<<

- COMPUTER INTERFACES U.S. MILITARY APPLICATIONS AND IMPLICATIONS

AN INITIAL ASSESSMENT

ANIKA BINNENDIJK TIMOTHY MARLER ELIZABETH M. BARTELS Cover : Peter Soriano Cover image: Adobe Stock/Prostock-studio

Limited Print and Electronic Distribution Rights This document and trademark(s) contained herein are protected by law. This representation of RAND intellectual property is provided for noncommercial use only. Unauthorized posting of this publication online is prohibited. Permission is given to duplicate this document for personal use only, as long as it is unaltered and complete. Permission is required from RAND to reproduce, or reuse in another form, any of our documents for commercial use. For information on reprint and linking permissions, please visit www.rand.org/pubs/permissions.html. RAND’s publications do not necessarily reflect the opinions of its research clients and sponsors. R® is a registered trademark.

For more information on this publication, visit www.rand.org/t/RR2996. Library of Congress Cataloging-in-Publication Data is available for this publication. ISBN: 978-1-9774-0523-4 © Copyright 2020 RAND Corporation Summary points of , adversary access to new informa- tion, and new areas of exposure to harm or avenues Brain-computer interface (BCI) represents an emerg- of influence of service members. It also underscores ing and potentially disruptive area of institutional vulnerabilities that may arise, includ- that, to date, has received minimal public discussion ing challenges surrounding a deficit of trust in BCI in the defense and national security policy commu- , as well as the potential erosion of unit nities. This research considered key areas in which cohesion, unit leadership, and other critical inter- future BCI technologies might be relevant for the personal military relationships. Finally, we consider warfighters of tomorrow. It sought to explore the potential future U.S. government ethical and legal operational value of current and future developments responsibilities to an individual BCI operator, as well regarding man-machine neural , the as the implications that BCI technologies might have associated vulnerabilities and risks, and the policy on the ethical and legal responsibilities of that indi- levers that should be in place before the technology is vidual. These considerations should be incorporated deployed. into research and development (R&D) efforts early The project drew from reviews of relevant tech- in the process and may warrant dedicated a depart- nical and security literature and discussions with mentwide oversight mechanism as the technologies subject-matter to develop a July 2018 game continue to mature. convening technical and operational experts. The Overall, our findings suggest that as the U.S. mil- game tested the potential utility of a functional “BCI itary increasingly incorporates artificial toolbox” against two future tactical urban operations (AI) and semiautonomous systems into its opera- vignettes. Game results indicated that BCI tech- tions, BCI could offer an important means to expand nologies are likely to have practical use on a future and improve -machine teaming. However, battlefield, particularly as the pace and volume of precautions will need to be taken to mitigate vul- human-machine interaction intensify. Within the nerabilities to DoD operations and institutions and vignettes, participants anticipated that BCI capa- to reduce potential ethical and legal risks associated bilities could enhance the speed of communication, with DoD’s development and adoption of BCI tech- improve common situational , and allow nologies. Specifically, we recommend that DoD operators to control multiple technological platforms simultaneously. Participants noted that the pragmatic • expand analysis to illuminate operational utility of each BCI capability would depend largely on relevance and risks. This research developed its fidelity and reliability during combat. Of the capa- a systematic approach to evaluating potential bilities assessed in the game, direct brain-to-brain operational applications of BCI by pairing communication facilitated by BCI appeared to offer operational experience with technological the most transformative applications for operational expertise and incorporating a disruptive and use but also carried the most significant operational creative Red team of RAND Corporation and institutional risks. experts. New analytical approaches such as Our analysis also explored possible areas of risk this could supplement existing internal exer- associated with the development and application of cises to help ensure that operational needs and BCI combat capabilities. As with many new techno- risks, rather than just technical opportuni- logical developments, BCI may create new military ties, drive BCI development and identify new operational vulnerabilities, new areas of ethical and adversary threats. legal risk, and potentially profound implications • address the trust deficit. The game and for existing military organizational structures. In associated research highlighted the extent to particular, the report highlights potential operational which cultural barriers to BCI, particularly vulnerabilities associated with the development and among infantry service members, are likely adoption of BCI technologies by the U.S. Department to be high. Trust barriers could be miti- of Defense (DoD), including the potential for new gated through heavy vetting and testing in

3 noncombat scenarios, introduction to service members that already rely on machine tech- Abbreviations nologies, and an initial focus on noninvasive measures and medical applications. AI • collaborate and anticipate. Our research ARL Army Research Laboratory highlighted examples of where DoD seed funding yielded successful BCI break- BCI brain-computer interface throughs, and examples of emerging BMI brain-machine interface private-sector . Where possi-

BRAIN Brain Research through Advancing ble, future could leverage Innovative private-sector advances to the benefit of the U.S. military and, if carefully pursued, could CaN CTA and improve trust gaps within the military. As the Collaborative Technology Alliance commercial market develops BCI technol- DARPA Defense Advanced Research ogies, this will help establish its capabilities Projects Agency and shortcomings. Although BCI applications DNI direct neural interface are currently still in the basic-research phase, DoD U.S. Department of Defense development of other technologies by the military, including , AI, and big data EEG electroencephalogram analysis, will need to consider the eventual availability of BCI. EW electronic warfare • plan ahead for institutional implications. As FDA U.S. Food and Drug Administration the U.S. government prepares to incorporate HCI human-computer interaction BCI technologies into future military capabili- ties, it will require institutional to IoT of Things address new ethical and policy issues at each MMI -machine interface stage of the process, from R&D to operational

MOUT Military Operations in Urban Terrain application to veteran care.

N3 Next-Generation Nonsurgical Introduction NCI neural-control interface The 86 billion of the represent NESD Neural System Design humankind’s primary evolutionary advantage and, perhaps, an area of untapped potential. Currently, NIH National Institutes of our interact with the world through our bod- OODA observe, orient, decide, act ies, sending electrical currents through the nervous R&D research and development system to vocalize with our mouths, to type—or swipe—with our fingers, or to move bipedally tDCS transcranial direct current through space. What will happen when human stimulation brains are freed of their corporeal confines and TNT Targeted Training can control machines directly? Neurotechnological TTX table-top advances have already given quadriplegics the ability to perform basic operations in an F-35 simulator with their thoughts1 and have given scientists the ability to decode speech that subjects are imagining in their —albeit imperfectly. Eventually, our physical

4 bodies might become a constraint that could be cir- Human-Machine Teaming cumvented with appropriate neurotechnology.2 The The research views BCI in the context of the antic- technical means for this brain-body bypass are BCIs, ipated future of warfare, including increases in defined as methods and systems for providing a human-machine teaming. The analysis begins from direct communication pathway between an enhanced the premise that human-machine teaming will or wired brain and an external device, with bidi- a major role in future combat and that BCI may rectional information flow (between the brain and provide a competitive advantage in future warfare. a device).3 Their potential impact is broad and far Former Deputy Secretary of Defense Robert Work, reaching, and policies on how to develop and manage who led DoD’s 3rd offset, a catalyst for defense-sector such technology should be proactive, not reactive. technology development focused on human-machine BCI technology is progressing. Such progress teaming, summarized trends with military technol- highlights the need to assess current and poten- ogy as follows: tial applications, and to ensure that the technology responds to actual needs in addition to the The coin of the realm during the Cold War of developers. As BCI transitions from basic research was armored brigades, mechanized infan- to more operational and commercial applications, try brigades, multiple launch rocket system it will be important to devote early to the battalions, self-propelled artillery battalions, tactical fighter squadrons, among others. Now, broader implications, to consider what policies and the coin of the realm is going to be guidelines might maximize its benefits while mitigat- machines and human-machine , ing potential downsides. Developing such technolo- which allows machines to allow to gies as AI, data analytics, and robotics have captured make better decisions; assisted human opera- headlines and fostered public discussion regarding tions, which means bringing the power of the potential benefits and risks. Limited comparable network to the individual; human-machine conversation has, as of yet, evolved for BCI. When combat teaming; and the autonomous compared to other prominent , network.4 BCI is relatively immature; few capabilities have been Although DoD R&D efforts include many deployed commercially. However, it has the potential dimensions of technology, certain aspects are to be no less influential. With profound potential particularly relevant to potential work with BCI. implications in fields from defense and national In particular, there is an increasing focus on security to health and wellness, BCI may represent human-machine collaboration for improved deci- a highly disruptive technology that, to date, has sionmaking, including human-computer interac- received insufficient analysis. tion (HCI) and cognitive teaming, assisted-human This report offers an initial assessment of what operations, and advanced manned and unmanned viable applications BCI may have in U.S. military combat teaming.5 Defense officials discussing the operations, and what risks and vulnerabilities may “centaur” model—collaborative human and AI be associated with its development and deployment. teams—have highlighted the relative advantages of The authors describe the current state of the art and the U.S. civilian and military workforce in develop- possible areas of technology development and growth ing and operating human-machine teaming technol- for BCI military applications and investigate key ogies.6 Human-machine teaming technologies that questions associated with the use of BCI capabilities effectively leverage the unique cultural strengths of in a future combat scenario. Fundamentally, we ask, the American warfighter, including critical think- (1) what is the potential operational significance of ing and creative problem-solving, represent an current and future developments regarding BCI, area of particular value to the future U.S. military.7 and (2) what are the policy considerations necessary Technology development programs within DoD have for effective management of the technology with an thus focused on human-machine collaboration.8 In of its potential impact on the warf- fact, the Defense Advanced Research Projects Agency ighter of the future?

5 Approach Ultimately, humans This analysis is structured to explore the operational implications of BCI technology. It pilots a repeatable and machines could process for systematically exploring the relevance and implications of emerging technologies in the collaborate cognitively context of military operations. Mapping techno- logical capabilities to practical applications—and and seamlessly—to understanding not just the state of the art but the state of the practical uses—can present a significant think together. challenge for assessments of emerging technology. Our process addresses that challenge by parsing the technology into operationally relevant capabilities, (DARPA) has suggested that “smart systems will testing their operational relevance with a table-top significantly impact how our troops operate in the exercise (TTX), and then exploring implications, future, and now is the time to be thinking about what risks, and risk-mitigation strategies. The process is human-machine teaming will actually look like and itemized as follows: how it might be accomplished. . . .”9 Future developments with human-machine 1. Through literature review and discussions teaming have the potential to prompt far-reaching with subject-matter experts, we summarize defense policy debates.10 Technological advances, the technology and identify key areas of devel- such as BCI, that allow humans to connect increas- opmental effort. ingly closely with machines on the battlefield 2. Specific topics of technology development are may yield fundamental strategic and operational used as a catalyst for discussions with military changes within each of the U.S. service branches experts to identify potential military applica- and will undoubtedly raise ethical and organiza- tions in theater. tional questions across the U.S. defense community. 3. We aggregate the results of the analysis Thus, as DoD pursues a forward-looking vision concerning technology development and for human-machine teaming, ambitious operational relevance to provide a general should prompt similarly far-reaching defense policy assessment of whether BCI could potentially debates.11 be valuable in a military setting and, if so, In preparation for a future world in which how. This preliminary finding becomes the human-machine teams represent the “coin of the overarching guide for a TTX. realm,” DoD has already invested in the development 4. Based on the technical and operational of technologies that can permit the human brain to assessments, we derive a of projected BCI communicate directly with machines, including the capabilities—a future BCI toolbox. development of implantable neural interfaces that 5. The BCI toolbox is used in a table-top game are capable of transferring data between the human based on exemplary scenarios involving brain and the digital world.12 On the future battle- tactical urban operations. This game explores field, human may well be channeled to AI more extensively the anticipated viability and software or to robots, with information transferred relevance of BCI in theater. back from sensors and machines directly to the 6. Throughout this process, we explore vulner- human brain.13 Ultimately, humans and machines abilities, risks, and risk-mitigation strategies could collaborate cognitively and seamlessly—to associated with BCIs across three dimensions: think together. technological, institutional, and legal/ethical.

6 Technology Summary humans to manage increasingly complex systems and information. BCI can improve such efficiency. Introduction Although human-machine teaming can be Although a comprehensive review of the field of BCI useful, it is just a subset of BCI applications. The lit- would be beyond the report’s scope, and although the erature regarding ongoing R&D and potential appli- literature on the topic is extensive enough to support cations extends beyond human-machine teaming, handbooks,14 dedicated societies,15 and dedicated and BCIs need not just link humans and machine journals,16 this section highlights prominent work, to provide value. More generally, BCIs provide a themes, and organizations to inform the assessment method to connect to the human brain. They pro- of potential military applications. By drawing on vide more data. This connection can then be tied to technical and popular literature, as well as discus- a machine, to software, to another human, or simply sions with subject-matter experts, we segment the to an output system for assessment. In fact, while topic into distinct areas of work, clarifying what human-machine teaming remains a cornerstone exactly BCI entails and what kind of research is of developing technology for warfare, the broader active. The section ends with a discussion of future advantages of BCIs point toward not just integrating trends and potential directions and summarizes humans and machines but leveraging human capa- primary technical challenges and risks. bilities in general. While we adopt the term BCI based on its prev- alence in the literature, a variety of related terms are Review used to describe similar capabilities: neural-control interface (NCI), mind-machine interface (MMI), While the practical significance of BCI has just direct neural interface (DNI), and brain–machine recently become more visible, work in this field has interface (BMI). been ongoing for nearly a century. In fact, work with Although there is some debate regarding the the first human electroencephalogram (EEG)—a precise nature of signals that are transmitted within device for tracking and recording brain wave the human brain, BCIs generally involve monitoring patterns—was published in 1929.18 Jacques Vidal or affecting those signals.17 Different BCI tools allow coined the term brain-computer interface in 1973, users to access and to use these signals with various and research in this field has continued since.19 levels of accuracy and invasiveness. In short, a BCI Work with BCI tends to fall into the follow- enables a bidirectional communication between a ing categories, which provide a framework for our brain and an external device, and there is a broad investigations of operational relevance and applied range of ongoing work on this topic. In this context, capabilities in subsequent sections: bidirectional generally includes direct neural readout • data transfer from the brain and feedback and direct neural write-in. • direct system control As reflected by the aforementioned definition • prosthetics and paralysis treatment and by the alternate terms, the focus is often on • cortically coupled AI (for training or running human-machine teaming, which aligns with needs AI systems) stemming from trends in warfare. Practically, BCI • data transfer to the brain, and brain-to-brain provides a mechanism for blending human strengths communication. and computer strengths, and much of the ongoing Each of these topics can be segmented further into work strives to link these two sets of capabilities and work involving invasive systems and noninvasive sys- yield synergistic advantages. The efficiency of the tems. Invasive systems involve implanting electronic interface between humans and machines—whether devices beneath the human skull, inside the brain. facilitating communication by screens, text, or The surgery allows practitioners to place the another form—is a significant factor in allowing exactly where desired to monitor precise sets of neu- rons that govern specific neurological functions, but it

7 carries health risks. Alternatively, noninvasive systems example, DARPA, APL, and the University of sit outside the skull. While this reduces risk to the user, Pittsburgh used a BCI implant to permit a quadriple- the skull essentially acts as a filter and muffles the gic woman to operate flight simulators.28 Researchers electrical signal.20 The signals picked up by external have also corrected robot mistakes through noninva- electrodes are less clear, and it is more difficult to sive measurement of EEG signals.29 ascertain which neurons are firing. Other research projects focus on drone control. A key effort in fueling most of these topics With funding from DARPA and the U.S. Army, has been the Brain Research through Advancing researchers at the Human-Oriented Robotics and Innovative Neurotechnologies (BRAIN) initiative at Control laboratory enable a user to control a swarm the National Institutes of Health (NIH). This effort is of drones.30 The lab’s researchers suggest the technol- broadly aimed at “revolutionizing our understanding ogy could be used practically in the military within of the human brain.”21 BRAIN partners include the five to ten years. Applications also include delivery National Science Foundation, DARPA, the U.S. Food of medical help, search and rescue, and exploration, and Drug Administration (FDA), and the Intelligence all in remote or inaccessible environments. Finally, Advanced Research Projects Agency (IARPA), as well using commercial hardware, researchers at the as foundations, institutes, universities, and indus- University of Florida have constructed, and demon- tries. NIH allocated $46 million in 2014 and strated the use of, a low-cost system that is capable of $81.4 million in 2015, reflecting growing interest in wirelessly controlling common small drones.31 the topic.22 Overall, DARPA has invested “hundreds Beyond military applications, the health-care of millions of dollars” transitioning into sector has advanced significant work with BCI direct neurotechnology since the early 2000s.23 control, especially involving invasive systems. Work With regard to the topic of transferring data at enables paraplegic patients to from the human, a primary goal is assessing cogni- control a computer mouse and computer software tive performance. To this end, the Army Research with their thoughts.32 Dr. Krishna Shenoy, one of Laboratory (ARL) is using 3-D printing to create the principal investigators, suggests, “The day will helmets that fit perfectly to each user and then come—closer to five than 10 years from now—when incorporate EEG sensors to monitor brain activi- a self-calibrating, fully implanted wireless system can ty.24 The Air Force is also pursuing a camera-based be used without caregiver assistance, has no cosmetic comprehensive cognitive monitoring system built impact and can be used around the clock.”33 into a pilot’s helmet to monitor cognitive workload The integration of prosthetics, which is essentially and stress.25 The helmets can adapt displays based a subset of the work with direct systems control, has on the pilot’s unique physical and mental condi- direct applications in the health-care sector. Much of tions. Separate research sponsored by ARL has this work involves invasive systems, primarily because investigated deep learning solutions for predicting of the necessity to target specific sets of neurons with drowsy and alert states based on EEG readings, and relatively high accuracy. Perhaps the most complex DARPA-sponsored teams have tested “closed-loop” prosthetic, conceptually, is the . Researchers brain implants that use to detect patterns have used electrodes to reconnect the motor cortex associated with mood disorders.26 BCI devices to and the spinal cord in monkeys and rats, restoring monitor performance and even emotional spikes the ability to walk.34 Case Western Reserve University related to , , or rage are increasingly used a similar procedure, called functional electri- prevalent in , where uses include factories, cal stimulation, to enable arm and hand movement public , state-owned companies, and the by bypassing the spinal cord and directly inducing military.27 muscle stimulation.35 As part of the Revolutionizing Related to systems for extracting data from the Prosthetics program, DARPA and APL have pursued human brain for assessment are methods for direct research concerning BCI-enabled prosthetic hands system control, whereby users control machines and arms that use intracortical microstimulation to wirelessly with brain activity. In one well-publicized provide feedback to the user (directly to the brain) and

8 evoke sensations that the user perceived as coming from his/her own hand.36 APL extended this work to study how BCI users could interpret external artifi- Noninvasive transcranial cially provided intracortical microstimulation, even if the type of information being provided varied from direct currents can what a brain region would normally have been pro- cessing. They then used this approach to provide a BCI be used to treat user with navigation information from a simulated Mooney Bravo aircraft.37 depression and , Data (or information) from a human brain can to increase focus and be used not only to inform assessment tools or to drive systems but also to inform software with cor- attention, to shorten tically coupled AI. Rather than using brain signals to control a computer or a system, a “cortically coupled training time, and, computer system opportunistically the brain state, capturing a user’s implicit or explicit computa- potentially, to improve tion, and then communicates this information to a traditional computer system via a neural interface.”38 physical training. This information can then potentially help train an AI system. This use of BCI represents a heightened prosthetic hand is without seeing it, unless they can level of human-machine teaming, allowing a human look directly at it to obtain visual feedback. to think with a machine (or a computer) or, more Research at Wake Forest Baptist Medical Center specifically, integrate human thoughts or data into and the University of Southern , funded a process conducted by machine. Real-time BCI by DARPA’s Restoring Active program, has interaction could negate the current requirement for shown initial success improving memory using surgi- predetermined computer codes to transfer informa- cally implanted electrodes, a strand of research with tion, addressing one of the key bottlenecks of tradi- promise for treating Alzheimer’s disease, strokes, tional human-machine integration.39 Such teaming and head injuries.41 By reinforcing recorded neural is of particular interest to members of the AI com- patterns from a patient’s experience (i.e., seeing a munity who are exploring methods and approaches particular image), researchers were able to improve for managing and “controlling” AI. BCI may be episodic memory, which is the most common type able to provide this tool. , the founder of memory loss in people with Alzheimer’s disease, of , a relatively new company focused on , and . Researchers saw a 35-percent integrating humans with AI, suggests, “Some high improvement in subjects’ baseline short-term memo- bandwidth interface to the brain will be something ry.42 Researchers noted, “This is the first time scien- that helps achieve a symbiosis between human and tists have been able to identify a patient’s own brain machine intelligence and maybe solves the control cell code or pattern for memory and, in essence, problem and the usefulness problem. . . .”40 ‘write in’ that code to make existing memory work In addition to the extraction of data from the better, an important first step in potentially restor- brain, there is also work exploring the ability to ing memory loss. . . .” While this work focused on implant or transfer information to the brain itself. A improving existing memory skills, future work might significant challenge in enabling efficient system and enhance the ability to retain specific as prosthetic control is feedback to the user, provid- memory skill begins to fail. Noninvasive transcranial ing information about the system being controlled. direct currents can be used to treat depression and Although patients in a lab setting may be able to strokes, to increase focus and attention, to shorten control a prosthetic hand using a BCI, for exam- training time, and, potentially, to improve physical ple, they will not necessarily realize where their training (with a focus on the motor cortex).43

9 authors extend this effort to involve five groups of three individuals.48 Two individuals sent informa- A natural extension tion, and the third person received information, with all three collectively playing a Tetris-like game. The from research that aims work follows earlier in transferring sig- to read brain signals nals between rats49 and from a human to a rat.50 and to send or implant Development Directions and Technical information in the Challenges Development Directions brain is brain-to-brain In general, the direction with most BCI work con- cerns the amount and quality of data being trans- communication. ferred. The fidelity with which data can be extracted and transferred from the human brain will likely increase.51 Signal bandwidth will likely improve. The DARPA System Design DARPA’s NESD program, for example, has invested (NESD) program is developing invasive systems in research on implantable neural interfaces for sen- that can communicate clearly and individually with sory restoration that may engage up to one million any of up to one million neurons in a given region neurons at once.52 As of July 2019, one of of the brain, and this includes the ability both to the grant recipients, led by Brown University, pre- transmit to the brain and read from the brain with sented 0.25-square-millimeter implants, called some neurons.44 While current invasive devices may “neurograins,” that permit wireless bidirectional incorporate something on the order of 100 channels, communication with an external device with an this project strives to read 106 neurons, write to 105 uplink rate of up to 10 megabits per second.53 As neurons, and interact with 103 neurons full-duplex, of January 2020, another grantee, Paradromics, a far greater scale than is possible with existing had announced a new high–data rate implantable neurotechnology.45 Another DARPA program, the BCI that can process and transmit neural data at Next-Generation Nonsurgical Neurotechnology (N3) 60-times-lower power dissipation than existing program involves a noninvasive system capable of approaches, allowing transfer of more data at lower reading from and writing to multiple points in the risk of overheating the brain.54 brain at once.46 Although no -of- currently A natural extension from research that aims to exists for this technology, one potential frontier in read brain signals and to send or implant information transferring data from the human brain may be in the brain is brain-to-brain communication. With long-distance standoff wireless assessments. These funding from the ARL, researchers at the University might allow commanders to assess their soldiers’ of Washington conducted a pilot study for a noninva- state or even the enemy’s state from long distances. sive system that uses EEG to read basic brain signals, Similarly, BCI might find use in aggregating the transmit them over the internet, and then transfer assessment of a group. For example, BCI could be motor responses to a second user using transcranial used to monitor the cognitive workload of a squad. magnetic stimulation.47 The signals represent very For direct system control, more work will be needed basic actions in the context of a simple video game, to transfer complex manipulations or strategies with such as move left or right. Nonetheless, especially resistance to distractions. given that these signals are transferred over the inter- With respect to prosthetics, the next step net, the potential to send even basic thoughts across is establishing new neurological connections. the internet inherently presents many opportunities Currently, BCI efforts regarding prosthetics involve and many risks concerning security and ethics. The

10 reconnecting existing neurons to physical systems. It these challenges by developing a portable noninvasive is more challenging to provide the ability to control system capable of reading from and writing to multi- a prosthetic (and associated neurons) that never ple points in the brain at once.63 existed. In addition, more work is also needed to Invasive systems, which provide higher-fidelity provide proprioceptive feedback directly to the brain signals, carry risks associated with any surgery, to improve bidirectional BCI control of prosthetic including hemorrhaging, infection, or brain dam- limbs. For data transfer to the brain, the ultimate age. Electrodes can also induce infections and can goal is to provide the brain with direct, high-fidelity degrade with time. Scarring and exhaustion (when information (e.g., isolated memory implants). Like neural substrates stop reacting) reduce signal long-distance standoff assessments, this capability is strength. Biocompatibility is also a significant limita- not yet feasible but nonetheless remains a target for tion. Furthermore, all current implants corrode and, the BCI field. thus, limit the duration for which they are useful. A particularly interesting prospect for BCI is its Sensors for most implanted BCIs currently last only integration with the (IoT), which about two to five years, although some primate work connects systems via the internet. DoD believes IoT involves sensors receiving signals for as long as seven can contribute to improved readiness by allowing to eight years. Reducing the sensor rigidity, size, and one to monitor the status of materiel and weapons tendency for deterioration while retaining signal systems in real time, and it is thus becoming perva- quality remains a persistent challenge. Developing sive.55 IoT has tactical applications, including giving sensors with additional channels to improve accuracy warfighters access to sensors and data, and BCI could and minimizing sensor power usage, which can cause enhance this ability.56 tissue damage, are additional challenges. Precise Some argue that, ultimately, the direction of placement of sensors on the brain is also challenging. this technology will follow the market.57 With the In general, the hardware necessary for BCI (amplifi- increasing focus on entrepreneurship, researchers ers, cables, sensors, etc.) is still too large for practical and academicians may likely spin off new technol- use outside a lab. ogy, and commercial entities will drive development With both invasive and noninvasive systems, based on market demand. To be sure, companies like the data gathered from neurons are then analyzed, Kernel,58 Neuralink,59 Paradromics,60 and Facebook61 and accurate decoding presents another challenge are actively pursuing BCI capabilities. A scenario with the advancement of BCI. This decoding often whereby commercial industry dominates this space involves some form of machine learning for perfor- can have two modes. Industry can, of course, respond mance assessment, which breaks down if the indi- to needs that pull the technology, but industry may vidual, task, or time frame changes. Furthermore, also push technology according to anticipated prof- decoding algorithms are unstable and require regular itability. It may be advantageous to have the relevant recalibration, in part because the position of neurons policy in place before the marketplace drives the relative to electrodes changes, and firing patterns technology. naturally change. To be sure, machine learning is an active area of research, including methods for gener- Technical Challenges and Risks alizing results such that an trained on one Despite the exciting and dynamic future that BCIs set of data may be used with acceptable accuracy on a may uncover, there are, of course, technological chal- new, slightly different set of input data. lenges and risks. Perhaps the most significant techno- Despite the aforementioned technical risks that logical challenge in BCI development is the trade-off must be addressed, the applications and underly- between signal clarity and the ability to target ing capabilities currently being explored with BCI specific neurons provided by invasive systems, and suggest that BCI may have viable uses for warfare. the ease of use with noninvasive systems.62 DARPA’s Subsequent sections thus consider operational reper- N3 program is currently seeking to address some of cussions of using BCIs in a military setting.

11 Operational Considerations those of recent decades. However, as a 2019 Royal United Services Institute report on the future oper- Introduction ational environment concludes, counter-Western Military theorists and practitioners have long lauded strategies are “dynamic and evolving” and “dif- the human mind as a critical determinant of military fer depending on the context, means, ways, and success. Could direct linkage between the brain and ground.”67 an external device improve a warfighter’s perfor- At the operational level, the character of warfare mance? Building on the overview of emergent BCI and combat experiences of the human warfighter technologies in the previous section, the following may be driven in large part by rapid advances in mil- section begins to explore how BCI technologies might itary technological innovation.68 As both the United be applied within a military context. States and potential adversaries develop and deploy Two lenses frame the discussion. The first con- new battlefield technologies, collaborative relation- siders relevant features of the future warfighter expe- ships between humans and machines are likely to rience to extrapolate potential applications for BCI. evolve and place new requirements on the cognitive The second draws from open-source material from workload for a future warfighter. Regarding the U.S. military research organizations and discussions potential application of BCI, the future warfighter is with subject-matter experts to identify existing con- likely to have increased requirements to cepts for how BCI might be applied even in today’s • digest and synthesize large amounts of data operational environment. This analysis informs the from an extensive network of humans and introduction of a BCI toolbox, which serves as the machines conceptual cornerstone for the RAND Corporation’s • make decisions more rapidly due to advances BCI TTX. in AI, enhanced connectivity, and autono- mous weaponry BCI and the Future Warfighter • oversee a greater number and types of robot- ics, including swarms. In his work on the history of the future of war, British One recurring theme in discussions of cognition scholar Sir Lawrence Freedman notes that “there and future combat is that decisionmaking is likely is no longer a dominant model for future war, but to be complicated by the synthesis of enormous instead a blurred concept and a range of speculative amounts of information.69 During ground combat, possibilities.”64 His work does identify a few recent for example, information sources might include such strategic themes addressed in the literature on future tools as acoustic sensor networks, which could pro- wars, including an increased prevalence of hybrid vide the location of distant gunfire, or drone trackers wars, cyberwar, use of robots and drones, and the to detect swarming robots.70 In a future battlefield advent of megacities and climate change as sites and defined by the IoT, smart devices, soldier-worn catalysts for future wars.65 A 2019 RAND Perspective sensors, and unmanned aircraft may flood service further projects several broad strategic trends for members with actionable data.71 Extensive data and the future of war: increased competition for regional new sources of information may improve future sit- hegemony, difficulty defending isolated countries uational awareness but could also complicate consid- and domains, a decline of American qualitative and erations for operational decisionmakers to process. quantitative military edge, blurred lines between Enhanced connectivity will expand the sources and war and peace, and continued war on terrorism.66 speed of information transferred between humans Additionally, increased focus on inter-state strate- or between humans and machines. Such information gic competition and the potential for conflicts with could range from networked connections between near-peer adversaries also raises the possibility that an F-35 and fourth-generation counterparts during future adversaries may wield more resources and an operation to real-time readiness status updates sophisticated technological capabilities than have of ground-based materiel and weapon systems.72

12 Unsurprisingly, services are already pursuing ways to facilitate rapid and extensive flow data between war- fighters and decisionmakers to improve the smooth To address the potential functioning of interconnected military systems.73 BCI systems could serve as a potential future tool support for , this endeavor, allowing human analysts and operator to monitor and exploit larger amounts of information future service members more effectively.74 To address the potential for information over- are likely to engage load, future service members are likely to engage more extensively with more extensively with AI. On the future battlefield, AI tools may help human operators assess an envi- AI. ronment, curate data, and ultimately allow oper- ators to digest greater volumes of information.75 Already, the U.S. Army has sought to leverage AI to experts. In general, BCI could theoretically be “lighten the ” on future warfighters applied to help future warfighters make more as a core capability objective of its 2017 Robotic and informed decisions within a shorter timetable or to Autonomous Systems strategy.76 more effectively engage with more robotic systems As AI and rapid connectivity are increasingly than their current counterparts. incorporated into military operations, the pace of Laboratory studies indicate that BCIs may be warfare will continue to accelerate.77 Thus the speed able to enhance both the speed and accuracy of at which decisions need to be made will also accel- human decisionmaking.82 In a future BCI team, erate. In the coming decades, the United States and AI could theoretically transfer initial near-peer competitors are likely to seek out new ways from a plane or drone directly to the relevant cen- to speed up decision cycles.78 ters of an operator’s brain to further reduce cogni- Finally, the future human warfighter may need tive load.83 In combat, BCI could thus accelerate an to oversee and interact with a larger number of operator’s observe, orient, decide, act (OODA) loop, autonomous and semiautonomous systems.79 Drone through new ways of presenting information and swarms may be incorporated at the tactical and bypassing physical senses.84 Thus, DARPA cites the operational levels in complex urban environments.80 potential ability of military personnel to “facilitate Future ground operations will incorporate robotics multitasking at the speed of ” and “interface into supply and logistics chains.81 Already, service with smart decision aids” as two rationales for its members can look to uninhabited aircraft flying investment in noninvasive or minutely invasive BCI above the combat zone for ISR and close air support. technologies.85 Additional combat applications for machines could BCI could also be used for more efficient engage- include, for example, a robot that would be the first ment with AI to help maintain human oversight to enter a building and take fire, currently one of the over operational decisions within a compressed time deadliest roles in urban warfare. frame. Some scholars have hypothesized that an AI-enabled battlefield could lead to a phase shift in Potential BCI Applications in Future warfare in which the tempo of operations outpaces the speed of human decisionmaking. Some Chinese Combat scholars have referred to this as a battlefield singular- Within the context of these technological and opera- ity.86 Some U.S. scholars have referred to this concept tional trends for military environments, this section as hyperwar.87 summarizes potential relevant applications, based If this hypothesis about the role of AI and on the literature and feedback from subject-matter in warfare is accurate, it may be the case that BCI is the only way to have humans remain

13 effectively engaged in decisionmaking in war and from subject-matter experts, to discuss additional keep pace with machines. In this world, adopting BCI potential applications for BCI during combat and effectively integrating humans with machines is One area where BCI technology could potentially not merely a tactical advantage but the central strate- prove useful for today’s military personnel would gic advantage in warfare. BCI systems could facili- be synthetic telepathy among human operators.92 tate centaur warfighting, leveraging “the precision In 2009, DARPA’s “Silent Talk” program awarded and reliability of automation without sacrificing the grants to research institutions to “allow user-to-user robustness and flexibility of .”88 communication on the battlefield without the use of On the battlefield, one critical question would be vocalized speech through analysis of neural signals,” whether BCI would permit humans to make mean- an application that could greatly facilitate covert ingful decisions within future AI-driven operations communication.93 An external analysis highlights the tempos. potential use of BCI technology to develop shared Additionally, BCI could yield potential advan- within and across units, improve col- tages for human operators seeking to manage future lective awareness of combat challenges, and provide robotics machines, or groups of machines, in combat. combatants with insights into perspectives and inter- As former DARPA program manager Al Emondi has nal deliberations of multiple operators.94 suggested, “As we approach a future in which increas- Direct access to the human brain could also ingly autonomous systems will play a greater role help commanders improve the understanding of the in military operations, neural interface technology cognitive and psychological states of their forces. As can help warfighters build a more intuitive inter- early as 2008, the Air Force investigated battlefield action with these systems.”89 As a practical matter, command-and-control systems that used EEG and the ability to achieve hands-free control of a vehicle, eye movements to “assess the operator’s actual cogni- robot, or a drone swarm though BCI could allow tive state” in an effort to “avoid cognitive bottlenecks operators to use their hands for other tasks, such as before they occur” and eventually to “anticipate carrying a traditional weapon. BCI could also poten- future mission state and operator functional state tially allow operators to do more with a swarm than ahead of time.”95 In its vision statement, the ARL’s manual operation would permit. One 2009 North Cognition and Neuroergonomics Collaborative Atlantic Treaty Organization study concluded that Technology Alliance (CaN CTA) makes the case for the goal of having a single human operator control developing the capacity to continually monitor oper- multiple vehicles was “at best, very ambitious, and, ator neurocognitive behavior, including depth, dis- at worst, improbable to achieve.”90 Current work on tribution, and shifting of human attention, appraisal brain-swarm interface posits that brain-computer of information, the emotional context of actions, technologies may be able to improve this challenge.91 and the impact of physiological state—, stress, —on cognitive and motor performance.96 Existing for BCI Combat This type of function could plausibly identify and facilitate operations for extremely fatigued convoy Applications drivers, or perhaps for gunners or tankers operating Even today, BCI technologies, if they were avail- in complex environments for whom mistakes could able and more readily deployed, could yield specific prove deadly.97 At a more complex level, a technology operational benefits because of the direct access they that could provide insights into the emotional state permit to the human brain. Some DoD research of a soldier might provide red flags as to whether programs publicly identify potential military oper- and when the soldier might “break” psychologically, ational applications of BCI that could be relevant in when a soldier might have psychotic tendencies, or the existing operational environment. The following perhaps when a soldier is shooting to miss. One study section draws from these themes, as well as insights on the use of functional magnetic resonance imag- ing to identify falsehoods suggests that the ability to detect whether a subject is concealing information

14 may be of particular interest to counterterrorism and counterinsurgency missions.98 DoD has also explored the application of BCI Beyond cognitive technologies to improve cognitive performance during or in preparation for combat.99 Potential enhancement, BCI military applications offered by enhanced cognitive abilities of service members, through electrical or could also be used chemical stimulation, might include improved mem- ory of battle assignments or of large amounts to reduce or to of information by a fighter pilot.100 Caffeine has been regulate such other used as a cognitive stimulant by the U.S. military for over a century.101 More recently, researchers from emotions as . the Air Force Research Laboratory have highlighted cognitive challenges associated with high-level multitasking environments as an impetus for applied research on transcranial direct current stimulation (tDCS) in the military context.102 DoD has also also arguably positive products of strong emotion invested in efforts to accelerate military training in combat, including an increase in adrenaline that 107 through the use of BCI.103 As DARPA’s Targeted improves physical capability. Neuroplasticity Training program description In the future, BCI that improves human observes, service members often need specialized sensors—eyes that could see in different spectra skills demanding perceptual acuity, rapid and accu- or ears that could hear sounds outside the usual rate judgment, and effective planning and execution human range—might improve situational awareness of complex actions. The existing training for these in infantry operations. As former DARPA program can be time consuming and require high aptitude.104 manager and former Army infantry officer Geoffrey Thus, DoD perceives utility in pursuing technologies Ling has observed, “If I gave you a third eye, and the that could reduce the time, investment, and innate eye can see in the ultraviolet, that would be incorpo- aptitude required for the acquisition of these special- rated into everything that you do. . . . If you can see at ized skills. night, you’re better than the person who can’t see at 108 Beyond cognitive enhancement, BCI could also night.” be used to reduce pain or to regulate such other emotions as fear. As one analyst with military Testing BCI Capabilities medical experience has observed, BCI capabilities that can physically manipulate the central ner- Through National Security vous system and disrupt pain could offer “practical Gaming applications as an electronic anesthetic.” DARPA’s Can BCIs support national security and future Electrical Prescriptions (ElectRx) program seeks to warfare? If so, how? When applied to the anticipated support military operational readiness by developing future of warfare, the technology summary for BCI nonpharmacological treatments for pain, general suggests there may, in fact, be operational benefits. , posttraumatic stress, severe anxiety, To test this further, the RAND team conducted a and other challenges through the stimulation of the TTX—a national security game—centered around 105 peripheral . Commanding officers a toolbox of projected future BCI capabilities. The have long grappled with how best to manage fear game brought together experts with technical and on the battlefield as warfighters make individual or operational experience and challenged them to collective decisions to fight or not fight when fearing make choices about what BCI technology they would 106 death. Application of BCI to improve manage- employ and why across tactical vignettes. The game ment could plausibly be of use, though there are

15 reinforced some preliminary hypotheses about the of plans or tactics on the battlefield, or improve com- appeal of specific BCI tools and clarified potential munication with headquarters to enhance command- use cases in an operational environment where the ers’ awareness of in-theater conditions. value of BCI was disputed. Importantly, the game Monitoring performance would enable awareness also provided insights into potential vulnerabilities of group or individual emotional, cognitive, and and risks, detailed in the next section. physical states. It could permit monitoring neural and cognitive state, thus detecting when a person is A Projected BCI Toolbox fatigued, paying attention, has high or low cognitive workload, or is significantly stressed. It might also Drawing from the technology summary and con- help a commander to better understand aggregated sideration of future operational requirements, we squad or platoon cognitive state and fatigue. projected a toolbox of six future sets of BCI capabil- Enhancement of cognitive and physical perfor- ities. The first three capabilities generally relate to mance includes improving a warfighter’s cognitive connectivity between humans, and between humans and physical states on the battlefield. Cognitively, it and machines. The second three relate to human could yield enhanced focus and alertness for rapid performance and training. These capabilities served and improved situational awareness and decision- as the core for the July 2018 BCI TTX—in effect, they making. The warfighter would also be afforded an served as our hypotheses about which BCI technol- enhanced emotional state that could, for example, ogies will be most relevant to tactical military units. disrupt fear and mitigate stress. It could also enhance They are listed as follows: cognitive skills training, per DARPA’s Targeted Human-machine decisionmaking involves trans- Neuroplasticity Training (TNT) program.109 ferring data to the human brain from sensor input With regard to physical performance, it might and from the brain to machines. It might help users include regulating or enhancing a warfighter’s aggregate and transfer information and assessments. psychological state.110 It could enhance sensory For example, a computer might sort and display capabilities through stimulation of the peripheral information in an easily digestible form for quick and nervous systems and possibly specific cortices (i.e., accurate response. Alternatively, with cortically cou- visual or audio). It could also enable the mitigation pled AI, data can be provided from the human brain of pain via pharmaceutical distribution. Finally, this to a computer. This kind of tool allows a warfighter tool could also include improved strength through to digest more information faster, to be used, for more efficient integration with mechanical exoskel- example, with theater assessment or risk and threat etons,111 which are natural extensions of the work on assessment. Warfighters ultimately can increase over- prosthetics. all reaction time, thus collapsing the OODA loop. The Training BCI tool could improve operator Human-machine direct system control involves learning and memory processing, allowing war- allowing warfighters to control systems with their fighters to retain more information. It could also thoughts wirelessly, as well as to supervise semiau- enable accelerated training, including deployable tonomous and AI systems, including robots, drones, training devices for rapid training in theater. It drone swarms, or jets. It might, for example, enable could allow for adaptive (and more effective) per- an immediate system shutdown or weapons launch sonalized mission-specific training. BCI could simply by exercising a thought. This, in turn, pro- provide more effective feedback during training vides the warfighter increased situational awareness and—someday—could enable implanted and again helps collapse the OODA loop. sets for immediate “training.” Human-to-human communication and man- Figure 1 lists the capabilities of a BCI toolbox agement entails wirelessly transmitting commands that may be available in a relatively near time frame, or basic ideas among warfighters and commanders, as well as longer-term projections. These capabili- lightening the load of systems. It ties are grouped with respect to the tools discussed could facilitate immediate and silent communication above. In general, the long-term capabilities reflect

16 FIGURE 1 BCI Toolbox for National Security Game

2030 2040 2050

BCI tool Near-term capabilities Long-term capabilities 1) Human-machine • Immediate transfer of operational risk • Transfer of risk and threats (increased bandwidth) decisionmaking • Faster decisions to deploy weapons • Augmented AI systems • Shorter preparation cycle with faster feedback from occurrences in battlespace (collapse OODA loop) • Increased speed and accuracy of targeting

2) Human-machine • Transfer basic commands to systems • Transfer of complex manipulations (increased direct system control • Increase situational awareness and reaction bandwidth and degrees of freedom) • Collapse OODA loop • Resistance to distraction (use in dynamic environments) • More speci c commands and control

3) Human-to-human • Transfer basic commands between individuals • Transfer complex strategies involving communication/ • Reduce (radio) weight commanders/headquarters (increased bandwidth) management

4) Monitor performance • Monitor state • Long-distance standoff assessment • Monitor individual and group cognitive • Monitoring of adversary emotional and cognitive workload, stress, breaking point states • Archived dynamic cognitive pro les

5) Enhance cognitive • Regulate emotional state (i.e., stress) • Modulate emotional state performance • Increase focus and alertness

6) Enhance physical • Improved strength augmentation • Implanted auto pharmaceutical distribution performance • Improved sensory capabilities • Pain disruption

7) Training • Increased learning retention • Implanted knowledge sets • Deployable training devices • Adaptive individualized training • More immediate and effective assessment

NOTE: This framework was used to support game play but does not reflect a technical maturity assessment. an improvement in the and bandwidth to make decisions about which, if any, of the BCI of data being transferred. Regarding direct system toolbox technologies they would utilize in two urban control, in addition to transferring more complex ground combat missions. This event allowed us to manipulations of a system, long-term capabilities better understand (1) whether players perceived may also reduce the sensitivity of BCI systems to any BCI technology as useful in a complex context, user distractions. Long-term capabilities related to (2) the perceived relative advantage of different monitoring performance will allow organizations to technologies for different tasks, and (3) the rationales archive cognitive performance and profiles over time. for why players did or did not opt to use different technologies, helping to unpack advantages and lim- Testing the Operational Relevance of itations of the tools. Our BCI TTX convened a small group of players BCI Capabilities from diverse military and technical backgrounds To provide an initial test of the utility of the seven to drive conversation and elicit a broad range of BCI toolbox areas to tactical military operations, we insights. The primary intent was to explore the ran a one-day game in which we convened experts relevance of BCI capabilities in a military setting and

17 drew two mission vignettes from doctrine for urban operations to allow players to drill down into com- Urban infantry mon tasks: clearing a building and responding to an ambush. Both vignettes included complex subtasks operations presented a that could address the range of warfighter functions. Additional details of the game design and execu- “most challenging” use tion are included in the appendix. case, with target users Insights on the Use of BCI from Game who have traditionally Play been skeptical of the The game provided several types of insights to help us understand the utility of BCI in urban operations. utility of technological First, we collected data about which BCI capabilities players opted to leverage to tackle which functional advances. areas. We used this information to understand which capabilities were seen as more useful to support com- plex ground operations. Second, we collected data discuss the nuances of how such capabilities might on the players’ discussion of why they saw specific be used. On the technical side, players included technologies as promising or risky. In this section, we researchers and managers with expertise in neuro- discuss which technologies players opted to use, and science, military technology, and human-machine what they saw as the advantages. These findings gen- teaming. A second set of players was drawn from erally align with our initial hypothesis that BCI has current and prior-service officers and experts on uses even in complex environments but showed that military affairs, many with experience in urban oper- players did not see the six capabilities as equally help- ations. The group was asked to represent U.S. forces ful. Beyond the potential promise for BCI that was as a whole in the vignettes. Two RAND analysts who supported by game play, we also noted many import- study disruptive technology took on the role of the limitations and risks introduced by using BCI. adversary during the last stage of the game. These findings are addressed in the next section. Players were asked to apply and weight the utility Our initial insight from the game is that BCI of each of the six BCI capabilities from the BCI capabilities can, in fact, be useful on the urban toolbox in the context of two tactical urban opera- battlefield, supporting the team’s initial hypothesis. tions and across the six warfighter functions: mission All BCI tools were used multiple times by a player, command, intelligence, fires, movement and maneu- and all warfighting functions had a majority of tools ver, force protection, and sustainment. This process applied at least once. When given a choice between allowed us to compare the players’ assessment of the using BCI technologies or not, players often decided utility of different technologies for different func- to use a BCI technology to confront the challenges of tions across two different missions. the tactical vignettes.113 More specifically, the game Urban infantry operations presented a “most provided evidence about the relative frequency at challenging” use case, with target users who have which BCI tools were selected to support different traditionally been skeptical of the utility of techno- warfighting functions, which suggests the perceived logical advances. As a former Assistant Secretary relative utility of different technology baskets. of Defense for International Security Affairs and Overall, human-machine decisionmaking, direct combat Marine, Bing West once bluntly suggested, systems control, and enhanced physical performance “Urban battle will remain a slugfest, with the basic were noticeably more popular than other tools. ingredient remaining heavy doses of high explosives. Perhaps not surprisingly, these tools were most often No technology is emerging to replace that.”112 We

18 used to support mission command, intelligence, fires, in BCI solutions that aided communication and and force protection tasks. Enhanced cognitive per- decisionmaking. formance and training were rarely used, but that may Drawing from the identified subtasks, partici- have been due to the tactical nature of the problem pants underscored the need to communicate clearly sets, which could have made long-term issues like and make swift and accurate decisions under condi- training less pressing. All of these results were in line tions with limited line of sight, visibility, and . with our initial projections, but the independent sup- As a result, players focused their discussion on BCI port from game players provides additional evidence tools that could assist with command and control, for the utility of BCI for specific operational tasks. communication, and intelligence. As one player sug- The game also underscored the need for fur- gested, BCI potentially offers the “totality of infor- ther analysis of the future technical capabilities of mation, getting voices and images together” while BCI-related military applications. Participants noted also providing tools to help make of otherwise that the pragmatic utility of each BCI tool would disorienting conditions. depend largely on its fidelity and reliability during Speed of decisionmaking and enhanced common combat. A human-to-human communication tool awareness were repeatedly mentioned as key qualities permitting fully shared situational awareness would of BCI tools that would be important in improving be of greater use than more rudimentary communi- current communication, command and control, and cation, which could be as easily transmitted over a intelligence. As one player observed, a unit leader traditional short-wave radio. A BCI device overseeing clearing a building would “need to know who is the operation of drone swarms is only as useful as it really scared, or dead, and would need to know that is dependable in maintaining control during unfa- without spending time to talk.” vorable conditions. Additionally, the incorporation Even for such functions as movement and of adversary automated lethal systems into the game maneuver, player discussion focused on the need to raised technical questions about the future ability of communicate the location of other service mem- BCI to overcome time constraints associated with bers and the status of different areas of the building human decisionmaking. One player commented that instantaneously, although there was some debate some of the capabilities presented in the BCI toolbox about the additional value of BCI for this task beyond did not feel sufficiently concrete to make an opera- the use of a traditional radio. tional determination, which may have contributed Players also selected tools that provided a com- to some players’ decisions. The next section further mander with more refined information on which to explores some of the key vulnerabilities and risks base decisions. To the extent that BCI could enhance identified during game play, which must be given a commander’s ability to rapidly cull information further thought before BCI can be productively from front-line soldiers, sort that information, and fielded in combat environments. make decisions on targeting and human manage- ment, it would potentially be useful in accomplishing Vignette 1: Clearing a Building the task. The first vignette, clearing a building, was drawn Control over robotics and physical enhance- directly from Marine Corps Warfighting Publication ments were also mentioned but discussed far less, Military Operations in Urban Terrain (MOUT).114 In and they seemed to be less central to players’ con- response to the vignette, players cited locating civil- ception of the vignette’s challenges. For example, ians and killing the enemy as two critical subtasks. many of the operators stressed that, in these types of Communication, control of forces through multiple conditions, the Blue shooter would still be a human rooms, and ensuring situational awareness were wielding a rifle, because the extra degree of judgment derivative subtasks. Overall, players identified the and accountability would be necessary. The lack of key problem as one of information collection and emphasis on controlling such emerging technologies human management and thus were more invested as drone swarms and battlefield robotics may also have been due in part to the fact that players were

19 asked to consider only current levels of non-BCI Additional Use Cases technology. Players also noted several additional areas where BCI would be helpful beyond the confines of the Vignette 2: Ambush and Casualty Evacuation vignettes. For example, players discussed the utility The second vignette adapted an ambush descrip- of providing BCI to a mechanic, who could then draw tion from MOUT to include a casualty evacuation on either machine learning–based diagnostics or the described in a firsthand account of the battle of experience of a more senior technician when mak- Fallujah.115 This vignette also explicitly incorporated ing repairs. Players felt that this would enable more future non-BCI technologies projected to be available capable support functions that, while not depicted to the United States and near-peer adversaries in the directly in the vignette, would be key to maintain- 2040 time frame, including autonomous lethal AI; ing forces over longer operations. Similarly, players air- and water-based swarming unmanned vehicles; mentioned that if there had been allies or partners advanced electronic warfare (EW); integrated intel- operating alongside the vignette forces, BCI could be ligence, surveillance, and reconnaissance; situational helpful in overcoming the barrier to smooth awareness; and fires support.116 communications. The short duration of the vignettes In this vignette, the focus shifted from commu- also minimized the role of cognitive and physical nication to control over multiple platforms and medi- fatigue in player decisions. Players noted that poten- cal support. The additional consideration of plausible tial offensive applications of BCI were not included future military technologies highlighted new areas in the toolbox but could be useful. Players also noted for potential human-machine interaction. While that the opportunities for BCI in the operational players did identify similar BCI technology as helping management of the fight would likely be extensive with command and control and intelligence, as in the but would be quite different from the tactical applica- first vignette, these warfighting functions were far tions that the game focused on. less prominent in discussion. Instead, players placed It is important to caveat that, while players felt more emphasis on the multiple platforms that needed BCI would be useful in the urban environment, they to be coordinated to provide situational awareness, were quick to note that it would not be a panacea. fire support, and medical evacuation capabilities. In addition to the potential of BCI, the exercise also Players also noted that the more-open lines of sight highlighted possible vulnerabilities, challenges, and in the vignette made air support and ground-based risks created by the use of BCI. These are captured in robots more useful, making direct control a more detail in the following section. attractive option. The presence of an active casualty moved appli- Summary of Game Findings cations of BCI to providing and monitoring medical While the insights of a single game should not be care to a central focus. Building from a firsthand overstated, our BCI TTX contributed to an emergent observation that “time is critical” when addressing discussion on BCI by identifying and considering urgent casualties, participants considered ways that tangible ways in which future BCI capabilities might 117 BCI could improve response time. One player contribute to combat operations. Participants chose suggested that transfer of advanced medical expertise BCI tools over traditional military approaches for through BCI could transform any combat medic into tasks across the spectrum of warfighter functions. a surgeon, potentially reducing transport time to crit- Of the seven BCI capabilities identified in the study, ical care. Players paid particular attention to using participants found the most uses across the two human-machine decisionmaking to support the cor- vignettes for three. First, participants prioritized BCI rect allocation of resources—for example, determin- support to human-machine decisionmaking, antici- ing how much service member time should be spent pating the benefits from the integration of informa- providing medical care, or allocating unmanned tion from many sources during a chaotic battle or systems to different tasks. the acceleration of decisionmaking during combat.

20 Second, direct systems control through BCI could offer combatants hands-free control of semiautono- mous systems and drone swarms. Third, enhanced Analysts of emerging physical performance would offer improved audi- tory and visual capabilities, or more fluid control military innovation of exoskeletons. Participants noted that utility of this function may become particularly pronounced often note the need for once technology for military applications of AI and robotics develops further, improving the fidelity and caution surrounding the reliability of BCI tools, and once adversaries have capability-vulnerability access to these capabilities. Further analysis of these three capabilities could further refine their uses and , whereby associated risks. Our participants noted that future direct new advantages brain-to-brain communication among service members, while requiring more advanced technology can introduce new than may be available in the 2040 time frame, could be revolutionary in allowing a team to coordinate vulnerabilities. actions while clearing a building. The discussion also suggested that the technologies might pose increased insights from a Red team dedicated to identifying risks of adversary exploitation and would have the BCI-related vulnerabilities. While the game con- greatest impact on current military organizational firmed the potential usefulness of BCI technolo- structures, depending on whether the capability can gies on the battlefield, it also highlighted potential be turned off and used selectively. Many of the chal- risks. With respect to policy, the value proposition lenges identified in the game highlighted the likely certainly suggests continued investment and devel- future amplification of cybersecurity risks by the use opment, but the risks highlight key areas where poli- of BCI. Our Red team underscored the potential vul- cymakers should be proactive. This section considers nerabilities associated with hacking, denial of service, potential operational vulnerabilities, institutional and EW. vulnerabilities, and ethical and legal risks, all associ- ated with combat applications of BCI technologies. Potential Risks Operational Vulnerabilities Analysts of emerging military innovation often note the need for caution surrounding the A Red team analysis of game-player decisions helped capability-vulnerability paradox, whereby new identify ways in which BCI technology could create advantages can introduce new vulnerabilities.118 The new vulnerabilities for the future warfighter. To introduction of any new technology may present new scope the analysis, game facilitators sought to distin- challenges, risks, and vulnerabilities. Thus, in addi- guish between new vulnerabilities specifically associ- tion to considering the operational utility of future ated with future BCI capabilities and those associated BCI technologies in combat, this project sought to with greater future reliance on technology in general. consider how the unique attributes of BCI might Participants generally agreed that within each new present new considerations for DoD and to identify area of vulnerability, the extent of the vulnerability core areas for further examination. Some of these would depend on both the reliance on BCI technol- considerations were derived from existing literature; ogy and specific features of the technology itself. most were drawn from discussions that were held in preparation for and during the BCI TTX, including

21 New Potential Points of Failure to the brains of service members could plausibly pro- One major area of vulnerability was that reliance on vide near-peer competitors with valuable information BCI use could present new ways for an adversary to regarding the U.S. disposition of forces, organiza- deny access to the technology, potentially rendering tional frictions, and vulnerabilities among individual a unit less effective. A 2014 examination of synthetic service members themselves. The degree of vulnera- telepathy concludes that “brain-to-brain communica- bility of operators’ brains would likely depend on the tion over the Internet may never be the best solution fidelity of the BCI technology employed, the amount for the battlefield, despite the millions of dollars of of sensitive information that operators had access to, Pentagon research money that’s gone into exploring and the robustness of physical and behavior counter- it.”119 This may be due in part to the potential for measures designed to thwart adversary hacking service denial. The security of the network among attempts. brains, or between brains and machines—and net- work vulnerability to electromagnetic pulse—would New Areas of Exposure to Harm or Influence thus be of paramount importance as EW attacks Red team participants noted that because BCI tech- began. In fact, this issue arises outside the field of nologies may directly connect to an operator’s brain, BCI as interests in secure networked communication they may present new areas of potential exploitation. on the battlefield increase. Physical vulnerabilities would likely be most acute Overreliance on new mediums of transmis- with the invasive variant of the technology. Already, sion could be problematic for any new technology, unconventional attacks are suspected of causing including future battlefields driven by the IoT, and traumatic brain injuries to U.S. government employ- maintaining communication channels is likely to be ees in China and Cuba.121 If adversaries are currently a priority. However, BCI could present a particular experimenting with disrupting the human brain at vulnerability because of its technical reliance on a distance using ultrasonic frequencies, microwaves, detecting very weak electrical signals. In a battlefield or other methods, implants could provide direct situation, these weak signals could potentially be entry into the brain for damage. Just as it is possi- jammed. As one Red team member noted, ble to hack a pacemaker or insulin pump, it is quite has significant EW capabilities embedded even at conceivable—albeit far in the future—that someone the lowest tactical level. Mitigation options could could hack a BCI and send cognitive commands or include helmets or masks that create a Faraday cage, even thoughts to the brain.122 shielding the user from jamming attempts. However, Reports of Russian pinpoint propaganda, the increased weight and the reduced visibility that text messages that seek to demoralize individual might result from use of such equipment could make Ukrainian soldiers through threats and false reports this a problematic option. of leadership desertion, offer one insight into how technology can enhance emotional manipulation Adversary Access to New Information tactics.123 Hacking BCI capabilities could theoreti- In addition to the risk of having signals jammed, cally provide adversaries with direct pathways into there is a risk of adversaries intercepting and using the emotional and cognitive centers of operators’ signals. Technologies that provide access to an oper- brains to sow confusion or emotional distress. In the ator’s emotional or cognitive states could potentially extreme, adversary hacking into BCI devices that be a treasure trove for adversary intelligence collec- influence the motor cortex of human operators could tion. Russia has reportedly targeted NATO soldier theoretically send false directions or elicit unin- smartphones for information on operations and tended actions, such as friendly fire, although such troop strength, while the Chinese government has influence may be technically difficult to achieve in reportedly hacked military contractor computers to the near term. Even an attack that broadly degraded extract highly sensitive data about future submarine gross motor skills could prove debilitating during warfare.120 BCI technologies that permit direct access combat.

22 Institutional Vulnerabilities absence. Participants highlighted the need for redun- dancy. Examples might include maintaining secure Trust radios to supplement computer-mediated telepathy, BCI research related to tactical applications has ensuring that members of military units continue to largely been driven by the “push” of technological be well-versed in traditional navigational techniques, advances, rather than the “pull” of requests for BCI and ensuring alternative means of communicating capabilities from the military field. Some participants with machines on the battlefield. in the game suggested that service members were likely to be skeptical about the practical utility of Erosion of Unit Cohesion BCI technologies. Although this issue of trust is not As humans become more closely intertwined with unique to BCI technology, it should be a focus point machines through BCI, technology could have for future policy. Those with direct infantry and close profound implications for the interpersonal relation- combat experience may be most wary of the potential ships that have traditionally played a preeminent role risks and downsides of the technology because of the in warfighting. The implications may be difficult need to balance technology with lethality.124 to predict. On the one hand, an ability to directly Acceptance of BCI may also be complicated by sense the thoughts and emotions of other members a general phenomenon referred to in the of a combat team could increase unit cohesion. On community as the “yuck factor,” in which a negative the other, there is evidence that advances in virtual emotional response is provoked by new advances in communication technologies that permit “chat-room biotechnology.125 Lack of trust would likely be more coordinated strikes” may already reduce emotional acute for invasive BCI, which requires alterations and psychological bonds among soldiers.128 To the to the human body and poses health risks, such as extent that they would replace traditional interac- infection. Trust could also be influenced by the scope tions among members of a military unit, future BCI of the information accessed by BCI technologies. capabilities could fundamentally alter the nature Service members may not want to provide the U.S. of these human relationships. An increasing use of government, or its machines, with access to the inner robotics and AI in combat could compound this workings of their minds. challenge. In fact, preliminary research on robots Additionally, the potential pace of decisionmak- on the battlefield has indicated the development of ing permitted by BCI—as well as the delegation of strong human-robot attachment, or even a feeling of tasks to AI—may raise particular concerns among “self-extension into the robot,” that might influence warfighters.126 Such concerns could exist in an operational decisionmaking.129 environment driven by technological advances even More broadly, the introduction of new BCI tech- in the absence of BCI. Some of these concerns also nologies raises questions about the future structure could be offset by the advent of civilian technologies that employ BCI, or generational shifts that yield a force for whom BCI technologies and biotechnolog- ical interventions are more acceptable.127 In other areas of , including study drugs, Service members may modafinil, or tDCS, civilian technologies and appli- cations have outpaced those of DoD. If this trend not want to provide the were to continue with BCI, military adoption could plausibly lag civilian adoption, potentially enhancing U.S. government, or its popular trust in the technology. Conversely, too much trust in BCI technology machines, with access could compound potential operational vulnerabil- to the inner workings of ities: As they became reliant on BCI technologies, warfighters might someday be unable to operate in its their minds.

23 of the human force. What does a company or platoon Institutes of Health established a work- look like when some or all of the force is neurally ing group that meets periodically to consider eth- plugged into various weapon systems, drones, or ical challenges in the development or application robots? Will these capabilities be integrated, or will of neurotechnologies at large, including BCI.132 As they be assigned to specialized detachments? How BCI technologies advance, government bodies that might it affect unit cohesion when senior officers support and use the technologies will need to develop can monitor service members’ emotions or even systems to supervise and manage BCI use to mitigate thoughts, and when some unit members have access abuse. A National Academy of Sciences study, in par- to BCI capabilities while others do not? ticular, emphasizes the need for organizational safety nets in the development and application of future Erosion of Unit Leadership BCI technologies.133 This section explores two issues Technologies that permit senior officers to monitor deserving of ethical and legal consideration: respon- and communicate directly with the brains of com- sibilities to a BCI operator both during and after mil- bat personnel could potentially undermine effective itary tenure, and responsibilities of the BCI operator squad-level leadership, extending micromanagement for actions taken in combat with BCI technologies. to new frontiers. In one hypothetical dynamic, BCI technologies could exacerbate an existing trend Responsibility to the BCI Operator toward what has been dubbed the tactical general: BCI breakthroughs will require consideration of the Senior officers, empowered through such new risks to operators, some of which would be amplified technologies as drone feeds, may tend to use these within a military context. Potential risks to opera- technologies to exert greater control over operators tors may occur during R&D, during operations, and in the field.130 Future BCI technologies that permit even long after exposure. As with any biotechnical direct brain-to-brain communication could poten- advance, the U.S. government will need to prepare tially exacerbate this dynamic, contributing to a for new responsibilities at each stage of use. A 2014 more robotic, less adaptive and resilient approach to report by the National Academy of Sciences, commis- unit-level leadership. sioned by DARPA, begins to address potential risks Unit-level leadership could also be jeopardized associated with the use of neurotechnologies within a by BCI technologies that provide senior leadership military setting, but additional questions are likely to with access to individuals’ physiological, emotional, arise as the technologies progress.134 and cognitive states. Traditionally, it has been the The safety of neuroscience-based interventions role of a squad leader to understand the physical and represents a primary concern for the U.S. govern- emotional states of his or her team through months ment.135 Currently, the health risks associated with of relationships, evaluation, training, and combat BCI are not fully known. Invasive techniques may experience. Technologies that allowed senior officers prompt the most vigorous public discussion of health to bypass unit-level leaders and second-guess their impacts, given the fact that they require the implan- judgments might be undermining unit leadership tation of a foreign object into a major organ and may rather than supporting it. carry immediate risks of infection from surgery. However, the long-term implications of noninvasive 136 Ethical and Legal Risks techniques are also unknown. The extent to which neurotechnologies rep- Two decades ago, biotechnology ethicists wrote that resent changes to the human brain and body has “the most frightening implication of [BCI] technol- legal as well as ethical implications. By statute, DoD ogy is the grave possibility that it would facilitate is responsible for any changes to a veteran’s body totalitarian control of human. . . . a paramount worry during their time of service. To the extent that BCI involves who will control the technology and what may change a service member’s baseline health will be programmed.”131 More recently, the National over the long term, the U.S. government would

24 be responsible for any service-related disability.137 Beyond the physical implications of the technology itself, the withdrawal of “” capabilities Ethicists also afforded by BCI—such as control over machines or cognitive enhancement—might elicit psychologi- emphasize the need for cal harm for which the U.S. government would be responsible. independent institutional Given the potential safety risks associated with BCI, any U.S. government agency developing and review, appropriate operationalizing the technologies would need to training, and adherence consider how to protect the principle of informed and voluntary consent of operators. Consent to any new to international and potentially risky technology may be complicated by the unique “competitive and coercive pressures” guidelines. of the military context.138 Limited personal autonomy among military personnel, as well as a lack of infor- mation about long-term health risks, have led some that “if N3 is successful . . . we could face questions ethicists outside of government to argue that BCI related to agency, autonomy, and the experience 144 interventions, such as noninvasive brain stimulation of information being communicated to a user.” techniques, are currently inappropriate for a military Within the military context, services might consider or security sector setting.139 arbitration mechanisms so that service members and Because BCI technologies could provide direct their commanding officers may discuss or object access to the human brain, the U.S. government will to unethical or harmful uses of BCI technology. need to consider implications for the privacy and lib- Such a mechanism might lie outside the chain of erty of the operators.140 is a related command—similar to the State Department “dissent concern for neuroenhancements, because, as one channel”—in an effort to counter some of the institu- bioethicist has articulated, it “concerns an organ that tional pressures of a highly hierarchical organization. mediates human identity.”141 Given these challenges, ethicists have proposed four emergent rights: “the Responsibility of the BCI Operator right to cognitive liberty, the right to mental privacy, As with many ethical and legal issues, responsibilities the right to mental integrity, and the right to psycho- rest not only on the institution but also on the indi- logical continuity.”142 As few are likely to volunteer vidual. The prospect of lethal autonomous systems for a mission that permanently eliminates personal on the future battlefield, for example, has prompted privacy, it would benefit DoD to develop meaningful extensive discussion about the potential legal and privacy policies surrounding BCI use before adopting ethical challenges associated with human account- the technology. In particular, how might mental and ability for the laws of war, as well as the need to retain cognitive privacy rights be applied in a combat envi- human agency and intent in decisions to use force.145 ronment? Over the longer term, would data extracted As one writer has noted, when an autonomous from service members’ brains through BCI be ano- system kills noncombatants, who is responsible? Is nymized? Should it expire? Alternatively, might this it the software programmers who coded for target technology be used to identify “super warriors” to identification, the system’s commanding officer, the form future elite forces? combatant commander who authorized the opera- Ethicists also emphasize the need for indepen- tion and use of the system, or someone else?146 One dent institutional review, appropriate training, and guiding concept in this debate has been meaningful adherence to international guidelines.143 DARPA human control, meaning a human should make program manager Al Emondi has highlighted some the final determination over whether or not to kill ethical questions surrounding BCI, anticipating another human. It was this that prompted DoD

25 Potential cognitive and emotional changes associated with BCI technologies that modify the BCI may also contribute human brain raise further questions about operator responsibility. Ethicists have concluded that altering to a diffusion of the levels of fear and aggression in service members could “expose soldiers, their missions, and society operator responsibility. in general to increased risk of injury or death.”152 A National Academy of Sciences report notes that to develop a directive requiring that autonomous electrical disruption of one region of the brain may weapon systems permit operators to exercise “appro- reduce the inhibitions of soldiers about morally 153 priate levels of human judgment over the use of problematic behaviors. If these regions are stimu- force.”147 As the International Committee of the Red lated by BCI during the course of combat and result Cross recently proposed, human engagement should in atrocities, the report asks, “How and under what be of the “type and degree of control that preserves circumstances might neurally-manipulated soldiers human agency and upholds moral responsibility in be accountable for activities that violate the laws of 154 decisions to use force”148 However, meaningful in this war?” Currently, a soldier is required to refuse an context can often be ambiguous and debatable and, illegal order. Would the soldier’s ability to refuse be thus, should be revisited as new technologies emerge more complicated if the instructions came directly and advance.149 from a ? If so, how might the ero- The incorporation of AI on the battlefield and sion of personal agency—and culpability—further the accelerating pace of future warfare may make influence combat decisionmaking? While such a it increasingly difficult to ensure that DoD pre- scenario appears unlikely to emerge in the near or serves “appropriate levels of human judgment” over even medium-term future, it is a potential element of use-of-force decisions. As noted earlier, BCI may per- the technology’s future trajectory that may need to be mit humans to accelerate decisionmaking in an effort considered. to maintain operational relevance on a battlefield 150 that incorporates AI. Should this be technically Conclusions and feasible, it is not clear that a decision made within the necessary time frame would allow for reasonable Recommendations moral and ethical judgment. Summary and Primary Findings BCI may also contribute to a diffusion of opera- tor responsibility. Once humans and machines work Former Secretary of Defense Ashton Carter has more closely (via BCI) to make use-of-force decisions observed that while “the accelerating pace of innova- during the heat of combat, it may be more difficult to tion is already bringing great progress . . . it would be 155 determine the meaning of meaningful human control foolish to let inertia set the agenda.” As an example, or “appropriate levels of human judgment.” In the BCI technologies, developed in part by DoD funding, traditional military “kill chain,” decisions can be have advanced significantly in recent years and are rolled back at each stage to determine legal culpa- likely to continue to progress, whether under govern- bility and the existing “kill web,” in which several ment, academic, or private-sector auspices. The U.S. people contribute to a decision error and no one is government thus has an opportunity to play a con- ultimately found culpable.151 Could this dynamic be structive role in the coming decades in supporting exacerbated by BCI-facilitated decisionmaking in elements of BCI technology that benefit U.S. national which individuals share thoughts and decisionmak- security and seeking to mitigate risks. ing instantaneously with machines and with one This report has provided an initial evaluation another? of the potential applications of BCI in a military setting and has highlighted potential policy issues that should be addressed. Our analysis and game

26 prototype contributed to nascent discussions on the a few ideas to mitigate operational risks, including extent to which BCI technologies might open up new potential use of EW shielding integrated into armor, areas of operational risk during combat. Our work secure networks, and steps to ensure that traditional also considered how an evolving relationship between backup methods are preserved. humans and machines, facilitated by BCI, might profoundly alter existing military organizational Recommendations structures and relationships and pose new ethical and legal challenges for DoD. Moving forward, we recommend that the U.S. gov- In addition to considering policy issues and the ernment conduct additional national security gaming potential tactical value of BCI, this research yielded a to further assess the operational risks and benefits of systematic approach to exploring the implications of BCI technology in combat, including provisions for emerging technology. This approach, which incorpo- additional domains and contingencies. Beyond oper- rated technology review, operational considerations, ational risks, the government will need to address a and technology deconstruction (into a set of practical potential lack of trust in BCI technologies, which is capabilities) into a TTX for testing in-theater impli- an issue that emerged during the game as a poten- cations, is scalable and can be applied to a variety of tial impediment to adoption by the armed services. emerging technologies. This, in turn, requires special attention to how BCI is Our game and associated research indicated that deployed as it matures. Our review of current techno- despite valid concerns, BCI can likely be useful for logical progress highlighted work done in academic future military operations, even in the most difficult and private-sector laboratories, and the U.S. govern- test case: infantry ground force combat. This utility ment should seek to leverage work in both, especially may become particularly pronounced once tech- as the commercial sector increasingly dominates nology for military applications of AI and robotics technology R&D. Developing and deploying BCI develops further, and once adversaries have access to technologies in the national security sector will these capabilities. Nonetheless, the application of BCI require institutional adaptation to operators at each would support ongoing DoD technological initia- stage of the process. Next, we offer some concluding tives, including human-machine collaboration for suggestions on each of these points. improved decisionmaking, assisted-human opera- tions, and advanced manned and unmanned combat Expand Analyses to Illuminate Operational teaming. Relevance and Vulnerabilities Of course, as with most significant technological Over the coming decades, it will be critical that oper- advances, there are potential risks. BCI falls subject ational needs and risks, rather than just technical to the capability-vulnerability paradox, with coun- opportunities, drive BCI development. To help sup- terweighted benefits and risks, and, as development port this need, we developed a systematic approach efforts and eventual acquisition efforts progress, to evaluating the potential operational applications requirements will need to account for such risks. of BCI and other over-the-horizon military technol- Cybersecurity will be a significant risk going for- ogies. During the TTX testing, pairing of operational ward, amplified by the use of BCI. Because cyber- experience with technological expertise yielded rigor- networks touch nearly all dimensions of BCI, further ous and fruitful discussions, and this process should development of BCI capabilities will have to integrate be replicated on a larger scale. These approaches with associated cybersecurity measures. Our game could supplement existing internal , such insights suggested that, while human-to-human as the Marine Corps Advanced Naval Technology communication had the highest reward and greatest Exercise, to explore the practical utility of BCI and number of opportunities for use, it also presented the other prospective technologies to future warfighters. greatest operational and organizational risks. Risks By incorporating a disruptive and creative Red will depend on whether this capability can be turned team of RAND experts, the game also highlighted on and off and used selectively. The game highlighted

27 are further developed, robust testing for failure in noncombat scenarios, including training and data It is important to processing and analysis, before introducing them into combat will help to strengthen trust and reduce consider ethical and the potential for unanticipated risk. policy issues before Collaborate and Anticipate emerging technologies Research for this project highlighted multiple examples in which DoD seed funding for BCI mature and are laboratory research yielded successes. Significant future advances may take place in the private sector, disseminated. and the U.S. government should seek to leverage private-sector R&D when possible. If carefully pur- sued, private-sector advances may also improve trust potential new areas of operational vulnerability, as gaps within the military: As the U.S. public begins to well as initial ideas to mitigate them. As the U.S. use BCI, there may be less skepticism about its use in government seeks to build resilience from early a national security setting. As private-sector technol- phases of BCI development, similar methods could ogy advances and begins to be applied to the military help to unearth the full range of adversary threats. sphere, International Traffic in Arms Regulations, Beyond BCI, the approaches developed in this pilot Export Administration Regulations, and other project are scalable and could be applied to a variety restrictions should be considered with respect to BCI. of emerging technologies. BCI intellectual property should be carefully mon- itored by DoD and the Department of Commerce Address the Trust Deficit during these early periods of development. One major theme to emerge from the study was that As emerging technology accelerates, it becomes cultural barriers to BCI, particularly among infan- increasingly important to consider integrated systems try service members, are likely to be high, and this and how different technologies depend on one other. is a common theme with many new and emerging BCI could prove an important tool for integrating technologies. These barriers can be mitigated with human-machine systems, whether by enhancing big the following steps. data analysis, accelerating accurate decisionmak- As BCI capabilities are integrated into the ing, or improving the control of exoskeleton, drone force, they may initially be more readily accepted swarms, or semiautonomous systems. However, there among service members who already rely heavily on is a risk that the research could occur in isolation machine technologies, and who experience greater without consideration of additional and related requirements for direct interaction with computers or emerging technologies. Thus, current development machines. efforts should make provisions for the eventual avail- During the R&D process in the coming decades, ability of BCI, even if its applications are currently noninvasive measures are less likely to encounter still in the basic-research phase. cultural resistance. They may also be easier to reverse and control. Similarly, work on medical or thera- Plan Ahead for BCI Institutional Implications peutic applications may offer near-term benefit for As the U.S. government prepares to incorporate today’s wounded warriors and is likely to encounter BCI technologies into future military capabilities, the least cultural resistance. appropriate institutional planning will help to ensure Unsurprisingly, service members are more likely a smooth rollout and execution. It is important to to trust capabilities that have been appropriately vet- consider ethical and policy issues before emerging ted and tested before use. Thus, once BCI capabilities technologies mature and are disseminated.

28 During the research phase, it will be important Appendix. Game Design and to continue to integrate ethical, legal, and societal Execution considerations into research funding. DARPA cur- rently requires that research teams conduct ethical The table-top game was designed to answer the analysis for many grants surrounding BCI. Rigorous question, “Can brain-computer interfaces support internal analysis should continue beyond basic national security and future warfare and, if so, how?” research throughout the development, design, and This question is well-suited to a game for several application of new BCI technologies for defense and . It is fundamentally speculative in nature, national security use. The U.S. government should because there is no way to capture observations of continue to implement National Academy of Sciences future technological capabilities today. Thus, many ethical recommendations in development and empirical approaches, such as prototyping and field implementation, particularly regarding (1) questions experiments, are premature. In the absence of data of consent that are specific to service members, (2) from early fielding, games represent one of the few potential health implications for invasive BCI, ways to gather data about a new technology in a (3) considerations surrounding enhanced human specific setting. We were also interested in questions performance, and (4) potential risks to privacy. of human decisionmaking—when given the option As BCI technologies are disseminated across of using this technology, would players take it, or national security institutions, services will want to would they prefer to depend on more traditional identify clear oversight mechanisms for BCI devel- approaches? Perhaps more important, why were opment and application. Given the broad range of they making these decisions? Unlike modeling and potential applications for BCI, there is significant risk , which can provide a detailed sense of of stovepiping with related R&D. A department-wide what the technology could be capable of, games focus oversight mechanism, potentially residing in the on human decisionmaking and how it will ultimately Office of the Secretary of Defense or the Joint Staff, impact the usefulness of a new tool. Finally, games should track and review BCI developments for senior provide a forum for various experts to pool their DoD approval. Once BCI is integrated into services, collective understanding of a technology like BCI individual services might consider coordinated arbi- that has not been used in an operational setting. Our tration mechanisms outside the chain of command game allowed us to synthesize the experience of dif- to allow service members and their commanding ferent players to gain a fuller picture than one-on-one officers to discuss or object to unethical or harmful interviews alone could provide. uses of BCI technology. Finally, DoD may need to plan for a range of Vignette Selection additional warfighter and veteran care needs after the incorporation of BCI technologies. BCI carries Drawing from our literature review, we selected the potential for new dimensions of care require- tactical scenarios related to urban operations to ments, potentially including BCI withdrawal, brain provide a challenging setting for the game. Several injuries, posttraumatic stress disorder, and ongoing criteria drove this selection. First, urban operations care for invasive devices through Veterans Affairs. are an area of focus for the future force based on global trends. U.S. Army analysis has highlighted the urbanization of global populations and the rise of megacities as key trends that may yield potential requirements for urban combat.156 Both the Army and the Marine Corps are updating their urban oper- ations doctrine, and their leaderships have publicly advocated for more urban-focused efforts.157 Second, urban operations are an area where the value of technological superiority is often called

29 into question. As Joint Publication 3-06, Joint Urban of the BCI toolbox and the challenges associated Operations, notes, “Cities may reduce the advantages with the tactical vignette. First, players were intro- of the technologically superior force.”158 duced to seven capabilities or tools in the BCI toolbox Finally, the selection of the urban environment (described in detail in the “Testing BCI Capabilities played into previously identified cultural barriers. Through National Security Gaming” section). This Infantry from the Army and Marine Corps were provided participants with the opportunity to gain likely to be the most skeptical of the value of BCI a common understanding of the scope of the tech- during combat. Infantry has traditionally been the nology to ensure shared terms of reference. Each primary fighting force in the urban environment, so tool from the toolbox was represented by a poker setting the game there would provide a strong test chip with an icon illustrating the tool. Players were of some of the cultural barriers that might prevent then confronted with a tactical vignette and given effective deployment of BCI. the opportunity to discuss the core tasks associated In the limited scope of our pilot research, we with each of the joint warfighting functions: com- could examine only a handful of vignettes related mand and control, intelligence, fires, movement and to one type of operation. Given this restriction, maneuver, protection, and sustainment.161 Use of the we opted to adopt a logic similar to that of the joint warfighting functions ensured that participants “critical case.”159 If the game showed that BCI was considered a wide variety of problems, while group useful—even under what were projected to be discussion allowed for a consensus on the types of difficult settings—that would offer evidence for the tasks involved in each area. potential utility of BCI on the battlefield.160 Selecting The fourth and fifth steps focused on how play- a stressful scenario increased the likelihood of a ers thought BCI could be used. For each warfighting robust debate about how and why BCI was or was not function, players had selected one or two BCI tools useful, giving us richer data to explore. they would employ. Players indicated this choice by moving a poker chip with the tool’s symbol onto an Game Process individual “placemat,” which showed each warfight- ing function (an example is shown in step 4 of Our game was focused around players deciding what Figure A.1). Players had the option to not use any aspect or component of BCI technology they would BCI tool and instead depend on traditional solutions, use in a particular operational context. For each of which was indicated by placing no chip on the mat. two vignettes, the game moved through six phases: Players had only three chips for each BCI tool, so (1) get chips representing the BCI capabilities item- while each tool could be used more than once, players ized in the toolbox, (2) receive a tactical vignette, could not use the same tool to address all warfight- (3) discuss the tasks that must be completed in the ing functions. After individuals had had a chance to vignette, organized by warfighting functions, (4) as make their selections, in the next stage, they pre- individuals, select which, if any, BCI capabilities play- sented to the group and discussed key similarities ers would opt to employ, (5) as a group, discuss selec- and differences between their selections. This pro- tions to note areas of alignment or clarify areas of vided an opportunity to clarify differences in under- disagreement, and (6) receive feedback from experts standing and identify trends in player preferences. about how adversaries could exploit the BCI capabil- ities and discuss the risks of the approach developed in steps 4 and 5. Figure A.1 provides a visualization of the game process as a whole. The first three stages of the game were designed to allow players to develop a common understanding

30 FIGURE A.1 BCI TTX Process

1 Get x 3 chips per technology

2 Receive vignette about an urban operation

3 Discuss key tasks in each war ghting function

4 Indicate 1st and 2nd choice technology Preferred for each solution war ghting function with Alternative chips solution

5 Aggregate choices as a group and discuss bene ts of BCI

6 Red team presents response to BCI on the battle eld; players respond

31 Notes 19 J. J. Vidal, “Towards Direct Brain-Computer Communication,” Annual Review of Biophysics and Bioengineering, Vol. 2, 1973, 1 Nick Stockton, “Woman Controls a Fighter Jet Sim Using Only pp. 157–180. Her Mind,” Wired, March 5, 2015. 20 There are some nonelectrical signals that also are muffled. For 2 Michael Joseph Gross, “’s Push to Program Sol- example, functional near-infrared spectroscopy systems use light diers’ Brains,” The Atlantic, November 2018. to measure bold activity in the brain. 3 M. O. Krucoff, S. Rahimpour, M. W. Slutzky, V. R. Edgerton, 21 “The BRAIN Initiative,” homepage, undated-a. and D. A. Turner, “Enhancing Nervous System Recovery Through 22 Neurobiologics, Neural Interface Training, and Neurorehabilita- “What Is the BRAIN Initiative?” blog post, Chicago Council on tion,” Frontiers in Neuroscience, Vol. 10, December 2016, p. 584. Science and Technology, September 24, 2015. 23 4 William Eliason, “An Interview with Robert Work,” Joint Force Defense Advanced Research Projects Agency, “Towards a Quarterly 84, January 26, 2017. High-Resolution, Implantable Neural Interface,” press release, July 10, 2017. 5 Mark Pomerleau, “DoD’s Third Offset Strategy: What Man 24 and Machine Can Do Together,” Defense Systems, May 4, 2016; Patrick Tucker, “The Military Is Altering the Limits of Human Aaron Mehta, “Work Outlines Key Steps in Third Offset Tech Performance,” The Atlantic, July 14, 2017. Development,” Defense News, December 14, 2015; Jason Sherman, 25 Roger Mola, “Super Helmet: F-35 Pilots Get X-Ray Vision and “DoD Unveils Technology Areas That Will Drive ‘Third Offset’ Other Magic Powers,” Air & Space Magazine, September 2017. in Investments, Experimentation,” Inside Defense, December 9, 2015. 26 M. Hajinoroozi, Z. Mao, and Y. Huang, “Prediction of Driver’s Drowsy and Alert States from EEG Signals with Deep Learn- 6 Sydney Freedberg, “Centaur Army: Bob Work, Robotics, and ing,” 2015 IEEE Sixth International Workshop on Computational the Third Offset Strategy,” Breaking Defense blog, 2015. See also Advances in Multi-Sensor Adaptive Processing (CAMSAP), Can- Arwen H. DeCostanza, Amar R. Marathe, Addison Bohannon, A. cun, December 13–16, 2015, pp. 493–496; Sara Reardon, “Brain William Evans, Edward T. Palazzolo, Jason S. Metcalfe, and Kaleb Implants for Mood Disorders Tested in People,” Nature, Vol. 551, McDowell, “Enhancing Human–Agent Teaming with Individu- November 30, 2017, p. 549. alized, Adaptive Technologies: A Discussion of Critical Scientific 27 Questions,” Human Research and Engineering Directorate, Army S. Chen, “‘Forget the Facebook Leak’: China Is Mining Data Research Laboratory, May 2018. Directly from Workers’ Brains on an Industrial Scale,” South China Morning Post, April 29, 2018. 7 Tim Marler, “The American Intellectual Property,” Real Clear Defense, December 19, 2017. 28 Stockton, 2015; M. Kryger, B. Wester, E. A. Pohlmeyer, M. Rich, B. John, J. Beaty, M. McLoughlin, M. Boninger, and E. C. Tyler-Ka- 8 Pomerleau, 2016; Mehta, 2015; Sherman, 2015. bara, “Flight Simulation Using a Brain-Computer Interface: A 9 Daniel Cebul, “DARPA Wants to Connect Human Brains and Pilot, Pilot Study,” Experimental , Vol. 287, 2017, Machines,” TechWatch, March 19, 2018. pp. 473–478. 29 10 Sydney Freedberg, “Iron Man, Not Terminator: The Pentagon’s A. F. Salazar-Gomez, J. Del Preto, S. Gil, F. H. Guenther, and Sci Fi Inspirations,” Breaking Defense blog, May 3, 2016. D. Rus, “Correcting Robot Mistakes in Real Time Using EEG Signals,” 2017 IEEE International Conference on Robotics and 11 Freedberg, 2016. Automation (ICRA), Singapore, July 2017, pp. 6570–6577. 12 Defense Advanced Research Projects Agency, “Neural Engi- 30 Human-Oriented Robotics and Control Lab, “Research Project neering System Design (NESD) Program,” press release, January 1: Brain-Swarm Interaction and Control Interfaces,” webpage, 19, 2015. University of Delaware, undated; Karavas, George, Daniel T. Larsson, and Panagiotis Artemiadis, “A Hybrid Brain-Machine 13 A. F. Salazar-Gomez, J. Del Preto, S. Gil, F. H. Guenther, and Interface for Control of Robotic Swarms: Preliminary Results,” D. Rus, “Correcting Robot Mistakes in Real Time Using EEG 2017 IEEE/RSJ International Conference on Intelligent Robots Signals,” 2017 IEEE International Conference on Robotics and and Systems (IROS), Vancouver, Canada, September 24–28, 2017, Automation (ICRA), Singapore, July 2017. pp. 5065–5075; A. Beall, “The Drones That Fly Using Mind 14 C. S. Nam, A. Nijholt, and F. Lotte, Brain-Computer Interfaces Control: Swarms of UAVS Developed for the US Military Could Be Handbook, Boca Raton, Fla.: CRC Press, 2018. Guided by Brain Waves, Daily Mail, July 19, 2016. 15 The International BCI Society was founded in 2015. 31 J. Dearen, “Mind-Controlled Drones Race to the Future,” Asso- ciated Press, 2016. 16 The first issue of the Journal of Brain-Computer Interfaces was published in 2014. 32 Chethan Pandarinath, Paul Nuyujukian, Christine H. Blabe,

17 Brittany L. Sorice, Jad Saab, Francis R. Willett, Leigh R. Hochberg, Douglas Fox, “Brain Cells Communicate with Mechanical Krishna V. Shenoy, and Jaimie M. Henderson, “High Performance Scientific American Pulses, Not Electric Signals,” , April 1, 2018. Communication by People with Paralysis Using an Intracortical 18 H. Berger, “Ueber das Elektroenkephalogramm des Menschen,” Brain-Computer Interface,” eLife, Vol. 6, February 21, 2017. Archiv fur Psychiatrie und Nervenkrankheiten, Vol. 87, 1929, pp. 527–570.

32 33 Stanford News Center, “Brain-Computer Interface T. Hatmaker, “DARPA Awards $65 Million to Develop the Advance Allows Fast, Accurate Typing by People with Paralysis,” Perfect, Tiny Two-Way Brain-Computer Interface,” TechCrunch, press release, February 21, 2017. July 10, 2017. 34 A. Regalado, “Reversing Paralysis,” MIT Tech Review, 45 Al Emondi, “Neural Engineering System Design (NESD),” February 22, 2017; Marco Bonizzato, Galyna Pidpruzhnykova, Defense Advanced Research Projects Agency, webpage, undated. Jack Digiovanna, Polina Shkorbatova, Natalia Pavlova, Silvestro 46 Cebul, 2018. Micera, and Grégoire Courtine, “Brain-Controlled Modulation of Spinal Circuits Improves Recovery from ,” 47 R. P. N. Rao, A. Stocco, M. Bryan, D. Sarma, T. M. Youngq- Nature Communications, Vol. 9, No. 1, 2018, p. 3015. uist, J. Wu, and C. S. Prat, “A Direct Brain-to-Brain Interface in Humans,” PLOS ONE, Vol. 9, No. 11, November 5, 2014. 35 Regalado, 2017. 48 L. Jiang, A. Stocco, D. M. Losey, J. A. Abernethy, C. S. Prat, 36 M. Parman, “Robotic Arm Reaches Toward Success,” Johns and R. P. N. Rao, “BrainNet: A Multi-Person Brain-to-Brain Hopkins News-Letter, January 31, 2013; S. N. Flesher, J. L. Interface for Direct Collaboration Between Brains,” Scientific Collinger, S. T. Foldes, J. M. Weiss, J. E. Downey, E. C. Tyler-Ka- Reports, Vol. 9, 2019, p. 6115. bara, S. J. Bensmaia, A. B. Schwartz, M. L. Boninger, and R. A. Gaunt, “Intracortical Microstimulation of Human Somatosen- 49 M. Pais-Vieira, M. Lebedev, C. Kunicki, J. Wang, and M. A. L. sory Cortex,” Science Translational Medicine, Vol. 8, October 19, Nicolelis, “A Brain-to-Brain Interface for Real-Time Sharing of 2016. Sensorimotor Information,” Scientific Reports, Vol. 3, 2013, p. 1319. 37 E. A. Pohlmeyer, M. Fifer, M. Rich, J. Pino, B. Wester, M. Johannes, C. Dohopolski, J. Helder, D. D’Angelo, J. Beaty, S. 50 S. S. Yoo, H. Kim, E. Filandrianos, S. J. Taghados, and S. Bensmaia, M. McLoughlin, and F. Tenore, “Beyond Intuitive Park, “Non-Invasive Brain-to-Brain Interface (BBI): Establishing Anthropomorphic Control: Recent Achievements Using Brain Functional Links Between Two Brains,” PLOS ONE, Vol. 8, Computer Interface Technologies,” in T. George, A. K. Dutta, No. 4, 2013. and S. Islam, eds., Micro- and Sensors, Systems, 51 and Applications IX, Proceedings of SPIE (Society of Photo-opti- I. H. Stevenson and K. P. Kording, “How Advances in Neural cal Instrumentation ), 2017. Recording Affect Data Analysis,” Nature Neuroscience, Vol. 14, 2011, pp. 139–142. 38 S. Saproo, J. Faller, V. Shih, and P. Sajda, “Cortically Coupled 52 Computing: A New Paradigm for Synergistic Human-Machine Defense Advanced Research Projects Agency, 2017. Interaction,” Computer, Vol. 45, September 2016. 53 Samuel Moore, “Wireless Network Brings Dust-Sized Brain 39 Darrell Etherington, “Elon Musk Could Soon Share More Implants a Step Closer,” IEEE Spectrum, Institute of Electrical on His Plan to Help Humans Keep up with AI,” TechCrunch, and Electronics Engineers, May 14, 2019. January 25, 2017. 54 Paradromics, “A New Paradigm in Neural Recording,” press 40 A. Kharpal, “Elon Musk: Humans Must Merge with Machines release, January 7, 2020. or Become Irrelevant in AI Age,” CNBC Tech Transformers, 55 Chief Information Officer, “DoD Policy Recommendations February 13, 2017; N. Statt, “Elon Musk Launches Neuralink, a for the Internet of Things (IoT),” U.S. Department of Defense, Venture to Merge the Human Brain with AI,” The Verge, March Washington, D.C., 2016; M. Weisgerber, “A Short-Staffed Air 27, 2017b. Force Wants Robots to Do More Human Jobs,” Defense One, 41 DARPA, “Progress in Quest to Develop a Human Memory 2017. Prosthesis,” press release, March 28, 2018; Tristan McClure-Beg- 56 V. Machi, “Internet of Things to Provide Intelligence for ley, “Restoring Active Memory,” Defense Advanced Research Urban Warfare,” National Defense, January 22, 2018a. Projects Agency, webpage, undated a. 57 Gross, 2018. 42 “Prosthetic Memory System Successful in Humans,” Neuro- science News, Vol. 15, March 28, 2018; R. E. Hampson, 58 N. Statt, “Kernel Is Trying to Hack the Human Brain—but D. Song, B. S. Robinson, D. Fetterhoff, A. S. Dakos, B. M. Roeder, Neuroscience Has a Long Way to Go,” The Verge, February 22, X. She, R. T. Wicks, M. R. Witcher, D. E. Couture, A. W. Laxton, 2017a. H. Munger-Clary, G. Popli, M. J. Soliman, C. T. Whitlow, V. Z. 59 Statt, 2017b. Marmarelis, T. W. Berger, and S. A. Deadwyler, “Developing a Hippocampal Neural Prosthetic to Facilitate Human Memory 60 Paradromics, homepage, undated. and ,” Journal of Neural Engineering, Vol. 15, 61 S. Samuel, “Facebook Is Building Tech to Read Your Mind. No. 3, June 2018. The Ethical Implications Are Staggering,” Vox, August 5, 2019. 43 H. Zhao, L. Qiao, D. Fan, S. Zhang, O. Turel, Y. Li, J. Li, G. 62 Stephan Waldert, “Invasive vs. Non-Invasive Neuronal Signals Xue, A. Chen, and Q. He, “Modulation of Brain Activity with for Brain-Machine Interfaces: Will One Prevail?” Frontiers in Noninvasive Transcranial Direct Current Stimulation (tDCS): Neuroscience, Vol. 10, June 27, 2016, p. 295. Clinical Applications and Safety Concerns,” Frontiers in Psy- chology, Vol. 8, 2017, p. 685. 63 Cebul, 2018. 44 P. Dockril, “The US Military Is Working on Tech That Could 64 Lawrence Freedman, The Future of War: A History, New Turn Soldiers into ,” Science Alert, January 22, 2016; York: Hachette Book Group, 2017, p. xxi.

33 65 Freedman, 2017. 82 Peng Yuan, Yijun Wang, Lian Gao, Tzyy-Ping Jung, and Xiaorong Gao, “A Collaborative Brain-Computer Interface for 66 John D. Winkler, Timothy Marler, Marek N. Posard, Raphael Accelerating Human Decision Making,” in C. Stephanidis and S. Cohen, and Meagan L. Smith, Reflections on the Future of M. Antona, eds., Universal Access in Human-Computer Inter- Warfare and Implications for Personnel Policies of the U.S. action: , Tools, and Interaction Techniques for Department of Defense, Santa Monica, Calif.: RAND Corpora- eInclusion, Lecture Notes in , Berlin: Springer, tion, PE-324-OSD, 2019. Vol. 8009, 2013, pp. 672–681. 67 Peter Roberts, “The Future Conflict Operating Environment 83 Jack Corrigan, “The Pentagon Wants to Bring Mind-Con- out to 2030,” Royal United Services Institute, Occasional Papers, trolled Tech to Troops,” Nextgov.com, July 17, 2018. June 3, 2019. 84 Carl Governale, “Brain-Computer Interfaces Are Game 68 “Future of Warfare,” Center for New American Security, Changers,” Real Clear Defense, August 31, 2017. website, undated. 85 Defense Advanced Research Projects Agency, “N3 Proposers 69 Cognition and Neuroergonomics Collaborative Technology Day,” press release, April 3, 2018. Alliance, "Vision,” webpage, undated. 86 Kania, 2017. 70 Jim Nash, “In the Fog of Battle Acoustic Sensors Pinpoint Gunfire by Measuring Air Movement” Scientific American, April 87 Amir Husain, Hyperwar: Conflict and Competition in the AI 30, 2013; Kelsey Atherton “Dedicated Drone Tracker Can Now Century, Austin, Tex.: SparkCognition, 2018. Detect Swarming Robots,” C4ISR.net, September 20, 2018; Lori 88 Scharre, 2018, p. 321. Aratani, “Officials Unveil New Facial Recognition System at Dulles International Airport,” Washington Post, September 7, 89 Corrigan, 2018. 2018; Jamie Condliffe, “Big Investors Are Placing Bets on China’s 90 Facial Recognition Start-ups” New York Times, July 24, 2018. P. W. Singer, Wired for War: The Robotics Revolution and Conflict in the Twenty-First Century, New York: Penguin Press, 71 Adam Stone, “The Internet of Battlefield Things Will Depend 2009, p. 126. on Modernized Networks,” C4ISR.net, August 3, 2018. 91 “Brain Swarm Interface (BSI)” fact sheet, Multi-robot Systems 72 Winkler et al., 2019. Lab, Boston University, undated. 73 Jim Lester and Russell Vieira, “U.S. Air Force Combat Cloud 92 Patrick Tucker, “Actual Telepathy Is One Step Closer to the Operating Concept,” March 2016, as cited in Aaron Kiser et Battlefield,” Defense One, September 2, 2014. al., The Combat Cloud: Multidomain Command and Control 93 Across the Range of Military Operations, Air Command and Staff Katie Drummond, “Pentagon Preps Soldier Telepathy Push,” College, Air University, Maxwell Air Force Base, Ala., March Wired, May 14, 2009. 2017. See also V. Machi, “Air Force Wants Platforms Built with 94 Patrick Cutter, “The Shape of Things to Come: The Military Connectivity in Mind,” National Defense, February 23, 2018b. Benefits of the Brain-Computer Interface in 2040,” research 74 Committee on Military and Intelligence Methodology for report for Master of Operational Arts and Sciences, Maxwell Air Emergent Neuropsychologic and Cognitive Neural Science Force Base, Ala., April 2015. Research in the Next Two Decades, Emerging Cognitive Neurosci- 95 Noah Shachman, Pentagon’s Mind Reading Computers Repli- ence and Related Technologies, Washington, D.C.: The National cate,” Wired, March 18, 2008. Academies Press, 2008. 96 Cognition and Neuroergonomics Collaborative Technology 75 Mark Pomerleau, “How Artificial Intelligence Went from an Alliance, undated. Advantage to a Worldwide Threat,” C4isr.net, April 4, 2018. 97 “Cognition and Neuroergonomics Collaborative Technology 76 “Robotic and Autonomous Systems Strategy,” Headquarters, Alliance August 2010,” briefing slides for U.S. Army Research Department of the Army, Washington, D.C., March 2017. Labs and Research, Development and Engineering Command, 77 Eliason, 2017. August 2010. 98 78 Elsa Kania, “Battlefield Singularity: Artificial Intelligence, D. D. Langleben, J. Loughead, W. B. Bilker, K. Ruparel, A. Military Revolution, and China’s Future Military Power,” Center Childress, S. I. Busch, and R. Gur, “Telling the Truth from Lie in for New American Security, November 2017, p. 29. Individual Subjects with Fast Event-Related fMRI,” Human , Vol. 26, No. 4, December 2005. 79 Paul Scharre, Army of None: Autonomous Weapons and the 99 Future of War, New York: W. W. Norton & Company, 2018, This potential application was identified in Committee on p. 326. Military and Intelligence Methodology for Emergent Neuropsy- chologic and Cognitive Neural Science Research in the Next Two 80 Todd South, “Drone Swarm Tactics Get Tryout for Infantry to Decades, 2008. Use in Urban Battlespace,” Army Times, January 8, 2018. 100 Jonathan Moreno, Mind Wars: Brain Science and the 81 Shawn Snow, “No Man’s Land: How New Tactics and Robots Military in the 21st Century, New York: Bellevue Literary Press, Will Push Marines Deeper into the Fight,” Marine Corps Times, 2012, p. 144. Moreno address the 2012 state of the art of a “brain August 8, 2018. prosthesis” chip, genetic alterations that could theoretically

34 enhance cognition, and electrical stimulation via transcranial 114 Marine Corps Warfighting Publication 12-1-B.1, Military direct current stimulation (tDCS) (pp. 143–149). Operations in Urban Terrain (MOUT), Washington, D.C.: Head- quarters U.S. Marine Corps, May 2, 2016, pp. C6–C8. 101 Murray Carpenter, “Inside the Crazy Lab Where the Army Spikes Its Rations with Caffeine,” Wired, March 14, 2014. 115 Elliot Ackerman, “Relearning Storm Troop Tactics: The Bat- tle of Fallujah,” Marine Corps Gazette, September 2006. 102 J. Nelson, R. A. McKinley, C. Phillips, L. McIntire, C. Good- year, A. Kreiner, and L. Monforton, “The Effects of Transcranial 116 Scott Cuomo, “Secretary Mattis’ ‘Guardian Angel’ and How Direct Current Stimulation (tDCS) on Multitasking Throughput Marine Corps Can Get Back on Target,” War on the Capacity,” Frontiers in Human Neuroscience, November 29, 2016. Rocks blog, July 20, 2017. 103 Defense Advanced Research Projects Agency, “Boosting 117 Ackerman, 2006. to Accelerate Learning,” press release, March 118 L. Andrews and J. MacDonald, “Five Costs of Military Inno- 16, 2016. vation,” War on the Rocks blog, February 18, 2016. 104 “Broad Agency Announcement: Targeted Neuroplasti- 119 Tucker, 2014. city Training (TNT),” Solicitation number DARPA-SN-16-20, DARPA Biological Technologies Office, April 1, 2016. 120 Thomas Grove, Julian Barnes, and Drew Hinshaw, “Russia Targets NATO Soldier Smartphones, Western Officials Say,” 105 Eric Van Gieson, “Electrical Prescriptions (ElectRx),” Wall Street Journal, October 4, 2017; Ellen Nakashima and Paul Defense Advanced Research Projects Agency, website, undated. Sonne, “China Hacked a Navy Contractor and Secured a Trove Patrick Cutter also makes this point on p. 24 of Cutter, 2015. of Highly Sensitive Data on Submarine Warfare,” Washington 106 Ben Connable, Michael J. McNerney, William Marcellino, Post, June 8, 2018. Aaron Frank, Henry Hargrove, Marek N. Posard, S. Rebecca 121 William J Broad, Zimmerman, Natasha Lander, Jasen J. Castillo, James Sladden, “Microwave Weapons Are Prime Suspect New York Times Anika Binnendijk, Elizabeth M. Bartels, Abby Doll, Rachel in Ills of U.S. Embassy Workers,” , September ; Emily Rauhala and Carol Morello, “State Department Tecott, Benjamin J. Fernandes, Niklas Helwig, Giacomo Persi 1, 2018 Warns U.S. Citizens in China After Employee Suffers Possible Paoli, Krystyna Marcinek, and Paul Cornish, Will to Fight: Sonic Attack,” Washington Post, May 23, 2018. Returning to the Human Fundamentals of War, Santa Monica, Calif.: RAND Corporation, RB-10040-A, 2019. 122 “How Medical Devices Like Pacemakers and Insulin Pumps Can Be Hacked,” CBS News, November 8, 2018. 107 Potential elimination of fear through neuroscience is addressed in Moreno, 2012, pp. 149–151. 123 “Sinister Text Messages Reveal High-tech Front in War,” Voice of America News, May 11, 2017. 108 Gross, 2018. 124 Shawn Snow, “Enhancing the Grunt: Sophisticated New 109 The TNT program seeks to “boost the neurochemical Tech Means Greater Responsibility, Heavier Load,” Marine signaling in the brain that mediates neural plasticity and Corps Times, March 2018. facilitates long-term retention of new cognitive skills.” Tristan McClure-Begley, “Targeted Neuroplasticity Training (TNT),” 125 L. Niemela, “What Puts the ‘Yuck’ in the ‘Yuck Factor,’” Defense Advanced Research Projects Agency, webpage, undated Bioethics, Vol. 25, No. 5, June 2011, pp. 267–279. b. 126 Matthew Cox, “The Army Wants Autonomous Aviation 110 J. J. Vidal, “Tomorrow Soldier: The Military Is Altering the Tech. But Do Pilots Trust It?” Military.com, September 6, 2018. Limits of Human Performance,” Defense One, July 12, 2017; 127 Y. Tadjdeh, “SOCOM’s Top 10 Tech Needs—Biotechnology,” “Thousands of Swedes Are Inserting Microchips Under Their National Defense, May 2018. Skin,” National Public Radio, All Things Considered, October 22, 2018. 111 Brian Wang, “TALOS Special Ops Exoskeleton Today and 128 Plans for Exoskeleton Divisions in 2030s,” nextBIG Future, Singer, 2009, p. 335. November 5, 2018. 129 Julie Carpenter, Culture and Human-Robot Interaction in 112 Bing West, “Urban Warfare, Then and Now,” The Atlantic, Militarized Spaces: A War Story, Abingdon, UK: Routledge, 2016. June 30, 2018, as cited in Todd South, “The Future Battlefield: 130 Singer, 2009, p. 335. Army, Marines Prepare for ‘Massive’ Fight in Megacities,” Mili- tary Times, March 6, 2018b. 131 Ellen M. McGee and G. Q. Maguire, Jr., “Ethical Assessment of Implantable Brain Chips,” paper presented to the Twentieth 113 As described in more detail in the appendix, players always World Congress of , Boston, Mass., August 10–15, had the option to play an unspecified, non-BCI technology. Play- 1998. ers were told to assume that this chip represented an extension of current approaches—for example, the C2 function might be 132 BRAIN Initiative, “Neuroethics Working Group,” webpage, provided by an upgraded radio with greater range and clarity. undated-b. However, because the technological alternative was not fully 133 National Research Council and National Academy of fleshed out for players, there is some risk that this approach may Engineering, Emerging and Readily Available Technologies and have biased players toward using BCI technologies. National Security: A Framework for Addressing Ethical, Legal,

35 and Societal Issues, Washington, D.C.: The National Academies 149 Heather M. Roff and Richard Moyes, “Meaningful Human Press, 2014. Control, Artificial Intelligence and Autonomous Weapons,” briefing paper prepared for the Informal Meeting of Experts 134 National Research Council and National Academy of Engi- on Lethal Autonomous Weapons Systems, UN Convention on neering, 2014. Certain Conventional Weapons, April 2016. 135 National Research Council and National Academy of Engi- 150 Carl Governale, “Brain-Computer Interfaces Are Game neering, 2014, p. 75. Changers,” Real Clear Defense, August 31, 2017. 136 Bernhard Sehm and Patrick Ragert, “Why Non-Invasive 151 Peter W. Singer, appearance on National Security and the Brain Stimulation Should Not Be Used in Military and Security Future of Warfare, CSPAN, March 18, 2018, 5:02 am–7:00 am Services,” Frontiers in Human Neuroscience, Vol. 7, September EDT; Megan Eckstein, “Interview: Rear Adm. Mike Manazir on 2013. Weaving the Navy’s New Kill Webs,” USNI News, October 3, 137 This point is highlighted in Moreno, 2012, p. 199. 2016. 138 National Research Council and National Academy of Engi- 152 Stockdale Center on Ethical Leadership, U.S. Naval Academy, neering, 2014, p. 74; Moreno, 2012, p. 156. “New Warriors and New Weapons: The Ethical Ramifications of Emerging Military Technologies,” report of the 2010 McCain 139 Sehm and Ragert, 2013; M. N. Tennison and J. D. Moreno, conference, Annapolis, Md., April 23, 2010, cited in Moreno, “Neuroscience, Ethics, and National Security: The State of the 2012, pp. 150–151. Art,” PLoS Biology, Vol. 10, 2012. 153 The right temporo-parietal junction; National Research 140 National Research Council and National Academy of Engi- Council and National Academy of Engineering, 2014, p. 77. neering, 2014, p. 53. 154 National Research Council and National Academy of Engi- 141 K. Brukamp and D. Gross, “—A Contro- neering, 2014, p. 77. versial Topic in Contemporary Medical Ethics,” in Contemporary Issues in Bioethics, D. P. A. Clark, ed., Rijeka, Croatia: InTech, 155 Ash Carter, “America Needs to Align Technology with a Pub- 2012, p. 48. Broader philosophical considerations surrounding lic Purpose,” Belfer Center for Science and International Affairs human identity and future concepts of “mind-machine merger” Magazine, November 25, 2018. are addressed in Susan Schneider, Artificial You: AI and the 156 Army Capabilities Integration Center, The Megacity: Opera- Future of Your Mind, Princeton, N.J.: Princeton University Press, tional Challenges for Force 2025 and Beyond, Army Chief of Staff’s 2019. Future Study Plan, Washington, D.C.: Headquarters Department 142 Marcello Ienca and Roberto Andorno, “Towards New of the Army, 2014, cited in Gian Gentile, David E. Johnson, Lisa Human Rights in the Age of Neuroscience and Neurotechnol- Saum-Manning, Raphael S. Cohen, Chara Williams, Carrie ogy,” Life Sciences, Society and Policy, Vol. 13, No. 5, 2017. Lee, Michael Shurkin, Brenna Allen, Sarah Soliman, and James L. Doty III, Reimagining the Character of Urban Operations for 143 J. Brunelin, J. Levasseur-Moreau, and S. Fecteau, “Is It Ethical the U.S. Army: How the Past Can Inform the Present and Future, and Safe to Use Non-Invasive Brain Stimulation as a Cognitive Santa Monica, Calif.: RAND Corporation, RR-1602-A, 2017. and Motor Enhancer Device for Military Services? A Reply to Sehm and Ragert,” Frontiers in Human Neuroscience, Vol. 7, 2013. 157 South, 2018b; Gidget Fuentes, “Marines with New Technology Concepts for the Urban Battlefield,” USNI 144 Corrigan, 2018. News, March 26, 2018. 145 “Ethics and Autonomous Weapon Systems: An Ethical Basis 158 U.S. , Joint Urban Operations, Joint Pub- for Human Control?” International Committee of the Red Cross, lication 3-06, Washington, D.C., November 20, 2013, pp. vii–viii, April 2018. cited in Gentile et al., 2017, p. 14. 146 Singer, 2009, p. 408. 159 Robert K. Yin, Case Study Research: Design and Methods, 2nd 147 Department of Defense Directive 3000.09, Autonomy in ed., Thousand Oaks, Calif.: Sage Publishing, 1994, pp 38–39. Weapon Systems, November 21, 2012; “A Human Touch: Autono- 160 Conversely, had the game not supported the hypothesis, it mous Weapons, DoD Directive 3000.09 and the Interpretation of was only the maximalist version of the hypothesis that would ‘Appropriate Levels of Human Judgment over the Use of Force,’” have been rejected—that is, that BCI might not be useful in all in N. Bhuta, S. Beck, R. Geiβ, H. Liu, and C. Kreβ, eds., Auton- settings but may offer enough utility in select areas to be worth omous Weapons Systems: Law, Ethics, Policy, Cambridge, UK: investment. However, we felt at this stage of exploration that Cambridge University Press, 2016, pp. 185–208. evidence about where a technology might not be appropriate was 148 “Ethics and Autonomous Weapon Systems: An Ethical Basis likely to be quite valuable in setting reasonable expectations. for Human Control?” 2018. 161 U.S. Joint Chiefs of Staff, Joint Operations, Joint Publication 3-0, Washington, D.C., January 17, 2017, pp. xiii–xix.

36 References Chen, S., “‘Forget the Facebook Leak’: China Is Mining Data Directly from Workers’ Brains on an Industrial Scale,” South Ackerman, Elliot, “Relearning Storm Troop Tactics: The Battle of China Morning Post, April 29, 2018. Fallujah,” Marine Corps Gazette, September 2006. Chief Information Officer, “DoD Policy Recommendations Andrews, L., and J. MacDonald, “Five Costs of Military for the Internet of Things (IoT),” U.S. Department of Defense, Innovation,” War on the Rocks, February 18, 2016. Washington, D.C., 2016. Aratani, Lori, “Officials Unveil New Facial Recognition System Chicago Council on Science and Technology, “What Is the at Dulles International Airport,” Washington Post, September 7, BRAIN Initiative?” blog post, September 24, 2015. As of May 5, 2018. 2020: https://www.c2st.org/what-is-the-brain-initiative/ Army Capabilities Integration Center, The Megacity: Operational Challenges for Force 2025 and Beyond, Army Chief of Staff’s Cognition and Neuroergonomics Collaborative Technology Future Study Plan, Washington, D.C.: Headquarters Department Alliance, "Vision,” webpage, undated. As of May 4, 2020: of the Army, 2014. https://www.cancta.net//vision.shtml Atherton, Kelsey, “Dedicated Drone Tracker Can Now Detect ———, “Cognition and Neuroergonomics Collaborative Swarming Robots,” C4ISR.net, September 20, 2018. Technology Alliance August 2010,” briefing slides for U.S. Army Research Labs and Research, Development and Engineering Beall, A., “The Drones That Fly Using Mind Control: Swarms of Command, August 2010. UAVS Developed for the US Military Could Be Guided by Brain Waves, Daily Mail, July 19, 2016. Committee on Military and Intelligence Methodology for Emergent Neuropsychologic and Cognitive Neural Science Berger, H. “Ueber das Elektroenkephalogramm des Menschen,” Research in the Next Two Decades, Emerging Cognitive Archiv fur Psychiatrie und Nervenkrankheiten, Vol. 87, 1929, Neuroscience and Related Technologies, Washington, D.C.: The pp. 527–570. National Academies Press, 2008. Bonizzato, Marco, Galyna Pidpruzhnykova, Jack Digiovanna, Condliffe, Jamie, “Big Investors Are Placing Bets on China’s Polina Shkorbatova, Natalia Pavlova, Silvestro Micera, and Facial Recognition Start-ups” New York Times, July 24, 2018. Grégoire Courtine, “Brain-Controlled Modulation of Spinal Circuits Improves Recovery from Spinal Cord Injury,” Nature Connable, Ben, Michael J. McNerney, William Marcellino, Communications, Vol. 9, No. 1, 2018, p. 3015. Aaron Frank, Henry Hargrove, Marek N. Posard, S. Rebecca Zimmerman, Natasha Lander, Jasen J. Castillo, James Sladden, BRAIN Initiative, homepage, undated-a. As of May 4, 2020: Anika Binnendijk, Elizabeth M. Bartels, Abby Doll, Rachel http://www.braininitiative.org/ Tecott, Benjamin J. Fernandes, Niklas Helwig, Giacomo Persi ———, “Neuroethics Working Group,” webpage, undated-b. As of Paoli, Krystyna Marcinek, and Paul Cornish, Will to Fight: February 23, 2020: Returning to the Human Fundamentals of War, Santa Monica, https://braininitiative.nih.gov/about/neuroethics-working-group Calif.: RAND Corporation, RB-10040-A, 2019. As of May 8, 2020: Broad, William J., “Microwave Weapons Are Prime Suspect in https://www.rand.org/pubs/research_briefs/RB10040.html Ills of U.S. Embassy Workers,” New York Times, September 1, 2018. Corrigan, Jack, “The Pentagon Wants to Bring Mind-Controlled Tech to Troops,” Nextgov.com, July 17, 2018. As of May 8, 2020: Brukamp, K., and D. Gross, “Neuroenhancement—A https://www.nextgov.com/emerging-tech/2018/07/pentagon- Controversial Topic in Contemporary Medical Ethics,” in wants-bring-mind-controlled-tech-troops/149776/ Contemporary Issues in Bioethics, D. P. A. Clark, ed., Rijeka, Croatia: InTech, 2012. Cox, Matthew, “The Army Wants Autonomous Aviation Tech. But Do Pilots Trust It?” Military.com, September 6, 2018. Brunelin, J., J. Levasseur-Moreau, and S. Fecteau, “Is It Ethical and Safe to Use Non-Invasive Brain Stimulation as a Cognitive Cuomo, Scott, “Secretary Mattis’ ‘Guardian Angel’ and How and Motor Enhancer Device for Military Services? A Reply to Marine Corps Aviation Can Get Back on Target,” War on the Sehm and Ragert,” Frontiers in Human Neuroscience, Vol. 7, Rocks, July 20, 2017. 2013, p. 874. Cutter, Patrick, “The Shape of Things to Come: The Military Carpenter, Julie, Culture and Human-Robot Interaction in Benefits of the Brain-Computer Interface in 2040,” research Militarized Spaces: A War Story, Abingdon, UK: Routledge, 2016. report for Master of Operational Arts and Sciences, Maxwell Air Force Base, Ala., April 2015. Carpenter, Murray, “Inside the Crazy Lab Where the Army Spikes Its Rations with Caffeine,” Wired, March 14, 2014. DARPA—See Defense Advanced Research Projects Agency. Carter, Ash, “America Needs to Align Technology with a Public Dearen, J., “Mind-Controlled Drones Race to the Future,” Purpose,” Belfer Center for Science and International Affairs Associated Press, 2016. Magazine, November 25, 2018. DeCostanza, Arwen H., Amar R. Marathe, Addison Bohannon, Cebul, Daniel, “DARPA Wants to Connect Human Brains and A. William Evans, Edward T. Palazzolo, Jason S. Metcalfe, Machines,” TechWatch, March 19, 2018. and Kaleb McDowell, “Enhancing Human-Agent Teaming with Individualized, Adaptive Technologies: A Discussion of Center for New American Security, “Future of Warfare,” website, Critical Scientific Questions,” Human Research and Engineering undated. As of January 8, 2019: Directorate, Army Research Laboratory, May 2018. https://www.cnas.org/research/future-of-warfare

37 Defense Advanced Research Projects Agency, “Neural Gentile, Gian, David E. Johnson, Lisa Saum-Manning, Raphael Engineering System Design (NESD) Program,” press release, S. Cohen, Chara Williams, Carrie Lee, Michael Shurkin, Brenna January 19, 2015. As of May 5, 2020: Allen, Sarah Soliman, and James L. Doty III, Reimagining the https://www.darpa.mil/news-events/2015-01-19 Character of Urban Operations for the U.S. Army: How the Past Can Inform the Present and Future, Santa Monica, Calif.: RAND ———, “Boosting Synaptic Plasticity to Accelerate Learning,” Corporation, RR-1602-A, 2017. As of May 5, 2020: press release, March 16, 2016. As of May 4, 2020: https://www.rand.org/pubs/research_reports/RR1602.html https://www.darpa.mil/news-events/2016-03-16 Governale, Carl, “Brain-Computer Interfaces Are Game ———, “Towards a High-Resolution, Implantable Neural Changers,” Real Clear Defense, August 31, 2017. Interface,” press release, July 10, 2017. Gross, Michael Joseph, “The Pentagon’s Push to Program ———, “Progress in Quest to Develop a Human Memory Soldiers’ Brains,” The Atlantic, November 2018. Prosthesis,” press release, March 28, 2018. Grove, Thomas, Julian Barnes, and Drew Hinshaw, “Russia ———, “N3 Proposers Day,” press release, April 3, 2018. Targets NATO Soldier Smartphones, Western Officials Say,” Wall Department of Defense Directive 3000.09, Autonomy in Weapon Street Journal, October 4, 2017. Systems, November 21, 2012. Hajinoroozi, M., Z. Mao, and Y. Huang, “Prediction of Driver’s Dockril, P., “The US Military Is Working on Tech That Could Drowsy and Alert States from EEG Signals with Deep Learning,” Turn Soldiers into Cyborgs,” Science Alert, January 22, 2016. 2015 IEEE Sixth International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP), Drummond, Katie, “Pentagon Preps Soldier Telepathy Push,” Cancun, December 13–16, 2015, pp. 493–496. Wired, May 14, 2009. Hampson, R. E., D. Song, B. S. Robinson, D. Fetterhoff, Eckstein, Megan, “Interview: Rear Adm. Mike Manazir on A. S. Dakos, B. M. Roeder, X. She, R. T. Wicks, M. R. Witcher, Weaving the Navy’s New Kill Webs,” USNI News, October 3, D. E. Couture, A. W. Laxton, H. Munger-Clary, G. Popli, 2016. M. J. Soliman, C. T. Whitlow, V. Z. Marmarelis, T. W. Berger, and S. A. Deadwyler, “Developing a Hippocampal Neural Prosthetic Eliason, William, “An Interview with Robert Work,” Joint Force to Facilitate Human Memory Encoding and Recall,” Journal of Quarterly 84, January 26, 2017. As of May 6, 2020: Neural Engineering, Vol. 15, No. 3, June 2018. https://ndupress.ndu.edu/Media/News/Article/1038783/an- interview-with-robert-o-work/ Hatmaker, T., “DARPA Awards $65 Million to Develop the Perfect, Tiny Two-Way Brain-Computer Interface,” TechCrunch, Emondi, Al, “Neural Engineering System Design (NESD),” July 10, 2017. Defense Advanced Research Projects Agency, webpage, undated. As of May 7, 2020: “How Medical Devices Like Pacemakers and Insulin Pumps Can https://www.darpa.mil/program/neural-engineering-system- Be Hacked,” CBS News, November 8, 2018. design Human-Oriented Robotics and Control Lab, “Research Project 1: Etherington, Darrell, “Elon Musk Could Soon Share More on His Brain-Swarm Interaction and Control Interfaces,” University of Plan to Help Humans Keep up with AI,” TechCrunch, January 25, Delaware, webpage, undated. As of May 6, 2020: 2017. https://sites.udel.edu/horclab/research-bsi/ Flesher, S. N., J. L. Collinger, S. T. Foldes, J. M. Weiss, J. E. Husain, Amir, Hyperwar: Conflict and Competition in the AI Downey, E. C. Tyler-Kabara, S. J. Bensmaia, A. B. Schwartz, M. Century, Austin, Tex.: SparkCognition, 2018. L. Boninger, and R. A. Gaunt, “Intracortical Microstimulation of Human Somatosensory Cortex,” Science Translational Medicine, Ienca, Marcello, and Roberto Andorno, “Towards New Human Vol. 8, October 19, 2016. Rights in the Age of Neuroscience and Neurotechnology,” Life Sciences, Society and Policy, Vol. 13, No. 5, 2017. Fox, Douglas, “Brain Cells Communicate with Mechanical Pulses, Not Electric Signals,” Scientific American, April 1, 2018. International Committee of the Red Cross, “Ethics and Autonomous Weapon Systems: An Ethical Basis for Human Freedberg, Sydney, “Centaur Army: Bob Work, Robotics, and the Control?” April 2018. Third Offset Strategy,” Breaking Defense blog, 2015. Jiang, L., A. Stocco, D. M. Losey, J. A. Abernethy, C. S. Prat, and ———, “Iron Man, Not Terminator: The Pentagon’s Sci Fi R. P. N. Rao, “BrainNet: A Multi-Person Brain-to-Brain Interface Inspirations,” Breaking Defense blog, May 3, 2016. for Direct Collaboration Between Brains,” Scientific Reports, Vol. 9, 2019, p. 6115. Freedman, Lawrence, The Future of War: A History, New York: Hachette Book Group, 2017. Kania, Elsa, “Battlefield Singularity: Artificial Intelligence, Military Revolution, and China’s Future Military Power,” Center Fuentes, Gidget, “Marines Experiment with New Technology for New American Security, November 2017. Concepts for the Urban Battlefield,” USNI News, March 26, 2018. As of May 14, 2020: Karavas, George, Daniel T. Larsson, and Panagiotis Artemiadis, https://news.usni.org/2018/03/26/marines-experiment-new- “A Hybrid Brain-Machine Interface for Control of Robotic technology-concepts-urban-battlefield Swarms: Preliminary Results,” 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Vancouver, Canada, September 24–28, 2017, pp. 5065–5075. Kharpal, A., “Elon Musk: Humans Must Merge with Machines or Become Irrelevant in AI Age,” CNBC Tech Transformers, February 13, 2017.

38 Kiser, Aaron, Jacob Hess, Mostafa El Bouhafa, and Shawn Multi-robot Systems Lab, Boston University, “Brain Swarm Williams, “The Combat Cloud: Enabling Multi-Domain Interface (BSI)” fact sheet, undated. As of May 8, 2020: Command and Control Across the Range of Military http://sites.bu.edu/msl/research/brain-swarm-interface-bsi/ Operations,” Air Command and Staff College, Air University, Maxwell Air Force Base, Ala., March 2017. Nakashima, Ellen, and Paul Sonne, “China Hacked a Navy Contractor and Secured a Trove of Highly Sensitive Data on Krucoff, M. O., S. Rahimpour, M. W. Slutzky, V. R. Edgerton, Submarine Warfare,” Washington Post, June 8, 2018. and D. A. Turner, “Enhancing Nervous System Recovery Through Neurobiologics, Neural Interface Training, and Nam, C. S., A. Nijholt, and F. Lotte, Brain-Computer Interfaces ,” Frontiers in Neuroscience, Vol. 10, Handbook, Boca Raton, Fla.: CRC Press, 2018. December 2016, p. 584. As of May 5, 2020: Nash, Jim, “In the Fog of Battle Acoustic Sensors Pinpoint https://doi.org/10.3389/fnins.2016.00584 Gunfire by Measuring Air Movement” Scientific American, April Kryger, M., B. Wester, E. A. Pohlmeyer, M. Rich, B. John, J. Beaty, 30, 2013. M. McLoughlin, M. Boninger, and E. C. Tyler-Kabara, “Flight National Research Council and National Academy of Simulation Using a Brain-Computer Interface: A Pilot, Pilot Engineering, Emerging and Readily Available Technologies and Study,” Experimental Neurology, Vol. 287, 2017, pp. 473–478. National Security: A Framework for Addressing Ethical, Legal, Langleben, D. D., J. Loughead, W. B. Bilker, K. Ruparel, A. and Societal Issues, Washington, D.C.: The National Academies Childress, S. I. Busch, and R. Gur, “Telling the Truth from Lie in Press, 2014. Individual Subjects with Fast Event-Related fMRI,” Human Brain Nelson, J., R. A. McKinley, C. Phillips, L. McIntire, C. Goodyear, Mapping, Vol. 26, No. 4, December 2005, pp. 262–272. A. Kreiner, and L. Monforton, “The Effects of Transcranial Lester, Jim, and Russell Vieira, “U.S. Air Force Combat Cloud Direct Current Stimulation (tDCS) on Multitasking Throughput Operating Concept,” March 2016, as cited in Aaron Kiser, Jacob Capacity,” Frontiers in Human Neuroscience, November 29, 2016. Hess, Mostafa El Bouhafa, and Shawn Williams, The Combat Niemela, L., “What Puts the ‘Yuck’ in the ‘Yuck Factor,’” Cloud: Multidomain Command and Control Across the Range Bioethics, Vol. 25, No. 5, June 2011, pp. 267–279. of Military Operations, Air Command and Staff College, Air University, Maxwell Air Force Base, Ala., March 2017. Pais-Vieira, M., M. Lebedev, C. Kunicki, J. Wang, and M. A. L. Nicolelis, “A Brain-to-Brain Interface for Real-Time Sharing of Machi, V., “Internet of Things to Provide Intelligence for Urban Sensorimotor Information,” Scientific Reports, Vol. 3, 2013. Warfare,” National Defense, January 22, 2018a. Pandarinath, Chethan, Paul Nuyujukian, Christine H. Blabe, ———, “Air Force Wants Platforms Built with Connectivity in Brittany L. Sorice, Jad Saab, Francis R. Willett, Leigh R. Mind,” National Defense, February 23, 2018b. Hochberg, Krishna V. Shenoy, and Jaimie M. Henderson, “High Marine Corps Warfighting Publication 12-1-B.1, Military Performance Communication by People with Paralysis Using Operations in Urban Terrain (MOUT), Washington, D.C.: an Intracortical Brain-Computer Interface,” eLife, Vol. 6, Headquarters U.S. Marine Corps, May 2, 2016 . February 21, 2017. Marler, Tim, “The American Intellectual Property,” Real Clear Paradromics, homepage, undated. As of May 7, 2020: Defense, December 19, 2017. https://paradromics.com/ McClure-Begley, Tristan, “Restoring Active Memory,” Defense ———, “A New Paradigm in Neural Recording,” press release, Advanced Research Projects Agency, webpage, undated a. As of January 7, 2020. May 5, 2020: Parman, M., “Robotic Arm Reaches Toward Success,” Johns https://www.darpa.mil/program/restoring-active-memory Hopkins News-Letter, January 31, 2013. ———, “Targeted Neuroplasticity Training (TNT),” Defense Pohlmeyer, Eric A., M. Fifer, M. Rich, J. Pino, B. Wester, M. Advanced Research Projects Agency, webpage, undated b. As of Johannes, C. Dohopolski, J. Helder, D. D’Angelo, J. Beaty, S. May 14, 2020: Bensmaia, M. McLoughlin, and F. Tenore, “Beyond Intuitive https://www.darpa.mil/program/targeted-neuroplasticity- Anthropomorphic Control: Recent Achievements Using Computer Interface Technologies,” in T. George, A. K. Dutta, McGee, Ellen M., and G. Q. Maguire, Jr., “Ethical Assessment and S. Islam, eds., Micro- and Nanotechnology Sensors, Systems, of Implantable Brain Chips,” paper presented to the Twentieth and Applications IX, Proceedings of SPIE (Society of Photo- World Congress of Philosophy, Boston, Mass., August 10–15, optical Instrumentation Engineers), 2017. 1998. Pomerleau, Mark, “DoD’s Third Offset Strategy: What Man and Mehta, Aaron, “Work Outlines Key Steps in Third Offset Tech Machine Can Do Together,” Defense Systems, May 4, 2016. Development,” Defense News, December 14, 2015. ———, “How Artificial Intelligence Went from an Advantage to a Mola, Roger, “Super Helmet: F-35 Pilots Get X-Ray Vision and Worldwide Threat,” C4isr.net, April 4, 2018. Other Magic Powers,” Air & Space Magazine, September 2017. “Prosthetic Memory System Successful in Humans,” Moore, Samuel, “Wireless Network Brings Dust-Sized Brain Neuroscience News, Vol. 15, March 28, 2018. Implants a Step Closer,” IEEE Spectrum, Institute of Electrical Rao, R. P. N., A. Stocco, M. Bryan, D. Sarma, T. M. Youngquist, and Electronics Engineers, May 14, 2019. J. Wu, and C. S. Prat, “A Direct Brain-to-Brain Interface in Moreno, Jonathan, Mind Wars: Brain Science and the Military in Humans,” PLOS ONE, Vol. 9, No. 11, November 5, 2014. the 21st Century, New York: Bellevue Literary Press, 2012. Rauhala, Emily, and Carol Morello, “State Department Warns U.S. Citizens in China After Employee Suffers Possible Sonic Attack,” Washington Post, May 23, 2018.

39 Reardon, Sara, “Brain Implants for Mood Disorders Tested in ———, “The Future Battlefield: Army, Marines Prepare for People,” Nature, Vol. 551, November 30, 2017, p. 549. ‘Massive’ Fight in Megacities,” Military Times, March 6, 2018b. As of May 14, 2020: Regalado, A., “Reversing Paralysis,” MIT Tech Review, https://www.militarytimes.com/news/your-army/2018/03/06/ February 22, 2017. the-future-battlefield-army-marines-prepare-for-massive-fight- Roberts, Peter, “The Future Conflict Operating Environment in-megacities/ out to 2030,” Royal United Services Institute, Occasional Papers, Stanford Medicine News Center, “Brain-Computer Interface June 3, 2019. Advance Allows Fast, Accurate Typing by People with Paralysis,” “Robotic and Autonomous Systems Strategy,” Headquarters, press release, February 21, 2017. As of May 4, 2020: Department of the Army, Washington, D.C., March 2017. https://med.stanford.edu/news/all-news/2017/02/brain- computer-interface-allows-fast-accurate-typing-by-people-with- Roff, Heather M., and Richard Moyes, “Meaningful Human paralysis.html Control, Artificial Intelligence and Autonomous Weapons,” briefing paper prepared for the Informal Meeting of Experts Statt, N., “Kernel Is Trying to Hack the Human Brain—but on Lethal Autonomous Weapons Systems, UN Convention on Neuroscience Has a Long Way to Go,” The Verge, February 22, Certain Conventional Weapons, April 2016. 2017a. Salazar-Gomez, A. F., J. Del Preto, S. Gil, F. H. Guenther, and ———, “Elon Musk Launches Neuralink, a Venture to Merge the D. Rus, “Correcting Robot Mistakes in Real Time Using EEG Human Brain with AI,” The Verge, March 27, 2017b. Signals,” 2017 IEEE International Conference on Robotics and Stevenson, I. H., and K. P. Kording, “How Advances in Neural Automation (ICRA), Singapore, July 2017, pp. 6570–6577. Recording Affect Data Analysis,” Nature Neuroscience, Vol. 14, Samuel, S., “Facebook Is Building Tech to Read Your Mind. The 2011, pp. 139–142. Ethical Implications Are Staggering,” Vox, August 5, 2019. Stockdale Center on Ethical Leadership, U.S. Naval Academy, Saproo, S., J. Faller, V. Shih, and P. Sajda, “Cortically Coupled “New Warriors and New Weapons: The Ethical Ramifications Computing: A New Paradigm for Synergistic Human-Machine of Emerging Military Technologies,” report of the 2010 McCain Interaction,” Computer, Vol. 45, September 2016. conference, Annapolis, Md., April 23, 2010. Saxon, D., “A Human Touch: Autonomous Weapons, DoD Stockton, Nick, “Woman Controls a Fighter Jet Sim Using Only Directive 3000.09 and the Interpretation of ‘Appropriate Levels of Her Mind,” Wired, March 5, 2015. Human Judgment over the Use of Force,’” in N. Bhuta, S. Beck, R. Stone, Adam, “The Internet of Battlefield Things Will Depend on Geiβ, H. Liu, and C. Kreβ, eds., Autonomous Weapons Systems: Modernized Networks,” C4ISR.net, August 3, 2018. Law, Ethics, Policy, Cambridge, UK: Cambridge University Press, 2016, pp. 185–208. Tadjdeh, Y., “SOCOM’s Top 10 Tech Needs—Biotechnology,” National Defense, May 2018. Scharre, Paul, Army of None: Autonomous Weapons and the Future of War, New York: W. W. Norton & Company, 2018. Tennison, M. N., and J. D. Moreno, “Neuroscience, Ethics, and National Security: The State of the Art,” PLoS Biology, Vol. 10, Schneider, Susan, Artificial You: AI and the Future of Your Mind, 2012. Princeton, N.J.: Princeton University Press, 2019. “Thousands of Swedes Are Inserting Microchips Under Their Sehm, Bernhard, and Patrick Ragert, “Why Non-Invasive Skin,” National Public Radio, All Things Considered, October 22, Brain Stimulation Should Not Be Used in Military and Security 2018. Services,” Frontiers in Human Neuroscience, Vol. 7, September 2013, p. 553. Tucker, Patrick, “Actual Telepathy Is One Step Closer to the Battlefield,” Defense One, September 2, 2014. Shachman, Noah, “Pentagon’s Mind Reading Computers Replicate,” Wired, March 18, 2008. ———, “The Military Is Altering the Limits of Human Performance,” The Atlantic, July 14, 2017. Sherman, Jason, “DoD Unveils Technology Areas That Will Drive ‘Third Offset’ in Investments, Experimentation,” Inside U.S. Joint Chiefs of Staff, Joint Urban Operations, Joint Defense, December 9, 2015. Publication 3-06, Washington, D.C., November 20, 2013. Singer, Peter W., appearance on National Security and the Future ———, Joint Operations, Joint Publication 3-0, Washington, D.C., of Warfare, CSPAN, March 18, 2018. January 17, 2017. ———, Wired for War: The Robotics Revolution and Conflict in Van Gieson, Eric, “Electrical Prescriptions (ElectRx),” Defense the Twenty-First Century, New York: Penguin Press, 2009. Advanced Research Projects Agency, website, undated. As of August 26, 2019: “Sinister Text Messages Reveal High-tech Front in Ukraine War,” https://www.darpa.mil/program/electrical-prescriptions Voice of America News, May 11, 2017. Vidal, J. J., “Towards Direct Brain-Computer Communication,” Snow, Shawn, “Enhancing the Grunt: Sophisticated New Tech Annual Review of Biophysics and Bioengineering, Vol. 2, 1973, Means Greater Responsibility, Heavier Load,” Marine Corps pp. 157–180. Times, March 2018a. ———, “Tomorrow Soldier: The Military Is Altering the Limits of ———, “No Man’s Land: How New Tactics and Robots Will Push Human Performance,” Defense One, July 12, 2017. Marines Deeper into the Fight,” Marine Corps Times, August 8, 2018b. Waldert, Stephan, “Invasive vs. Non-Invasive Neuronal Signals for Brain-Machine Interfaces: Will One Prevail?” Frontiers in South, Todd, “Drone Swarm Tactics Get Tryout for Infantry to Neuroscience, Vol. 10, June 27, 2016, p. 295. Use in Urban Battlespace,” Army Times, January 8, 2018a.

40 Wang, Brian, “TALOS Special Ops Exoskeleton Today and Winkler, John D., Timothy Marler, Marek N. Posard, Raphael S. Plans for Exoskeleton Divisions in 2030s,” nextBIG Future, Cohen, and Meagan L. Smith, Reflections on the Future of Warfare November 5, 2018. and Implications for Personnel Policies of the U.S. Department of Defense, Santa Monica, Calif.: RAND Corporation, PE-324-OSD, Weisgerber, M., “A Short-Staffed Air Force Wants Robots to Do 2019. As of May 7, 2020: More Human Jobs,” Defense One, 2017. https://www.rand.org/pubs/perspectives/PE324.html West, Bing, “Urban Warfare, Then and Now,” The Atlantic, June 30, 2018. As of May 14, 2020: https://www.theatlantic.com/international/archive/2017/06/urban- warfare-hue-mosul/532173/

41 Yuan, Peng, Yijun Wang, Lian Gao, Tzyy-Ping Jung, and Xiaorong with a focus on multiobjective optimization, training sim- Gao, “A Collaborative Brain-Computer Interface for Accelerating ulators and virtual environments, digital human modeling, Human Decision Making,” in C. Stephanidis and M. Antona, eds., Universal Access in Human-Computer Interaction: Design Methods, and emerging technology. Tools, and Interaction Techniques for eInclusion, Lecture Notes in Computer Science, Berlin: Springer, Vol. 8009, 2013, pp. 672–681. Elizabeth M. Bartels is an associate policy researcher and Yin, Robert K., Case Study Research: Design and Methods, 2nd ed., specialist in national security policy analysis gaming. Her Thousand Oaks, Calif.: Sage Publishing, 1994. work explores a wide range of strategic and operational Yoo, S. S., H. Kim, E. Filandrianos, S. J. Taghados, and S. Park, concerns, with a focus on games on novel topics and those “Non-Invasive Brain-to-Brain Interface (BBI): Establishing that integrate other analytical techniques. Functional Links Between Two Brains,” PLOS ONE, Vol. 8, No. 4, 2013. Zhao, H., L. Qiao, D. Fan, S. Zhang, O. Turel, Y. Li, J. Li, G. Xue, A. Chen, and Q. He, “Modulation of Brain Activity with Noninvasive Acknowledgments Transcranial Direct Current Stimulation (tDCS): Clinical Applications and Safety Concerns,” Frontiers in , Vol. 8, We thank Andrew Parasiliti, former director of RAND’s 2017, p. 685. Center for Global Risk and Security, for his support through the CGRS Security 2040 initiative. Our senior advisers, John Plumb and Rand Waltzman, provided sound advice About the Authors and feedback during the early phases of the project. We are indebted to Sonni Efron for her invaluable edits, to Anika Binnendijk is a political scientist at the RAND Matt DeNardo for his thorough review and feedback, and Corporation with a research focus on defense and national to Robin Meili for ushering our work from draft to pub- security issues, including decisionmaking, technology, lication. We also thank our reviewers, Joel Predd, senior allied modernization, and national resilience. Her prior and director of the National Security Research experience includes service at the Office of the Secretary Division Acquisition and Technology Policy Center at the of Defense and U.S. State Department Office of Policy RAND Corporation; Eric A. Pohlmeyer, senior researcher, Planning. Johns Hopkins University Applied Laboratory; and Tim Marler is a senior research engineer at the RAND Paul Scharre, senior fellow and director of the Technology Corporation and a professor at the Pardee RAND Graduate and National Security Program at the Center for a New School. His work revolves around modeling and simulation American Security, for their insight and feedback.

42 43 C O R P O R A T I O N

The RAND Corporation is a research organization that develops solutions About This Report to public policy challenges to help make communities throughout This report reviews the operational utility and policy considerations of the the world safer and more secure, potential future application of brain-computer interface technologies in military healthier and more prosperous. combat. This report is part of a broader effort, an initiative of RAND Ventures, RAND is nonprofit, nonpartisan, and to envision critical security challenges in the world of 2040, considering the committed to the public interest. effects of political, technological, social, and demographic trends that will shape those security challenges in the coming decades. The research was conducted RAND’s publications do not within the RAND Center for Global Risk and Security. necessarily reflect the opinions of its research clients and sponsors. is a registered trademark. The RAND Center for Global Risk and Security Limited Print and Electronic The Center for Global Risk and Security works across the RAND Corporation Distribution Rights to develop multidisciplinary research and policy analysis dealing with systemic This document and trademark(s) risks to global security. The center draws on RAND’s expertise to complement contained herein are protected and expand RAND research in many fields, including security, economics, by law. This representation of health, and technology. A board of distinguished business leaders, philanthro- RAND intellectual property is pists, and former policymakers advises and supports the center’s activities, provided for noncommercial use which are increasingly focused on global security trends and the impact of only. Unauthorized posting of this disruptive technologies on risk and security. For more information about the publication online is prohibited. RAND Center for Global Risk and Security, visit www.rand.org/international/ Permission is given to duplicate this cgrs. document for personal use only, as long as it is unaltered and complete. Permission is required from RAND Security 2040 to reproduce, or reuse in another This report is part of a RAND initiative to envision critical security challenges form, any of our research documents in the world of 2040, considering the effects of political, technological, social, for commercial use. For information on reprint and linking permissions, and demographic trends that will shape those security challenges in the coming please visit www.rand.org/pubs/ decades. The research was conducted within the RAND Center for Global Risk permissions. and Security. For more information on this publication, visit Funding www.rand.org/t/RR2996. Funding for this research and analysis was provided by gifts from RAND sup- © 2020 RAND Corporation porters and income from operations. Support for this project was also provided, in part, by the generous contributions of the RAND Center for Global Risk and Security (CGRS) Advisory Board. www.rand.org

RR-2996-CGRS C O R P O R A T I O N

The Department of Defense has invested in the development of technologies that allow the human brain to communicate directly with machines, including the development of implantable neural interfaces able to transfer data between the human brain and the digital world. This technology, known as brain-computer interface (BCI) may be used to monitor a soldier’s cognitive workload, control a drone swarm, or link with a prosthetic, among other examples. Further technological advances could support human-machine decision making, human-to- human communication, system control, performance enhancement and monitoring, and training. However, numerous policy, safety, legal, and ethical issues should be evaluated before the technology is widely deployed. This project developed a national security game to explore the use of BCI in future combat scenarios, convened experts in military operations, human performance, and neurology to explore how the technology might affect military tactics, which aspects may be most beneficial, and which aspects might present risks, and offered recommendations to policymakers. The project sought to assess current and potential BCI applications for the military to ensure that the technology responds to actual needs, practical , and legal and ethical considerations, in addition to the intentions of developers.

$24.00

ISBN-10 1-9774-0523-1 ISBN-13 978-1-9774-0523-4 52400

www.rand.org 9 781977 405234

RR-2996-RC