matches a general identifier or particular class or physiological state.

Although identifying individuals based solely on their collected personal neurodata is likely a difficult proposition, such identification has been shown to be possible with relatively little data (less than 30 seconds-worth) within a lab setting,31 and some ex- perts believe that such identification is feasible if not today, then in the near-term.32 This possibility has implications for definitions pertaining to biometric data, as well as its permitted use. Personal neuroda- ta can vary in levels of sensitivity, as certain personal neurodata can reveal seemingly innocuous data leading to few, if any, inferences about an individual; health information associated with an individual; or provide insight into an individual’s private feelings or intentions. For example, a BCI might reveal what object a gamer intends to select in a video game,33 which may or may not be innocuous; infer that a truck driver is becoming less alert while driving,34 which could reveal an individual’s sleeping habits; or it could reveal whether a patient is depressed, information pertaining to their health.35

In the future, BCIs could progress into new arenas, recording increasingly sensitive personal neuroda- ta, leading to intimate inferences about individuals. Those arenas include transcribing a wide-range of a wearer’s thoughts into text, serving as an accu- rate lie detector, and even implanting information directly into the brain. These uses are still in the early research phases and could be decades from fruition, or perhaps never emerge.36

FUTURE OF PRIVACY FORUM | IBM | PRIVACY AND THE CONNECTED MIND | NOVEMBER 2021 10

D. Both Invasive and Non-invasive BCIs Pose Technical Challenges for Effectively Recording Neurodata and Modulating Neural Signals

Regardless of the technique used, recording and processing brain signals to derive usable neu- rodata is a technologically challenging process. Wired BCIs—typically associated with the clinical and medical context—include complex wiring that involves a prolonged preparation time before use, while wires limit user movements.37

Wireless BCIs avoid some of the hardware chal- lenges of wired BCIs, but present new challenges associated with battery life—especially in the case of health-related BCIs that are intended to be on and active for extended sessions—and device weight, comfort, and practicality.38 Other hard- ware challenges include the need for commercial non-invasive headsets to record small neural sig- nals through a physical barrier of hair, skin, flesh, and bone, all of which can interfere with the signals and add noise to the data. Meanwhile, invasive BCIs require expensive, high-risk surgery.39

Once signals are collected, the device must process and separate actionable nerve impulses from those that are created by passive activities,

including artifacts derived from the wearer’s mus- cle movements, eye blinking, and electrical activity from the heart. Sometimes this extra data is used in conjunction with BCIs for various purposes, but these artifacts often have to be removed for neu- rodata to be usable. Most neurodata derived via BCIs is noisy (especially in the case of non-invasive applications) and creating computer systems that can classify and remove noise is a complex and cumbersome undertaking.

Collepals.com Plagiarism Free Papers

Are you looking for custom essay writing service or even dissertation writing services? Just request for our write my paper service, and we'll match you with the best essay writer in your subject! With an exceptional team of professional academic experts in a wide range of subjects, we can guarantee you an unrivaled quality of custom-written papers.

Get ZERO PLAGIARISM, HUMAN WRITTEN ESSAYS

Why Hire Collepals.com writers to do your paper?

Quality- We are experienced and have access to ample research materials.

We write plagiarism Free Content

Confidential- We never share or sell your personal information to third parties.

Support-Chat with us today! We are always waiting to answer all your questions.