Practical autonomy
Although brainjacking might potentially threaten the target’s practical autonomy, the sorts of threat posed by brainjacking to this aspect of autonomy are not particularly novel, as we shall explain. Accordingly, our analysis in this section shall be somewhat brief.
To see why the issues raised by brainjacking for practical autonomy are not particularly novel, consider first the point that the intended therapeutic end of any medical intervention is often the enhancement of the practical dimension of an agent’s autonomy. In the current context for instance, patients such as Alex in case one may undergo DBS to alleviate severe motor symptoms associated with PD. Such symptoms often prevent patients from actively pursuing various activities that may previously have been central to their conception of the good life. In so far as stimulation serves to alleviate these impediments to the patient’s acting on the basis of their desires to pursue these goods, stimulation can be understood to enhance their practical autonomy.
As such, non-consensual third-party control of a DBS could feasibly infringe upon an individual’s practical autonomy by ceasing the individual’s stimulation without their consent. This would serve to re-instate impediments to the individual’s practical autonomy. For instance, a targeted attack of a patient undergoing DBS for PD could plausibly impair the patient’s motor function by changing stimulation parameters (Pycroft et al.
2016).
However, the third party initiation of stimulation that serves to alleviate an impairment (or the third-party deployment of a BCI application) could also be detrimental for the agent’s practical autonomy all things considered. Notice that this would be true even if the stimulation nonetheless enhances the individual’s positive freedom, by virtue of enhancing their physical capacities. To see why, suppose that an individual has refused to consent to an instance of stimulation that would alleviate her motor impairment; if so, then initiating stimulation in spite of this (let us presume valid) refusal would still infringe upon the individual’s practical autonomy in an important sense; it would frustrate her desire not to be stimulated in that instance. Compare this to a case in which a patient rejects analgesics because she would prefer to be lucid and suffering, rather than pain-free but delirious. In so far as the individuals in both cases have validly refused these interventions, they have made the decision that not undergoing the intervention is more important to them than any potential positive effects that the intervention might have on her ability to act in accordance with other desires. The principle of respect for autonomy thus requires that the interventions are not carried be out in each case.
In such cases, it is crucial to distinguish the individual’s freedom to not have their stated preference frustrated, and the increase in freedom that non-consensual stimulation might afford. We should not assume that non-consensually increasing an agent’s physical capacities will enhance their practical autonomy all things considered just because it will enhance their capacity to act in certain ways—doing so will fail to respect autonomy if the agent herself prefers the absence of stimulation to the increase in physical capacity that stimulation affords. Indeed, Carl in case three has very good reasons to prefer not to undergo stimulation in certain contexts. Even though it may enhance his physical capacities by removing his motor impairment, it may give rise to impulsive desires that he strongly does not want to act upon.
What about instances in which the individual has neither consented to nor refused stimulation, perhaps because they have not been asked to consent? Here third parties face epistemic barriers to knowing whether initiating stimulation would frustrate the recipient’s preferences. Whether or not third party stimulation should be construed as enhancing the individual’s practical autonomy in this case would depend on the evaluative weight the recipient ascribes to the potential positive effects that stimulation might have on their ability to act in accordance with some desires, in comparison to the evaluative weight they might ascribe to not undergoing stimulation.
With respect to practical autonomy then, brainjacking may pose a significant potential threat to patients such as Alex in case one: although the attack may not directly affect Alex’s decisional autonomy, the unauthorized cessation of stimulation would rob Alex of the positive liberty afforded to him by stimulation. Conversely, the unauthorized initiation of stimulation might frustrate Alex’s preference not to undergo stimulation, even if it enhances his positive liberty more generally by ameliorating his motor impairment.
However, it is important to note that these sorts of threat are not restricted to brainjacking attacks. Rather, it seems that the hacking of any other medical devices could plausibly be construed as undermining practical autonomy in the ways described above. Hacking a pace-maker can have significant effects on one’s practical autonomy, in much the same way that hacking a DBS system can.
6
Decisional autonomy
Brian-jacking thus raises familiar issues with regards to the practical dimension of autonomy. However, it raises more complex issues about decisional autonomy. The reason for this is that it raises the prospect of a third-party holding direct control over the very processes that undergird our status as autonomous decision-makers. Hackers could plausibly exert direct influence over an individual’s deliberations by brainjacking DBS via a number of different mechanisms, as we shall explore in this section.
Of course, the possibility of taking advantage of these mechanisms depends on the victim having electrodes implanted in the appropriate area. However, with that caveat in mind, it is possible to identify at least three possible mechanisms that might be exploited by brainjackers. First, stimulation of the subthalaminc nucleus (STN) in patients suffering from PD assists in the management of impulse control disorders (ICDs) that are relatively common amongst such patients. By disrupting stimulation parameters in this context, hackers could remove the protection that DBS affords against such ICDs, or even induce aberrant impulse control (Pycroft et al.
2016). Second, hackers could stimulate inappropriate electrode contacts in order to induce personally and socially undesirable emotional changes. Alteration of emotional processing is often an unintended side-effect of STN- DBS in the treatment of PD (and stimulation of the Nucleus Accumbens for other indications). However, some uses of DBS in the psychiatric context have deliberately targeted areas associated with emotional processing in order to modulate the dysregulated affective states that characterise certain psychiatric disorders (Lipsman and Lozano
2014). Third, a number of emerging DBS indications in psychiatry target neural circuits associated with reward processing, such as the Nucleus Accumbens; with a sufficient degree of control of the IPG, a hacker could plausibly initiate stimulation to reinforce certain behaviours (such as eating behaviours in the case of an anorexic patient) as a form of operant conditioning to modify the behaviour of the victim (Pycroft et al.
2016).
Existing discussions of the implications of DBS for decisional autonomy have focused on whether recipients of
consensual DBS can be autonomous with respect to actions undertaken as a result of the
unintended side effects of consensual stimulation (Klaming and Haselager
2010; Kraemer
2013b; Maslen et al.
2015). The focus on consensual DBS is not surprising; due in large part to the highly invasive nature of the procedure required to implant the physical components of a DBS system, and its experimental nature for certain indications, there is a broad consensus that valid consent should be obtained prior to DBS treatment (Nuttin et al.
2014). Moreover, as various authors have pointed out, individuals who undergo consensual DBS sometimes report experiences of self-estrangement or alienation following stimulation, in part due to the development of novel psychological characteristics. In turn, it has been argued that such estrangement could plausibly serve to undermine personal autonomy (and/or moral responsibility).
The phenomenon of self-estrangement following consensual DBS, and its implications for autonomy (and/or moral responsibility) has received a great deal of attention in the neuroethics literature that we lack the space to fully address here (Baylis
2013; Clausen
2010; Gilbert
2013,
2017; Gilbert et al.
2017; Klaming and Haselager
2010; Kraemer
2013a; Lipsman and Glannon
2013; Schermer
2011). However, in discussing brainjacking, we are considering an unprecedented potential avenue for the non-consensual stimulation of patients who have previously provided valid consent to the implantation of physical components of a DBS system. This, we believe raises a new set of questions for the autonomy (and/or moral responsibility) of recipients of DBS. Whilst these questions are our primary focus in this section, the existing prior discussions of the implications of unintended side-effects of consensual DBS for decisional autonomy provide a useful background for understanding the implications of brainjacking DBS for autonomy.
Sharp and Wasserman (
2016) have recently addressed the implications of unintended side-effects of consensual DBS for decisional autonomy, although, as we mentioned above, they phrase their discussion in terms of moral responsibility rather than autonomy. According to this account, an agent is only autonomous with respect to an action if it is issued from a psychological characteristic that she would not be alienated from following hypothetical reflection (unconstrained by distorting factors) on the historical processes that gave rise to it (Sharp and Wasserman
2016). Crucially, Sharp and Wasserman also incorporate a notion of tracing into this historical account, according to which our assessment of responsibility for some act at
t +
1 can
trace back to an action at
t for which the agent was responsible. So, although a drunk driver may not exert control over their driving behaviour whilst drunk, we can hold them responsible for their driving in so far as we can trace back their responsibility for their choice to drink prior to driving (Sharp and Wasserman
2016). Similarly, in the context of DBS, an individual can be held responsible for impulsive behaviour under stimulation if he does not feel alienated from the historical processes that gave rise to his initial decision to undergo stimulation, and if he foresees the likely effects of stimulation on his behaviour.
Although the unintended side-effects of consensual DBS may plausibly undermine decisional autonomy, particularly on externalist accounts, we should note that there will likely be some cases in which the intended therapeutic effects of the consensual procedure may appropriately be construed as
facilitating decisional autonomy. This will be so if stimulation can be construed as enhancing the agent’s competence, or their ability to critically reflect on their motivating desires (or their aetiology). As some of us have argued elsewhere, one scenario in which DBS might do this is if stimulation would serve to increase the individual’s ability to exert top–down control over competing compulsive or impulsive desires, or by reducing the motivational force of such desires, so that they do not move the agent to act prior to carrying out critical reflection about what to do (Maslen et al.
2015). On internalist accounts, stimulation might also arguably enhance decisional autonomy by serving to amplify the motivational force of desires that the agent herself endorses, but which lack sufficient motivational force to move her to act in the absence of stimulation. Alternatively, stimulation might remove other impediments to the sort of critical reflection that autonomy requires, such as pain.
Assessments of the effect of consensual DBS on autonomy are more complicated in cases in which stimulation has unintended deleterious effects on the competencies required for decisional autonomy, whilst simultaneously facilitating the patient’s ability to act in certain ways. In a much-discussed case described by Leentjens et al. DBS was able to effectively alleviate a patient’s severe motor incapacitation, but also led to stimulation-related mania that was not responsive to treatment. Whilst the patient was competent, he was asked to choose between continuing stimulation and being committed to a psychiatric ward, or remaining bed-ridden for the rest of his life due to his motor incapacitation. The patient chose stimulation (Leentjens et al.
2004).
7
This case highlights the importance of the distinction between local and global understandings of autonomy, introduced in section II. With the terminology of this distinction in mind, we might understand the patient in Leentjen’s case to have made a locally autonomous decision to sacrifice his competence to make future locally autonomous decisions, by choosing to continue undergoing stimulation that renders him decisionally incompetent. Nonetheless, we may still understand his choice here as facilitating his global autonomy, in so far as stimulation allows him to achieve the end over time that he himself believes he has most reason to achieve, namely, not remaining bed-ridden due to his severe motor incapacitation. This choice can thus be construed as a kind of Ulysses contract (Unterrainer and Oduncu
2015); just as Ulysses tied himself to the mast in order to hear the Sirens’ song without swimming to his death, so might this patient have decided to live in a state of mania in order to be able to live the remainder of his life in a physically active manner. In both cases, the chosen impairment can be understood to enhance autonomy in an important sense, in so far as it is necessary for the agent to effectively pursue the goal that they have decided is most important
to them, even if the impairment takes away the liberty to pursue other competing goals. In some cases of deciding whether or not to use a medical technology, it is not simply a case of choosing whether or not we want to increase the number of options available to us; sometimes, the relevant choice is between different kinds of mutually exclusive option sets.
As the above discussion makes clear, ascertaining the effects of even consensual stimulation on these different dimensions of autonomy raises complex questions. In some ways, third party interference via brainjacking may seem to make the issue simpler; if brainjacking means that a competent individual does not undergo stimulation voluntarily, it is difficult to see how they could be autonomous with respect to the actions that they performed as a result of that stimulation.
8 Furthermore, it seems clear that third parties can seriously infringe upon another’s autonomy, either by ceasing stimulation that maintains an individual’s competence (so that the individual is no longer able to carry out the sort of critical reflection that decisional autonomy requires), or by ceasing stimulation that maintains an individual’s capacity to act, as we suggested above. In fact, brainjacking seems to be a paradigm case of the sorts of manipulative interference with autonomous decision-making that externalists in the autonomy literature have been at pains to highlight.
However, there are some less clear cases; for instance, it is less clear what we should say about cases in which a third party initiates non-consensual stimulation in order to
increase the agent’s competence to make autonomous decisions. For instance, consider case two above; it might be claimed that Betty’s desire to refuse further stimulation evidences a lack of decision-making competence, in so far as her decision seems to be grounded by pathological values (Tan et al.
2007; Geppert
2015). On this approach, it might be claimed that we should respect her treatment decision when she is undergoing stimulation, in so far as stimulation serves to correct the aberrant neurological processes that underlie her pathological values without stimulation. Moreover, according to internalist accounts of decisional autonomy, if Betty, following non-consensual stimulation, decides to act in accordance with a desire to eat that she endorses following the appropriate sort of critical reflection, she is, by virtue of that fact, autonomous with respect to that behaviour.
However, as Sharp and Wasserman (
2016) make clear, externalist accounts might plausibly object that third-party stimulation undermines decisional autonomy
even if it increases the agent’s competence. It would do so if the individual herself would hypothetically take herself to have reasons to object to the causal history of a psychological characteristic that was efficacious in moving them to act. For instance, in case two, Betty might take herself to have reasons to object to the causal history of her post-stimulation desire to consume food, even though that stimulation might be understood as enhancing her decision making competence, at least on some understandings of that latter concept. Furthermore, on such externalist accounts, Betty could potentially qualify as autonomous with respect to her (non-stimulated) decision to refrain from eating if she embraces the ‘pathological’ process by which she developed this plan in a minimally rational sense (Christman
1991).
There are other circumstances in which it might plausibly be claimed that third parties may be in a better position than the agent herself to readily identify instances in which stimulation might be appropriate. Consider, for example the potential use of DBS to combat paraphilias in convicted sex offenders (McMillan forthcoming); even if stimulation could be effective in reducing sexual urges, such a device would fail to achieve the aim of reducing the targeted behaviour if the recipient were unable or unwilling to reliably initiate stimulation before his urges take hold. Criminal justice authorities might welcome the possibility of hacking such an offender’s device in the interests of public safety. Third party control here would be broadly analogous to closed-loop DBS systems in which the device itself interprets signals from sensory electrodes in the brain in order to determine when stimulation is required, as well as delivering stimulation (Wu et al.
2015). The benefit of third party control is that it would allow non-physiological predictors to be taken into account when deciding when stimulation might be necessary.
What are the implications of third party control for autonomy here? It seems that much will depend on the precise effects of stimulation, how the recipient views his sexual urges, and the basis of the recipient’s prior consent to initially having the physical components of the DBS device implanted. If the recipient authorized third party control over his device, then autonomy-based concerns significantly reduce; however, such authorized third party control would not accurately be construed as ‘brainjacking’.
9
Suppose though that the agent consented to having the device implanted on the basis that he alone would be in charge of his stimulation. Here, concerns relating to the agent’s global autonomy become far more salient, even though the agent’s capacity to make locally autonomous decisions might plausibly be enhanced by stimulation. To illustrate, if stimulation served to increase the agent’s top–down control over an uncontrollable impulsive behaviour from which they feel alienated, then the third party initiation of stimulation could serve to increase his capacity to make locally autonomous decisions with respect to his behaviour on either internalist or externalist accounts of decisional autonomy. Nonetheless, if the agent did not feel alienated from their impulsive behaviour, and their lack of control over their urges, such stimulation would globally frustrate the agent’s practical autonomy to act in accordance with these urges. In such circumstances, an all things considered autonomy-based justification of brainjacking would not be applicable.
As such, the implications of brainjacking for decisional autonomy will depend not only on the particular features of the target’s situation, but also on the particular account of decisional autonomy that we endorse. For instance, in the case of Betty, the key to interpreting the case lies in the assessment of whether Betty is autonomous with respect to her decision under stimulation, or in the absence of stimulation. As we have described above, depending on the theory of decisional autonomy invoked, non-consensual stimulation could be construed as enhancing Betty’s decisional autonomy; however, if we maintain that we should respect Betty’s treatment wishes when she is off stimulation, non-consensual stimulation would amount to a serious violation of her negative freedom to act in accordance with this preference.
Carl’s situation in case three arguably raises a further set of relevant considerations here. It might be claimed that the internalist and externalist accounts we have been focusing on overlook the crucial relational aspect of why we might plausibly have autonomy-based concerns about third-party control over stimulation devices. Such concerns, it might be claimed, are best captured by relational understandings of autonomy, rather than non-relational accounts, as we shall now explain.
As we discussed at the end of section II, relational views make the strong claim that only intentional third party interference can undermine autonomy. Bublitz and Merkel put the point as follows:
Persons can be manipulated through various means, from the presentation of false evidence, hypnosis and advertisements, through to pharmaceutical interventions. It is not the means that render them nonautonomous but the fact that someone else
illegitimately infringed upon their rights. (Bublitz and Merkel
2009)
Such accounts arguably lead to some implausible outcomes, as Sharp and Wasserman (
2016) point out. For instance, it seems highly plausible that an individual may come to lack autonomy in the absence of a third party infringing their rights in some way. For example, those suffering from paraphilias may plausibly claim that they are not autonomous with respect to their behaviour because they are moved to act by irresistible impulsive desires that they repudiate at a reflective level. They seem analogous to Frankfurt’s unwilling addict, described in section I.
Rather than adopting a strong relational view, it seems more plausible to adopt a compromise position that rejects the strong relational claim, but that also affords some significance to third party interference. Such a compromise position might claim that although third-party interference is not necessary for undermining autonomy all things considered, it is necessary to undermine a particular kind of freedom incorporated into our overall conception of agency, namely freedom from domination.
10 The thought here is that in so far as a third party is able to exert control over the individual’s stimulation, the third party is able to exert significant power over that individual.
This sense of freedom has its roots in the republican tradition of political liberty.
11 A salient aspect of this conception of freedom for our purposes is that the violation of this freedom does not require actual interference with one’s liberty; it can be violated simply by the fact that one is in a situation in which another could hypothetically exert arbitrary power over another.
12 To illustrate, we may say a slave lacks freedom from domination, even if he is subject to a benign slave-owner who never tells him what to do, but who could exert this power if he so chose. This is relevant to the present discussion because those who stress the importance of this freedom might plausibly claim that the
mere possibility of a third party hacking a DBS device renders individuals with those devices lacking this sort of freedom. On this sort of account, the need to develop sophisticated forms of cyber-security to protect stimulation devices becomes all the more salient.
Indeed, this is the sort of situation in which Carl in case three seems to find himself in. The damage of the hack is not just that it leads Carl to commit an act of sexual harassment; it robs him of his ability to understand himself as an autonomous agent. Whilst a degree of scepticism about one’s ability to act and decide in a perfectly autonomously manner is a good thing, complete scepticism can in fact undermine autonomy. These reflections raise the daunting prospect that if brainjacking became frequent and difficult to detect, individuals with DBS devices could no longer guarantee that their behaviour had been autonomous, since any action of theirs could feasibly have been influenced by (undetectable) hacking. Not only would this undermine claims of responsibility after a wrongful action has been undertaken, it would also undermine the agent’s ability to take responsibility for their own future actions. This would amount to clinical teams being put into the curious, self-defeating position in which they may be motivated try an improve a patient’s autonomy by implanting a device that de facto casts doubts on the recipient’s autonomy.