OPINION ARTICLE

published: 30 July 2014

SYSTEMS NEUROSCIENCE doi: 10.3389/fnsys.2014.00136

Ethical issues with brain-computer interfaces

Walter Glannon*
Department of Philosophy, Faculty of Arts, University of Calgary, Calgary, AB, Canada
*Correspondence: wglannon@ucalgary.ca

Edited by:
Mikhail Lebedev, Duke University, USA
Reviewed by:
Pim Haselager, Radboud University, Netherlands

Keywords: benefit-risk ratios, brain-computer interface, communication, expectation, auditory feeedback, visual feedback

INTRODUCTION caregivers’ expectations about recovering Ordinarily, motor skills are performed
Brain-computer interfaces (BCIs), or motor function with a BCI might not be unconsciously and automatically follow-
brain-machine interfaces (BMIs) involve reasonable given the cognitive challenges ing an initial period of conscious atten-
real-time direct connections between the in operating the system. This might result tion and learning. For those with severe
brain and a computer (Kubler, 2009; in psychological harm when the subject’s paralysis, however, sustained attention is
Wolpaw and Wolpaw, 2011). Bidirectional desires and intentions to produce actions required both while being trained to oper-
feedback between the user and the system fail to be realized. Second, the different ate the interface and effectively operating
produces physical changes that can restore types of electrodes used to detect and it to execute motor tasks. Subjects whose
some degree of motor or communicative respond to motor cortical neural signals cognitive capacity for planning has been
control for individuals with lost limbs, involve different levels of invasiveness and impaired by injury to the central nervous
extensive paralysis or who are significantly different benefit-risk ratios that have to be system may have difficulty in translating
neurologically compromised (Hochberg weighed with a view to the probable suc- their thoughts into actions or fail to do
et al., 2006, 2012). In these respects, a BCI cess or failure of the technique. Third, the so. Failure to meet the expectation to pro-
can enable an individual with severe brain use of a BCI for communication in neuro- duce certain actions may cause distress and
or bodily injury to regain some degree of logically compromised patients prompts harm in some subjects by defeating their
agency. By providing the subject with the the question of whether their responses interest in recovering some, albeit lim-
relevant type of feedback, the device may would be evidence of the capacity to make ited, degree of motor control. Planning is
enable her to translate an intention into informed decisions about their care. a critical component in moving a pros-
an action despite the inability to perform thetic limb, for example. The subject must
voluntary bodily movements. There are EXPECTATIONS indicate with his brain and mind where
two types of feedback with a BCI. The The user of a BCI can execute an intention the limb should go before executing the
first concerns feedback about the out- to perform a motor task through changes intention to move it. The cognitive work-
come of a self-initiated, BCI-mediated in the system caused by electrodes detect- load requires considerable time and effort.
action, such as moving a computer cursor ing signals in, for instance, the motor cor- This may cause frustration and anxiety
or robotic arm. It provides only indirect tex mediating the intention. Success in and increase the probability of failure for
feedback about brain activity. The second operating the system depends on a com- some in trying to achieve their goal. It
type concerns direct feedback about the bination of unconscious operant condi- can exacerbate the feeling of a loss of
level of brain activity itself. The first is tioning of brain responses and conscious behavior control. To minimize the prob-
more pertinent to the potential to restore goal-directed expectation of the subject. ability of harm, investigators and practi-
some behavior control in the sense that These depend in turn on how effective the tioners must educate users on the potential
one can perceive the success or failure of practitioner is in training the subject how positive effects and limits of BCIs. They
their mental act. Although it is still at an to operate the system. As in other cases should also adopt strict selection criteria
early stage of development, an EEG- or of traumatic brain injury, goal-directed and include only those with largely pre-
fMRI-based BCI might also enable mini- thinking in some patients with tetraple- served cognitive functions who could give
mally conscious individuals or those with gia may be impaired if there is signifi- informed consent and would more likely
complete locked-in syndrome to commu- cant damage to neural networks in frontal be trained to successfully operate it. This
nicate wishes about medical treatment regions mediating planning and decision- may seem unfair to those with impaired
when they are unable to do this verbally making. This may also impair the subject’s levels of cognition who lack these capac-
or gesturally (Sellers, 2013). These appli- capacity to understand the benefits and ities. Nevertheless, the idea of providing
cations of interface technology raise a risks of the technique and give informed equal opportunity for all paralyzed indi-
number of ethical issues (McCullagh et al., consent to participate in BCI research viduals to access to a BCI would have to
2014), three of which I will discuss in this and treatment (Hochberg and Cochrane, be weighed against the potential for emo-
article. First, in some cases patients’ and 2013). tional harm if a subject cannot meet the

Frontiers in Systems Neuroscience www.frontiersin.org July 2014 | Volume 8 | Article 136 | 1

Glannon Ethical issues with brain-computer interfaces

cognitive demands of operating the sys- they could also cause adverse changes in Some investigators have claimed that
tem and his expectations are not met. the surrounding tissue and result in neuro- fMRI-guided BCIs could enable minimally
Discriminating on the basis of levels of logical and psychological sequelae. A safe conscious patients with a high level of
cognitive function may be justified on and effective array that could function for cognitive function to make these deci-
these grounds. many years would be one in which the sur- sions (Peterson et al., 2013). But emotion-
rounding neuropil grew into the electrode. ally laden decisions about life-sustaining
BENEFITS AND RISKS This would be more stable and allow mye- treatment reflect a person’s values and
BCIs utilize wired or wireless systems to lated axons to be recorded using implanted attitudes about quality of life. It is ques-
detect and allow transmission of signals amplifiers (Kennedy et al., 2011). If this tionable whether these values and atti-
in the motor cortex into actions. The occurs, then invasive systems can be func- tudes can be expressed by simple “Yes”
significance of these systems for bene- tionally superior to and as safe as non- or “No” responses to questions (Monti
fit and risk to patients depends not so invasive systems. The first type can have a et al., 2010), and yet they have to be
much on the type used but their level more favorable benefit-risk ratio than the included in any robust sense of “com-
of invasiveness. Theoretically, the distinc- second. munication.” This involves more than
tion between wired and wireless systems is being aware, even fully aware. More
orthogonal to this level. The non-invasive COMMUNICATING WITH A BCI sophisticated interface systems enabling
type consists of scalp-based electrodes that EEG- and fMRI-based BCIs might enable the expression of complex semantic pro-
are part of the equipment required to individuals to reliably communicate when cessing may or may not confirm that
record EEG. Because they do not involve they are unable to communicate behav- the patient had the requisite capaci-
intracranial surgery and implantation of a iorally (Birbaumer et al., 2008, 2014). ties. Hochberg and Cudkowicz point out
device in the brain, they do not involve This involves three distinct patient groups. that among completely locked-in patients
a risk of infection or hemorrhage. At the Minimally conscious patients have resid- there have been “no reports of restor-
same time, though, they may not readily ual awareness of self and surround- ing communication using a neural signal-
read signals from the motor cortex because ings. Locked-in patients are fully aware based BCI in this most severely affected
the cranium can smear them. despite being almost completely paralyzed. population” (Hochberg and Cudkowicz,
In electrocorticography (ECoG), elec- Some of these patients can communi- 2014, p. 1852; Birbaumer et al., 2014).
trodes are implanted epidurally or subdu- cate through voluntary eyelid movements. Moreover, Fernandez-Espejo and Owen
rally (Leuthardt et al., 2004). These can These in turn are distinct from com- acknowledge that, with current interface
decode motor cortical signals more readily pletely locked-in patients who lack the technology, simple affirmative or nega-
than scalp-based electrodes because they capacity for any voluntary bodily move- tive responses to questions about whether
are not susceptible to cranial smearing. ments. Conscious perception and expres- a minimally conscious patient wanted to
But they entail some risk of infection and sion of intentions in locked-in patients continue living would not be sufficient to
hemorrhage. Like the non-invasive system, is different from that of minimally con- establish that the patient had the “cog-
both forms of ECoG BCIs impose con- scious patients, and this may better facil- nitive and emotional capacity to make
straints on the subjects’ freedom from the itate communication through a BCI. One such a complex decision” (Fernandez-
wires running from the electrodes to the challenge for this intervention would be Espejo and Owen, 2013, p. 808). But
machine. Wireless systems consisting of that BCIs typically utilize visual feed- they also say that “it is only a matter
a microelectrode array implanted in the back, and minimally conscious and com- of time before all of these obstacles are
motor cortex avoid this problem and are pletely locked-in subjects have limited or overcome” (p. 808).
less burdensome for subjects. Because they no capacity to receive feedback from and This last point may be overly optimistic.
can decode and transmit signals from this respond to a visual stimulus in learning Even advanced BCIs that could detect
region more directly, implanted arrays are how to operate the system. Alternatively, neural activity correlating with complex
more likely to facilitate the execution of tactile or auditory feedback could be semantic processing might not be suffi-
the subject’s intentions in actions. Still, used to enable communication (Kubler, cient to show that the subject had the cog-
this would depend on the specifics of 2009; Hochberg and Cudkowicz, 2014). nitive and emotional capacity to make an
the neurological deficit and the patient’s Yet even if this modality could over- informed and autonomous decision about
ability to manipulate the BCI. Moreover, come the limitations associated with a life-sustaining treatment. Some form of
in addition to the risk of infection and lack of visual feedback, questions would behavioral interaction may be necessary to
hemorrhage, microelectrode arrays raise remain about the meaning of “communi- confirm that the subject had this capacity.
the issue of biocompatibility between the cate.” Specifically, it is not clear whether Medical professionals and caregivers must
implanted objects and surrounding neu- the responses of linguistically impaired be cautious not to read too much into BCI-
ral tissue. The electrodes may reorganize minimally conscious or even fully con- enabled responses and interpret them as
and induce changes in the tissue. These scious locked-in patients would be evi- having a meaning they lack.
changes may be salutary, especially if they dence of the cognitive and emotional
promote neuroplasticity and the gener- capacity to give informed consent to con- CONCLUSION
ation of new neuronal connections that tinue or discontinue artificial hydration BCIs can benefit individuals by restor-
could bypass the site of brain or spinal cord and nutrition (Brady Wagner, 2003; Jox, ing varying degrees of motor control and
injury causing loss of motor function. But 2013). possibly the ability to communicate. But

Frontiers in Systems Neuroscience www.frontiersin.org July 2014 | Volume 8 | Article 136 | 2

Glannon Ethical issues with brain-computer interfaces

expectations of some subjects and their REFERENCES Tononi (Amsterdam: Elsevier), 217–233. doi:
caregivers may exceed what they can rea- Birbaumer, N., Gallegos-Ayala, G., Wildgruber, M., 10.1016/B978-0-12-374168-4.00017-4
Silvoni, S., and Soekada, S. (2014). Direct Leuthardt, E., Schalk, G., Wolpaw, J., Ojemann, J.,
sonably achieve with the technology and
brain control and communication in paralysis. and Moran, D. (2004). A brain-computer inter-
result in psychological harm. Selecting Brain Topogr. 27, 4–11. doi: 10.1007/s10548-013- face using electrocorticographic signals in humans.
candidates with higher levels of preserved 0282-1 J. Neural Eng. 1, 63–71. doi: 10.1088/1741-2560/1/
cognitive function for BCI research and Birbaumer, N., Murguialday, A., and Cohen, L. 2/001
treatment and educating them on the (2008). Brain computer interface in paralysis. Curr. McCullagh, P., Lightbody, G., Zygierewicz, J.,
Opin. Neurol. 21, 634–638. doi: 10.1097/WCO. and Kernohan, W. (2014). Ethical challenges
potential benefits and limitations of the
0b013e328315ee2d associated with the development and deploy-
technique is advisable to prevent or at Brady Wagner, L. (2003). Clinical ethics in the context ment of brain computer interface technology.
least minimize harm. The fact that a of language and cognitive impairment: rights and Neuroethics 7, 109–122. doi: 10.1007/s12152-013-
particular BCI system is more invasive protections. Semin. Speech Lang. 24, 275–284. doi: 9188-6
than others does not imply that it has 10.1055/s-2004-815581 Monti, M., Vanhaudenhuyse, A., Coleman, M., Boly,
Fernandez-Espejo, D., and Owen, A. (2013). M., Pickard, J., Tshibanda, L., et al. (2010). Willful
an unacceptable degree of risk and may
Detecting awareness after severe brain injury. modulation of brain activity in disorders of con-
instead indicate a more favorable benefit- Nat. Rev. Neurosci. 14, 801–809. doi: 10.1038/ sciousness. N. Engl. J. Med. 362, 579–598. doi:
risk ratio if it does more to enable the nrn3608 10.1056/NEJMoa0905370
execution of intentions in actions and pro- Hochberg, L., Bacher, D., Jarosiewicz, B., Masse, N., Peterson, A., Naci, L., Weijer, C., Cruse, D.,
mote neuroplasticity. Perhaps the most Simeral, J., Vogel, J., et al. (2012). Reach and grasp Fernandez-Espejo, D., Graham, M., et al. (2013).
by people with tetraplegia using a neurally con- Assessing decision-making capacity in the behav-
significant application of BCIs would be iorally nonresponsive patient with residual covert
trolled robotic arm. Nature 485, 372–375. doi:
in enabling minimally conscious or com- 10.1038/nature11076 awareness. AJOB Neurosci. 4, 3–14. doi: 10.1080/
pletely locked-in patients to communicate. Hochberg, L., and Cochrane, T. (2013). “Implanted 21507740.2013.821189
Yet it is questionable that even the most neural interfaces: ethics in treatment and research,” Sellers, E. (2013). New horizons in brain-computer
sophisticated system alone could demon- in Neuroethics in Practice, eds A. Chatterjee and M. interface research. Clin. Neurophysiol. 124, 2–4.
Farah (New York, NY: Oxford University Press), doi: 10.1016/j.clinph.2012.07.012
strate that these subjects have the ability 235–250. doi: 10.1093/acprof:oso/9780195389784. Wolpaw, J., and Wolpaw, E. (2011). Brain-Computer
to clearly and meaningfully communicate 003.0017 Interfaces: Principles and Practice. New York, NY:
their wishes and make informed decisions Hochberg, L., and Cudkowicz, M. (2014). Locked Oxford University Press.
about life-sustaining treatment. Decision- in, but not out? Neurology 82, 1852–1853. doi:
making capacity falls along a continuum 10.1212/WNL.0000000000000460 Conflict of Interest Statement: The author declares
Hochberg, L., Serruya, M., Friehs, G., Mukand, J., that the research was conducted in the absence
correlating with a continuum of cogni- Saleh, M., Caplan A., et al. (2006). Neuronal of any commercial or financial relationships
tive and emotional capacities, and there is ensemble control of prosthetic devices by a that could be construed as a potential conflict of
no algorithm providing a definitive answer human with tetraplegia. Nature 442, 164–171. doi: interest.
to the question of where the threshold at 10.1038/nature04970
which one has a sufficient degree of these Jox, R. (2013). Interface cannot replace interlocution: Received: 16 June 2014; accepted: 10 July 2014;
why the reductionist concept of neuroimaging- published online: 30 July 2014.
capacities lies. All relevant parties need to based capacity determination fails. AJOB Citation: Glannon W (2014) Ethical issues with brain-
be cautious and not infer that a BCI indi- Neurosci. 4, 15–17. doi: 10.1080/21507740.2013. computer interfaces. Front. Syst. Neurosci. 8:136. doi:
cating certain levels of brain activity and 827279 10.3389/fnsys.2014.00136
semantic processing in a subject is evi- Kennedy, P., Andreasen, D., Bartels, J., Ehirm, P., This article was submitted to the journal Frontiers in
Mao, H., Velliste, M., et al. (2011). Making the Systems Neuroscience.
dence of an understanding of the ethical
lifetime connection between brain and machine Copyright © 2014 Glannon. This is an open-access
magnitude of life-and-death decisions and for restoring and enhancing function. Prog. Brain article distributed under the terms of the Creative
the ability to make them. Res. 194, 1–25. doi: 10.1016/B978-0-444-53815-4. Commons Attribution License (CC BY). The use, dis-
00020-0 tribution or reproduction in other forums is permitted,
ACKNOWLEDGMENT Kubler, A. (2009). “Brain-computer interfaces for provided the original author(s) or licensor are credited
communication in paralysed patients and impli- and that the original publication in this journal is cited,
I am grateful to the reviewer for very help- cations for disorders of consciousness,” in The in accordance with accepted academic practice. No use,
ful comments on earlier versions of this Neurology of Consciousness: Cognitive Neuroscience distribution or reproduction is permitted which does not
article. and Neuropathology, eds S. Laureys and G. comply with these terms.

Frontiers in Systems Neuroscience www.frontiersin.org July 2014 | Volume 8 | Article 136 | 3