Locked-in Syndrome and BCI - Towards an Enactive Approach to the Self
It has been argued that Extended Cognition (EXT), a recently much discussed framework in the philosophy of cognition, would serve as the theoretical basis to account for the impact of Brain Computer Interfaces (BCI) on the self and life of patients with Locked-in Syndrome (LIS). In this paper I will argue that this claim is unsubstantiated, EXT is not the appropriate theoretical background for understanding the role of BCI in LIS. I will critically assess what a theory of the extended self would comprise and provide a list of desiderata for a theory of self that EXT fails to accommodate for. There is, however, an alternative framework in Cognitive Science, Enactivism, which entails the basis for an account of self that is able to accommodate for these desiderata. I will outline some first steps towards an Enactive approach to the self, suggesting that the self could be considered as a form of human autonomy. Understanding the self from an enactive point of view will allow to shed new light on the questions of whether and how BCIs affect or change the selves of patients with LIS.
KeywordsLocked-in syndrome Enactivism Extended cognition Self Autonomy BCI
Imagine being consciously trapped within an immovable body [1, 2]. You notice what is going on around you but your only way to react or to maintain and initiate interactions consists in blinking movements of your left eye. Locked-in syndrome (LIS) describes a neurological condition in which a patient’s body is severely paralyzed, inhibiting voluntary muscle movement [3, 4, 5]. This prevents the patient from communicating verbally and non-verbally, placing a distance between herself and her social environment. Recent developments in so-called Brain Computer Interfaces (BCI) may help to narrow it [6, 7, 8]. BCIs make use of brain activation patterns voluntarily produced by the patient to control external devices such as a wheelchair, or, in case of communication a cursor for selecting letters on a screen. They augment and facilitate communication and enable actions the LIS patient would not be able to perform otherwise (see  for a comprehensive review on BCI technologies). Understanding both the impact of LIS on a person’s life and the role of BCI therein is a challenging problem for researchers, not only in neuromedicine, but also philosophy, psychology and ethics [10, 11, 12]. To contribute to this is the main purpose of the present paper. In particular I suggest that a major issue of concern should be to clarify how LIS affects the self of the patient and how BCI technologies can alter or sustain it.
In a recent paper Fenton and Alpert have already tackled this issue and suggested that BCIs affect the “autonomous capacity” and selves of patients with LIS. They have argued that Extended Cognition (EXT), a recently much discussed framework in the philosophy of cognition, would serve as the theoretical basis to account for this . According to EXT, our minds do not reside within our heads, but transcend the boundaries of brain and body. Under certain circumstances, they comprise particular features of the external environment [14, 15]. Fenton and Alpert suggest that this can be applied to the self: BCIs have “the potential to change the self” of individuals with LIS, they extend their selves (pp.120,123) .
This is an interesting proposal and, at a first glance, offers a compelling account of the role of BCI in LIS. However, it has two major shortcomings. These are partly due to the way the proposal is presented, the authors do not clarify what they mean with “self” and how it is related to autonomy. On the other hand, the authors uncritically assume that EXT is the appropriate framework to account for the effect of BCI on the LIS patient’s self and mind, not paying attention to the implications and limitations of this approach.
A goal of this paper is to identify some of these limitations and consequently show that EXT is not the right theoretical background for understanding the role of BCI in LIS, but that Enactivism would be the better framework within the philosophy of cognitive science to serve this purpose.
First, I will consider a critical response to Fenton and Alpert’s proposal by Sven Walter, who made a similar suggestion and denied that BCIs are a case of extended cognition . I will argue that BCIs may be considered as a vehicle of some extended cognitive process, and that it is not with respect to this that EXT fails as the appropriate framework, but to the self. Secondly, I will thus consider what a theory of the extended self would look like and provide a list of desiderata, I believe, it fails to account for. These constraints are required for a theory of self in order to determine what the role of BCI is in LIS.
Thirdly, Walter argues, that Fenton and Alpert’s proposal is the result of a confusion of the various, but distinct perspectives within the so-called situated approach to cognition. He has sought to clarify these distinctions; subsequently assigning the role of the appropriate situated approach to the enactive approach to cognition . Walter is right in this, but for the wrong reason. The reference to Enactivism as the better framework is not sufficiently substantiated, because the author misconceives the enactive notions of autonomy and normativity and, more importantly, because there is no enactive account of the self yet. Enactivism, however, entails the basis for developing a theory of self, which may be able to accommodate the abovementioned desiderata. In the fourth and final part of the paper I will thus provide a couple of hints for an Enactive approach to the self, basically suggesting that it could be considered as basic human autonomy. This enables us to begin to intelligibly address the questions of whether and how BCIs affect or change the selves of patients with LIS.
Note, that the following considerations are neither an attempt to offer a rich account of the phenomenology of BCI in LIS, nor a full-fledged analysis of the role of BCI for the Locked-in patient. This paper considers the case of LIS and BCI as informing and challenging with respect to the conceptual, empirical and phenomenological plausibility of two major theoretical frameworks in cognitive science and its main goal is to raise new questions and to highlight problems that remain unarticulated so far. The reader is therefore asked to consider the following suggestions as a kind of road map, providing new ideas on the relations between aspects of an admittedly highly complex case. It does not offer a definite answer, but rather an invitation for further development, both in theory and practice.
Are BCIs a Case of Extended Cognition?
Fenton and Alpert take EXT to serve as a “lens through which we learn to re-see particular aspects of human cognitive engagement with the relevant physical or social environment” (p.126) . Defenders of EXT hold that parts of the external environment can, when functionally integrated become part of the constitutive mechanisms of our mind, and that since specific cognitive processes can transcend our brains and bodies the mind is extended [14, 15].1 Fenton and Alpert assume that this naturally implies that selves are extended, too. Sven Walter rejects Fenton and Alpert’s attempt to apply a theory of the Extended Self (p.67) . He formulates the problem in a more narrow way, focusing on the question whether or not BCI use is a case of extended cognition. In order for BCI to count as a vehicle of cognition it “would have to realize the functional role characteristic of a cognitive faculty or process” (ibid.). In the case of BCI, Walter suggests the following three candidates for cognitive extension: a) the voluntary control over an external tool, b) allowing the patient to act and c) communication. According to Walter, a) and b) are bad examples, because the movement of an arm or cursor is not a cognitive, but a bodily process. At a first glance, c) seems to be thus more suitable: BCIs allow patients with LIS to communicate and express their thoughts. The problem with this, he says, is that the cognitive part in communication is still done in the head because the EEG does not “contribute to formation of thought or intention to communicate” (p.68). The extension Fenton and Alpert refer to is thus only an extension of bodily, but not cognitive processes; BCIs do not substitute or extend communication or any process that we would call cognitive. This is why, according to Walter, the EXT framework is not a suitable framework to account for how BCI affects patients with LIS.
I believe this is wrong for a couple of reasons. Firstly, the proponent of EXT does accept that certain bodily processes count as cognitive. Examples in support include cases of mental rotation  and gestures (pp.123–127) . Now imagine that the LIS patient uses a BCI to control an artificial arm, for performing “artificial gestures”. If gestures are cognitive, then following a principle of parity , BCIs used for moving an artificial limb can be seen as a part of constitutive mechanism realizing the cognitive process of gesturing and could count as a case of extended cognition after all. An analogical example would be the game Jigsaw. Pieces of a puzzle are moved, probed and turned around as to compose a bigger image. In all of these cases bodily movements realize a process that is cognitive.2
Secondly, consider the BCI control of an external cursor. The patient might use a cursor to draw arrows between squares and bubbles of a virtual mind map on a screen, thereby ordering them in a specific logical sequence or connecting them in a way so as to display an abstract property of the relations that hold between different parts of a more complex whole. The proponent of EXT could hold that this is a case of extended cognition, where in order to perform a complex mental task movement of an external tool allows outsourcing parts of the memory process as well as of our abstraction capacities. Bringing items into a logical order, or conceptualizing a dynamic relationship clearly counts as cognitive. The only difference in the case of BCI is that it is partly externally realized.
Thirdly, there is a recent, highly innovative elaboration on BCI technology used for letter selection on a screen allowing patients to learn painting with their brain-activity. Instead of letters or numbers and links for communication or maneuvering the Internet, patients can select “painting commands” comprising “colours, colour intensities, figures, shapes, brush size…” . Painting does not solely happen in the head, in the sense that the painter thinks of an image and then just puts it out there. It involves bodily movement and checking back constantly with the result on canvas. If painting counts as a cognitive process, involving bodily movement and the emerging product on canvas, then BCIs that enable one to move a virtual brush on a screen might be considered as a case of extended cognition.
Consider finally Walter’s claim against BCIs as constituting communication. He says that EEGs don’t “contribute to formation of thought or intention to communicate” (p.68) , which is why BCI cannot be considered as a case of extended cognition. This line of argument, it seems, implies a classical Fodorian view of the relation between language and thought . Fodor considers language and thought to be two disparate things. Thinking might have structures similar to those of natural language; the latter, however, is just a means to communicate the first. BCIs only help a LIS patient to express thoughts in a linguistic and visually tacit fashion, but they are not involved in any thinking process. However, such a perspective on the relation of language and thought stands in contrast to the one advocated by proponents of extended cognition who reject the Fodorian distinction between language and thought, and consider language to be in fact a prime example for extended cognition (pp.50–58) . If language and thought are intimately intertwined and words “augment human computation”, as Clark has put it, then BCIs realizing the formation of words based on letter selection, could be in principle seen as a part of cognition. One might object that both Walter’s and my claim are misguided because communication based on BCI differs so greatly from ordinary communication that a discussion of whether or not BCI communication is extended is futile to begin with.3 While ordinary communication may be extended, BCI communication is not. This objection seems to be a spin-off of the classic argument by Adams and Aizawa who claim (vs. Clark/Chalmers) that the differences between processes that are supposedly extended and original cognitive processes are so great that the former could not possibly count as cognitive . Now, Clark has conceded that the symbols in Otto’s notebook vary greatly from the encodings of our biological memory. However, in order to qualify as a physical vehicle of mental content, sameness of kind is not a necessary condition (p.90) . Importantly, by accepting that “ordinary communication” is extended, all of this seems to be warranted. The problem with the objection is that it reintroduces Adams and Aizawa’s worry while at the same time rejecting it. If one allows for Clark’s vehicle-liberalism for the case of “ordinary communication”, why would one resist doing the same for the case of BCI communication?
Walter’s rejection of BCI as a case of EXT was based on analyzing a very specific cognitive process. The above reasons show that this is not substantiated. BCIs can be seen as extended vehicles of some particular cognitive processes. The EXT framework may have to say something about the relation of LIS and BCI after all. The question remains, however, to which extent this is useful for the present context and the case of self, because the original explanatory target was, as Walter himself summarizes “to understand why Brain Computer Interfaces (BCIs) have the potential to change the self of patients suffering from Locked-in Syndrome (LIS)” (p.61) . The reason why Fenton and Alpert think that EXT is the appropriate framework to account for the effect of BCI on LIS is not because it is seen as constituent of a particular cognitive process. They assume that it entails a theory of the extended self.4
Whether EXT could be applied to the self is a question that proponents of EXT have only marginally dealt with. If we were to accept that specific mental content (dispositional beliefs) is extended, and that such content would be seen as constitutive of self, then it “seems so”, the extended mind implies an extended self (p.16) . Unfortunately we do not find much more than this vague suggestion, but maybe this is too weak a reason to reject Fenton and Alperts’ claim. I propose to have a closer look at what a theory of self—extended style—would look like.
Why Extended Cognition Fails as a Framework
In this section I briefly outline what a theory of the extended self would look like and argue that even though EXT may be in principle able to provide some story about the self, it would not be sufficient to explain the role of BCIs in LIS. It does not imply an account of the experience of self and lacks a proper definition of what self is.
Selves Without Subjectivity
As seen in the above quote EXT is concerned with the unconscious vehicles of cognition. Were we to construct a theory of extended self, we would accordingly look out for its material vehicles (in brain, body and environment) and ask whether parts of the environment could play the same functional role for constituting the self as internal processes. We would try to identify the unconscious constituents of the self. With respect to the present context we might then argue that BCIs could play a role for the LIS patient’s self in that they would be an external part of its unconscious realizing mechanism.
However, the perspective of the subject herself experiencing her condition and the effect of BCI is entirely neglected, but might be of equal importance and considered for the discussion of how BCI can affect the self of patients with LIS (p.41) [12, 19]. Fenton and Alpert, for instance, emphasize, that “there is little doubt that many individuals with LIS experience a loss of autonomy, or autonomous capacity” (my emphasis, p.128). In LIS “the relevant self suffers reduction (…) and patients experience a “sense of loss” (my emphasis, p.129) . The authors clearly appeal to a personal and phenomenological level of description. A theory of the extended self would be mute about the subjective perspective of the extended agent herself; it would tell us nothing about the experience of the person being the extended self. This is because EXT as a functionalist account of cognition is explicitly concerned with the cognition- (or self-) constituting mechanisms only; phenomenological considerations are irrelevant .
Secondly, proponents of classical EXT usually reject the possibility of extended consciousness. While the environment might very well be causally relevant, the constitutive basis for consciousness remains within the head (p. 987)  – the vehicles of conscious experience of self would thus be brain-bound. As a consequence, even if we were to accept that BCIs are realizing some extended cognitive process, they would not be involved in realizing conscious processes. BCIs would not play a role in providing the basis for the experience of self.
The use of BCI and its impact on self could not possibly be seen through the extended lens, as Fenton and Alpert suggest, because the extended lens is blind with respect to the subjective, or experiential level of description and it rejects the idea that the basis of consciousness may be extended. As conscious experience and its basis seem to play a role within the present context, this limitation of EXT would be a good reason to reject it as the appropriate framework.
What is Extended?
There is another major shortcoming of EXT, Fenton and Alpert have to face. The authors reject a disembodied view of the self, according to which “(t)he body is presented as a vessel in which the self is housed and through which the self must express itself”. They assume that the body constitutes the self (p.120). Viewed from the EXT perspective the same holds for the role of the external environment - in this case, BCIs. Selves, the authors hold, “develop from, emerge out of, interactions with our physical and social worlds”, which is why BCI have the “potential to change the self” (pp.120, 127) . Fenton and Alpert share with EXT a deep commitment to what I would call the principle of “location inheritance”, according to which the nature of the cognitive is characterized by the location of its realizing mechanisms: Almost all participants in the extended cognition debate, whether they are proponents or opponents of EXT seem to make this confusing presupposition. Adams and Aizawa identify neural mechanisms as the constitutive vehicles of cognition. Cognition is therefore bound to the brain, and defined in terms of neural activity; it is “a species of causal processing involving non-derived content” , (pp.55, 70) .5 Rupert in turn takes the whole system of brain and body to be part of the underlying mechanism of cognition. The locus of cognition is therefore found within the individual, i.e. the organism as a whole (p.10) . Clark and other proponents of the extended view on cognition then cross the boundaries of brain and body and include certain external features. Notebooks and other external tools can be considered as vehicles of cognitive processes, too. Since these vehicles are located outside of the organism, the cognitive processes they realize are extended. The general attitude within the debate seems to be this: take any possible cognitive process, make a claim as to whether or not you think it can be realized by an external feature and then argue that cognition is or is not extended. This polemic is to hint at a fundamental problem of the extended cognition debate. Because, as Di Paolo so aptly points out: “Before asking where it is we must first say what it is” (p.10) . The greatest shortcoming of EXT is the lack of a definition of cognition.
And this shortcoming applies similarly to Fenton and Alpert’s discussion of the self. They claim that the body does not enclose the self, but that it is constitutive of the self, in that it forms part of its underlying mechanism. Self thus partly acquires a bodily location. Were we now to find vehicles of self that transcend the agent’s organism, such as BCI, we would have to conclude that proper parts of the self are extended, too. But what is the self? Fenton and Alpert remain silent about what the self actually is that is supposedly extended.
Note however, that Fenton and Alpert do not entirely remain with EXT. They partly assume a perspective pretty similar to the Enactivist’s. According to Enactivism “cognition is a relational phenomenon and thereby has no location” (p.9). Neural, bodily and environmental processes all contribute to cognition in that they realize a cognitive identity “that seeks to preserve its mode of life” (p.19) . In other words and very broadly put, a process is cognitive if it contributes to the maintenance of the autonomy of a particular system. Before even attempting to indicate the locus of the cognitive system, before arguing for or against the extension of cognitive processes, Di Paolo thus proposes a definition of the cognitive system as an autonomous, self-producing system that engages in sense-making interactions with the environment. Something like this seems to underlie Fenton and Alperts’ attempt to understand the role of BCI for self, when they say that the self is based on “interactions with our physical and social worlds” (p.126) . They focus on what the different vehicles are realizing: the interaction with the environment. It is the interaction of these vehicles, not the vehicles in themselves, that gives rise to self. Fenton and Alpert paradoxically adopt both positions: they identify parts of the self based on the location of its underlying vehicles (EXT) arguing that BCIs can be seen as a part of self, and that it is therefore extended. But at the same time they adopt a completely different level of description, describing self as arising out of the interaction of a specific mixture of brain-body and environmental features (which is rather in line with Enactivism). If cognition is only the result of the interaction with the environment, then it cannot be identified with the components of its constituting mechanism. The location of different vehicles that are part of a mechanism of a cognitive process does not entail any information about the cognitive process itself. Analogously, what the self is and how BCI can impact on the self is not explained by simple reference to some of its vehicles.
Crucially, if we do not know what the self is in the first place, we might pick out the wrong vehicles. A claim about the extension of the vehicles of the self is not telling and might only be made in a second step. In order to account for how BCI impacts the LIS patient’s self, we need to first determine what the self is. Then we can see what its constituents are and to which extent BCIs may contribute to it.
The very same issue could be phrased in terms of the boundary of the cognitive system or, in this case, the self. Following EXT, its boundary would be physically determined. Since its underlying vehicles, consisting of brain, body and the particular parts of the environment can be physically localized the boundary of self would be drawn around them. The environment of the self would then consist of everything that is not its vehicle. But if it is only the interaction between the vehicles giving rise to self, and these interactions are not reducible to their singular components, the boundary of the self cannot be determined by simply considering the location of the components. The boundary exists not between some material vehicles or others. It is relational, determined and constructed by making a distinction between all those processes that contribute to the emergence of self, and those that do not.
A Definition of Self. Only if we have an idea of what the self is, can we determine what its constituting processes are.
Determining the constitutive processes, or what the interactive dynamics underlying self are. This helps to distinguish self-constituting processes from others, and thus determining the boundary of self. Once we have an idea of what self is and which processes constitute it, we can try to determine their underlying vehicles.
Understanding how these self-constituting processes are realized. What are the vehicles and how do they interact as to bring them about.
An understanding of the interrelations of self, its constitutive processes, their underlying vehicles is required in order to tell what role BCIs play, how they are related to the self, i.e. what exactly they constitute and change.
The role of consciousness and the subjective experience of having or being a self. This is to explain the impact of LIS on the patient’s experience of her self and how is it changed through the use of BCI.
Towards an Enactive Account of the Self
In response to Fenton and Alpert, Walter has argued that taking EXT to be the right theoretical framework is based on “a confusion about some related, but significantly different, approaches to cognition that all fall under the heading of “situated cognition.” (p.62). Having spelled them out, he is convinced that “it should come as no surprise that the adequate “lens” through which to view LIS and BCIs” is Enactivism (p.69). Both, the proponents of Enactivism, and Fenton and Alpert, appeal to similar concepts—autonomy and normativity (p.69). Given the enactivist’s emphasis on agency, Walter argues, BCIs can be seen as a means to “restore the capacity for exploration and interaction” .
Things are unfortunately not that easy. While I agree that the Enactive notions of autonomy, normativity and agency are at play in Fenton and Alpert’s proposal, the similarities Walter claims to make out are somewhat misguided; and even if correct, they would not substantiate the claim that Enactivism is the appropriate framework. Firstly, the concept of autonomy, as used by Fenton and Alpert, differs from the one employed by Enactivism. Secondly, Fenton and Alpert’s main target—the effect of BCI on the self—cannot be accounted for with the enactive approach, because at the moment there is no account of the self in Enactivism. In what follows I am going to clarify the notions of autonomy and normativity and, based on this, suggest some first steps towards a theory of the enactive self. As we will then see in the final part of the paper, these constraints will already help to clarify the role of BCI in LIS.
Autonomy and Normativity
According to Fenton and Alpert autonomy is “the capacity of an individual to freely acquire or develop values or interests and then act in accordance with them in contexts that are minimally coercive” (p.128) . Walter claims that this bears a “significant overlap” with the notion of autonomy as it is conceived of within the enactive approach to cognition: both emphasize “that autonomy is crucially dependent upon the system’s behavior in unfavorable, potentially dangerous or damaging conditions” and both refer to an “ethical or normative, dimension of autonomy” (p.70) .
However, the way Fenton and Alpert see autonomy seems to stand in for something that is quite close to our everyday language use of the concept: somebody is autonomous if he acts “freely”, as Fenton and Alpert say, if he is in control. If somebody has an “autonomous capacity“, then this usually means that he is independent of the support of others; to have autonomy is to “self-rule” (p.128) . Although this is linked to the notion of autonomy in Enactivism, I want to stress a different, more fundamental reading of the term. For Enactive cognitive science autonomy is viewed as a core building block for understanding cognition and the cognitive system [26, 27, 28, 29]. The cognitive system is defined as an autonomous system that is “composed of several processes that actively generate and sustain an identity under precarious circumstances” (p.15) . Autonomy is thus not “just” some capacity of the cognitive system, as Walter’s as well as Fenton and Alperts’ interpretation seem to suggest, but it refers to its identity as a cognitive system, whose basic purpose is to sustain a “coherent, self-determining unity” (p.37) , (p.117)  and, based on this, to “regulate the conditions of their exchange with the environment” (p.8) . As such, autonomy is part of the description of the very nature of the cognitive system.
Secondly, for Enactivism, normativity is based on autonomy. In order for an agent to be autonomous it needs to “have its own goals” and “actively regulate their interactions according to some norm” (p.372) . Norms arise as a consequence of a system’s need to maintain, monitor and regulate itself and its interaction with the environment in such a way that it does not violate the organization of its identity . It is the agent herself generating them. The enactive notion of normativity does initially not imply “mak[ing] judgments on behalf of the agent about the “adequacy” of its behavior to some of our own norms, standard or goals” (ibid.). The ethical implications of BCI use in LIS, as considered by Fenton and Alpert are something different than the Enactivist’s attempt to define an autonomous agent as acting according to some “self-defined normativity” (p.373) . This does not rule out that the enactive notions of autonomy and normativity have ethical implication. In fact, there is recent work, beginning to account for the ethical dimension in Enactivism .
From Autonomy to Autonomy
Fenton and Alpert’s style of presentation and the lack of conceptual clarification make it a bit difficult to see how their proposal could actually be enactive in nature. They say that BCIs impact the “autonomy of individuals with LIS in at least two ways”: with respect to, firstly, their autonomous capacity and secondly, their “self-perception” and “self-nature”. (p.128) . This seems to suggest, that autonomy and self are distinct and interrelated such that self—as it is experienced, as well its constitution—is an aspect of autonomy.
I have argued that this notion of autonomous capacity is not directly in accordance with the enactive approach to cognitive science, and that there is no theory of self in Enactivism. Yet, to regard autonomy and self as interrelated and claim that the self emerges out of the social and physical interactions with the environment, might at least provide an interesting basis for developing an Enactive approach to self.
According to Maturana and Varela all living beings can be described as autonomous systems striving to maintain their identity. According to their theory of autopoiesis, living beings are self-organized and self-producing systems. At one level this is based on the formation and maintenance of a physical boundary that arises through metabolic processes inherent in each system. This physical boundary marks a distinction between the individual system as the biological organism and its environment (the system’s organismic identity, if you wish). Maturana and Varela have claimed that these principles similarly apply to more complex organisms, as well as human beings, which is reflected in the so-called “life-and-mind continuity thesis” . In order to determine a system’s autonomy, we are not restricted to the metabolic and biological level of description. A boundary “can appear at different scales responding to a variety of autonomous processes” (p.378) . The single cell’s environment consists of biochemical processes, but animals and human beings are embedded in an environment that is also socially and, at least clearly in the case of human beings, linguistically structured. They develop a higher, more complex sense of autonomy, the constitutive mechanisms of which are not bound to the organism. Human beings and animals are clearly autonomous in the fundamental sense that they create and sustain an organismic identity, but they differ quite dramatically from other “simpler” living beings in that they also distinguish themselves from other individuals in a way that transcends the metabolic level. As Barandiaran et al. remark: “we should expect natural agents to operate at a single level of organization. Most will involve different scales of autonomy (metabolic, immune, neurodynamic, social, etc.)” (p.378) . In order to establish that BCI affects the autonomy, i.e. agency of the LIS patients, one thus needs to first provide an account of agency that accounts for the autonomous organization of a system based on its linguistically and socially structured couplings with the environment, shedding light on how metabolic and non-metabolic autonomous organizations go together. This would then possibly be closer to what Fenton and Alpert have in mind with respect to the self. As we move from single cells to multi-cellular organisms, and human beings, the principle of autonomy, as a self-organizing and self-producing dynamic does not change, for “every class of entities has an identity which is peculiar to them”—what changes is the domain in which the autonomous system is embedded (from the biochemical to the psychological and social), and with it the specific “kind of organization” (p.77) . Humans do not simply engage in social interactions, but their interactions will have an effect on how their autonomy is constituted; the specific characteristic of autonomy in humans might be that it is intersubjectively constructed. Enactivism provides the basis for a theory of self in that, at the highest level of description, its characterization of living systems as autonomous can be seen as falling together with what we think of as self. Instead of regarding self and autonomy as different things, as Fenton and Alpert do, I therefore suggest that self basically is autonomy—it is exactly this non-metabolic and intersubjectively constituted autonomy, which occurs at the level of human beings, i.e. in agents that live in a socially pre-structured world and are genuinely directed towards each other. In contrast to Fenton and Alpert, autonomy in human does not only mean being “self-ruled” and independent of others, because to be autonomous necessarily requires the constant engagement with others. The independence lies in the fact that while the system is in constant exchange with the environment desirably only those interactions are chosen that do not endanger the self-organizing principle, which is established by the system itself. Based on Varela and Weber “two modes of purposiveness”, i.e. the preservation of identity and the resultant sense-making activities (p.146) , human autonomy, too, is conceived of as teleological in nature. Firstly, in that it strives for the fundamental and self-inherent goal of preserving itself and secondly, that it is directed to successfully interacting in an environment of other human beings. All other goals are dependent on this.
Socially Enacted Autonomy, BCI and Locked-in Syndrome
To conceive of the self as autonomy in the enactive sense is not to provide yet another definition, but offers a novel perspective, a radical suggestion for a general, biology-based and minimal definition of self. It should ultimately be able to provide the basis for integrating further aspects or other models of the self. This is unfortunately not the place to provide a full enactive account of the self. However, going back to the list of constraints from the second section I want to provide at least some very basic idea as to how an enactive approach to self as autonomy would account for them. These considerations will be of an exploratory nature, but I believe that they will already be useful to account for the effect of BCI on LIS patients, therefore offering a better alternative than EXT.
The Constitution of Enacted Autonomy
Self as autonomy arises in agents that are embedded and situated in a socially and linguistically structured environment. It involves an intimate coupling of two kinds of autonomies, which Colombetti and Thompson have called “agent-autonomy” and “interactive-autonomy” (pp.518–520) . Agent-autonomy refers to the “ways in which an individual agent maintains a trajectory that enables her to survive in material and social environments” (p.518) . Interaction-autonomy is the result of participatory sense-making, i.e. the “combination of individual mechanisms” (p.38) which give rise to an “autonomous entity” based on coordinated process of social interaction between agents (p.40) . I take the notion of agent-autonomy to refer only to the individual’s organismic identity, which indeed enables its survival in a material world. But in order to survive in a social environment, another sense of autonomy is required, not entailed in the above-mentioned notion of agent-autonomy. This is the self, an autonomy, brought forward in a non-metabolic domain through the dynamics of (two kinds) of social interactions, in the ongoing back and forth between the processes of distinction and participation. Distinction refers to the fact that the interactions are related to and will affect the subject herself. They help to characterize the differences of the individual, distinguishing it from the other individuals, psychologically, not in terms of a material boundary. Participatory processes refer to the fact that human autonomy also requires and enables being directed to the other subjects and engaging collaboratively in goal-directed actions.7
Vehicles of Socially Enacted Autonomy
The vehicles of a particular human autonomy include all components or features that contribute to the realization of the processes of distinction and participation. They may thus be parts of the human organism, i.e. be neurological or bodily, or external to it, including technologies, and other human beings. In order to describe human autonomy we have to carefully distinguish the levels of description. Following Di Paolo’s characterization of cognition, I suggest that human autonomy is a “relational phenomenon” and has no physical location (p.19) . We cannot solely rely on the individual (as the particular organism for instance), but have to consider its interactions with its material and especially with its social environment. In a sense it is observable at the level of its constitutive processes. But since it is the dynamic interrelation of these processes giving rise to it, autonomy cannot be reduced to a singular process of constitution. It has to be accounted for based on a systemic, interwoven network of social interactions. This also implies—with respect to the realizing vehicles—that human autonomy cannot be found within the brain, or the body alone. Features of the brain and the body contribute to the realization of its constitutive dynamics—they cannot themselves be identified with autonomy. It does not matter for human autonomy what exactly the vehicles realizing its constitution are, as long as they enable the two basic constitutive processes.8
The enactive approach has no clear-cut separation of consciousness from cognition [26, 35], it holds that “cognition and experiential consciousness seem to be best seen as two parts of the same process, that of the lived, embodied action of the organism within its world” (p.4) . Putting conscious experience at the heart of cognitive science will be therefore equally central for understanding the self. Enactivist takes first-and third-person methodologies to be fundamentally intertwined and views them as complementary. A full account of self requires a shift from mere abstract and theoretical considerations to the pragmatic level of the lived experience, because with the emergence of autonomy the system creates a viewpoint on the world, i.e. “a subjective perspective is established” (p.119) , “from which the world is never neutral” (p.506)  and interactions with the social and material environment acquire subjective meaning. This would be impossible to account for without the subject’s own sense of autonomy. It is therefore crucial for understanding human autonomy that it always acts with respect to its own subjective experience, and its own norms based on the ultimate goal to preserve itself. This is not accommodated within the framework of Extended Cognition. Enactivism is decisively committed to the role of lived experience and therefore provides a better basis for an experientially informed theory of self. It is an important task to elaborate on the details of the subjective experience of LIS and BCI, and this is one of the major tenets of the present paper, but not in that I attempt to provide a rich phenomenology of LIS/BCI, but in the sense of bringing the role of experience to attention. This role is undermined if not entirely neglected within the EXT framework, and in the papers by Walter and Fenton and Alpert. This paper points out a lack of consideration of an important issue in the framework of EXT and provide reasons why, if we were to account for LIS and agree on the importance of experience, we should (given that ENC as a theoretical framework decidedly acknowledges the role of subjective experience) rather take ENC as a lens to understanding self and experience in LIS/BCI than EXT.
The Role of BCI in Human Autonomy
It is often assumed, by caregivers as well as family members, that the quality of life of the severely paralyzed patient is poor and her life not worth pursuing. But reports of the quality of life with LIS have shown, to the contrary, that the patients’ assessment of their situation was not significantly different from normal subjects. Patients with heavy physiological impairments adapt to their new situation. This seems to imply that the “physical state does not predict psychological adjustment” (p.344) . Well-being depends rather on the “subjective feeling of control over one’s life and the feeling of a purposeful life irrespectively of the actual physical conditions” as well as one’s ability to communicate with others, i.e. engage in social interactions (ibid.). This supports the enactive approach to the constitution of self. Self is not inherent in the individual organism (or in the body), simply living on, even when completely separated from any social interrelation. It requires a dynamic of ongoing social interactions (in acts of communication) giving rise to it. In Locked-in Syndrome, this interwoven network is heavily affected because of the physiological impairment. Given that, from the point of view of the enactive self, the body is seen as one of the vehicles realizing a human being’s autonomy, the paralysis thus makes it difficult to impossible for the patient to engage in the constitutive processes for the emergence of her autonomy; she cannot or only hardly communicate. She is close to other human beings but in the same time as distant as it gets, because her severe disability eliminates or radically reduces possibilities to interact with them. This is a terrifying situation, because the patient is still embedded and directed to a socially structured environment. She strives towards engaging with the others because this is (according to the enactive self) what makes her herself, the autonomous system she is. Herein lies the important role of Brain Computer Interfaces. They do not extend the self, (if “extended” only refers to the fact that its constitutive mechanism is not found within brain or body alone) because self is already (prior to LIS) and necessarily based on and directed to social engagement with others, involving non-individual, i.e. non-cranial and non-organismic processes. The human body plays a realizing part in the mechanism of its constitution, but it is not sufficient for it and once social engagement is made possible, it may to some extent even be negotiable. BCIs are a means of substituting the body to some extent in that, by acting as a vehicle for communication, they sustain a part of the constitutive dynamics of self, i.e. they secure the communicative exchange with her social environment necessary for self-emergence. Contrary to EXT, for Enactivism, the basis of consciousness is not in the head alone . By sustaining the constitutive dynamics of self, BCIs can be seen as also contributing to the emergence of conscious experience of being that self.9
Implications and Further Issues
As we have seen above, humans are not exhaustively described at one level of autonomy. A question of fundamental concern is how the different autonomies are interrelated. For instance, is bodily (organismic/agent) autonomy a necessary prerequisite for socially enacted autonomy? In “normal” circumstances this seems to be the case. However, LIS patients suffer from heavy bodily impairment, and still engage with their social environment. Especially in BCI communication, where patients rely on distinct patterns of brain activation, not bodily movement, it is not clear what exactly the body’s contribution is. Does it show that bodily autonomy is not required for socially enacted autonomy? An answer to this would require a clarification of what is meant by bodily autonomy or embodiment (for a challenging assessment see ) in the first place.
Although I have not explicitly dealt with it, I wish to emphasize that the above considerations entail important ethical issues. These may be of importance for us, the more theoretically interested scientist, but crucially also for the practice of caregivers and medical doctors. Firstly, recognition of other subjects comes naturally and passively when encountering their moving and lived bodies (p.508) . It is however easy to think of situations, where we do perceive the presence of another subject, even when their bodies are absent. Speaking on the telephone we only hear a voice, but we do not doubt that there is somebody on the other side. The same applies to an online chat-conversation; it is only strings of letters that tell us that there is another subject speaking to us. In usual encounters with others, either the body is present and, if the person is not dead, usually taken as an indicator of there being somebody, or it is not present at all, and we infer from other behaviors (mostly linguistic) that we are dealing with another subject. LIS creates a very paradox situation for the social surrounding. We are presented with somebody, and nobody at the same time. A body that is present, but immovable is alienating in the beginning, making it difficult to recognize that the patient is present. Because of this, special efforts have to be made to secure that the patient is not treated as an object of behavioristic, mechanistic observation , but is recognized, as alive, to be addressed and treated as a subject with her own perspective.
According to the enactive view the socially enacted processes underlying autonomy (as self), are not a one-way endeavor, but rely on communicative (re)-actions of the social environment. It is of vital importance that not only the patient herself is seen as a full participant in the social interaction that needs to engage with the others, but that the others actively participate in this process, too. Self requires the other, and the intersubjective relations will have specific dynamics depending on the conditions of the individuals contributing to them. BCIs facilitate the vital social exchange between the patient and her environment. But communication with somebody who uses BCI technology is extremely different from what we are used to. As Wolpaw remarks in summarizing the downside of BCI research, the development of BCI is still far from offering solutions that are “at once practical, reliable and capable of high-speed complex communication” (p.614) . The changes in the individual condition of the LIS patient using the BCI, crucially affect the dynamics of the social interactions (in terms of complexity of expression, variability of actions, and especially temporality), and, in turn, they impact on both, the patient herself and the person communicating with her. Efforts should be made to determine and then accommodate these differences. This will be important for both, the everyday interaction and treatment of LIS patients, as well as the development of novel technologies and ideas on how to amend the patient’s situation. Even though there seem to arise a greater responsibility on the part of the social environment to accommodate the change in interactive dynamics, the patient’s subjective experience of the change—when they become locked-in, and when they begin to use a facilitating technology—is indispensable: in order to improve the vital social exchange in LIS we need to ask the patients themselves, because only their own sense of autonomy can inform us whether or not a facilitating technology is doing the job that we “normal”-bodied others, suppose it does.
I have argued that a theory of the extended self fails to account for how BCI can affect a LIS patient’s self, because it does not offer a definition of self. This is, however, required in order to determine what exactly its constituting components are and which role BCIs play for the self of LIS patients. On the other hand, the theory of extended self would be mainly focused on the unconscious vehicles of self, but it would not account for consciousness and the personal perspective of the subject. I have argued that in order to explain how the self of LIS patient is changed and the impact of BCI, we have to include the subjective experience of the patient being that self.
I showed why Walter’s reasons to treat Enactivism as an alternative are unwarranted and argued that Enactivism is the better candidate because, other than Extended Cognition, it accommodates for the desiderata of consciousness, first-person perspective and the relation between the mechanistic level of explanation and the level of cognition as a dynamically evolved phenomenon.
I have offered some basic ideas as to what a theory of self from the enactive viewpoint would look like. According to this, self is conceived of as a socially enacted autonomy that arises in humans that are embedded and situated in socially and linguistically structured environments. Human autonomy is based on the ongoing coupling of processes of distinction and participation, and thus not inherent in the individual. As a consequence of this, BCIs do not extend the self, but are considered as one component of the complex interactive dynamics constituting autonomy: by acting as the body’s substituting for communicative expression, they enable the vital social exchange with the environment, thus sustaining the LIS patient’s autonomy. I have emphasized that LIS and the use of BCI change the dynamics of interactions dramatically, and suggested that this should be, firstly, accommodated within the various interactive contexts (family, care givers, medical doctors), but secondly, might also be taken into account for evaluating present and developing novel facilitating technologies.
In accordance with Fenton and Alpert, as well as Walter I refer to “Extended Cognition” as the position originally put forward by Clark and Chalmers.
I owe this example to Ezequiel Di Paolo.
Thanks to an anonymous reviewer for raising this worry.
Even if we take the self to be based on cognitive processes, a claim about the nature of the vehicle of some cognitive processes would not be telling without understanding what self is, and what exactly the cognitive processes involved are.
They do admit that extended cognition is in principle possible, but argue that “as a matter of contingent empirical fact” cognition takes place in the brain (pp.55,70) .
As an anonymous reviewer has correctly pointed out, this does not rule out that it is in principle possible to combine both approaches. The crucial point is that while EXT maybe applied to the relation of BCI and LIS with respect to specific cognitive processes, ENC is a better theoretical framework to account for the self in LIS/BCI. However, given the more general problems of EXT, e.g. the lack of definition of cognition, and its widespread ignorance of the role of the subjective experience a combination of both approaches may be unwarranted after all. Furthermore, if explanatory efficiency matters, than it maybe better to not combine a framework that is explanatorily less encompassing with a second one, but opt for a single, explanatorily more fruitful framework, if such a framework is available.
I propose that these social interactions of distinction and participation are by and large established by communication, which is not restricted to linguistic or verbal communication—bodily postures, touch, gestures and facial expressions etc. are included.
The above considerations have at least two implications. First, the constitutive processes of distinction and participation have to be constantly realized. This is to say that human autonomy would necessarily be affected to the degree that its constitutive processes are; it might even cease to exist, when these processes are absent for a critical time. Secondly, if self is seen as coinciding with human autonomy, which is not constituted by the organism alone, but through social interactions, it cannot be innate and it is also questionable whether very young infants would have it.
Note that I by no means assume that these considerations exhaustively account for the role of BCI. This would indeed require a much more thorough investigation, and is a task for future work. The purpose here is to highlight new avenues for addressing this question and rethinking the role of BCI for LIS. I have to thank an anonymous reviewer for raising this worry.
I would like to thank an anonymous reviewer, Ezequiel Di Paolo, Markus I. Eronen, Rudolf Müllan, Frank Schumann and Sven Walter for valuable comments on earlier versions of the manuscript. My research is funded by a scholarship of the Research Training Group “Adaptivity in Hybrid Cognitive Systems”, University of Osnabrück, Germany.
- 3.Laureys, S. et al. 2005. The locked-in syndrome : what is it like to be conscious but paralyzed and voiceless? In Laureys, S. (Ed.). Progress in Brain Research: 150.Google Scholar
- 5.Bauby, J.D. 1997. The diving bell and the butterfly. London: Harper Perennial.Google Scholar
- 10.Kyselo, M. and Di Paolo, E. in progress. Through the enactive eye—locked-in syndrome as a challenge for embodied cognition.Google Scholar
- 15.Clark, A. 2008. Supersizing the mind: embodiment, action, and cognitive extension. Oxford: Oxford University Press.Google Scholar
- 17.Kübler, A., A. Furdea, S. Halder, and A. Hösle. 2008. Brain Painting—BCI meets art. In 4th international brain-computer interface workshop and training course. Graz University of Technology, ed. G.R. Müller-Putz, C. Brunner, R. Leeb, G. Pfurtscheller, and C. Neuper, 361–366. Austria: Verlag der Technischen Universität Graz.Google Scholar
- 18.Fodor, Jerry. 1987. Psychosemantics: The Problem of Meaning in the Philosophy of Mind. MIT Press.Google Scholar
- 19.Bauby, J.-D. 1997. The diving bell and the butterfly. London: Harper Perennial.Google Scholar
- 22.Adams, F., and K. Aizawa. 2010. Defending the bounds of cognition. In The extended mind, ed. R. Menary. Aldershot, Hants: Ashgate.Google Scholar
- 23.Adams, F., and K. Aizawa. 2009. The bounds of cognition. Blackwell: Wiley.Google Scholar
- 24.Rupert, D.R. 2004. Challenges to the hypothesis of extended cognition. Journal of Philosophy 101: 389–428.Google Scholar
- 26.Thompson, E. 2007. Mind in life: biology, phenomenology, and the sciences of mind. Cambridge: Harvard University Press.Google Scholar
- 27.Di Paolo, E.A., M. Rohde, and H. De Jaegher. 2007. Horizons for the enactive mind: values, social interaction and play. In Enaction: towards a new paradigm for cognitive science, ed. J. Stewart, O. Gapenne, and E.A. Di Paolo. Cambridge: MIT Press.Google Scholar
- 33.Maturana, H. R. and Varela, F. J. 1980. Autopoiesis and Cognition: The Realization of the Living. Boston Studies in the Philosophy of Science. 42. Dordrecht: D. Reidel.Google Scholar
- 34.De Jaegher, H. and Di Paolo, E. 2008. Making Sense in Participation: An Enactive Approach to Social Cognition. In Morganti, F., Carassa, A., Riva G. (Eds.) Enacting Intersubjectivity: A Cognitive and Social Perspective on the Study of Interactions Amsterdam, IOS Press.Google Scholar
- 35.Varela, F.J., E. Thompson, and E. Rosch. 1991. The embodied mind: cognitive science and human experience. Cambridge: MIT Press.Google Scholar
- 37.Lulé, D. et al. 2009. Life can be worth living in locked-in syndrome, Progress in Brain Research, 177.Google Scholar
- 38.Lutz, A., and E. Thompson. 2003. Neurophenomenology: integrating subjective experience and brain dynamics in the neuroscience of consciousness. Journal of Consciousness Studies 10(9–10): 31–52.Google Scholar