Operationalizing Agency in Brain Computer Interface (BCI) Research

As researchers involved in examining ethical issues related to neurotechnologies, we see potential utility in tools like the Qualitative Agentive Competency Tool (Q-ACT) designed to help clinicians, researchers, users, and caregivers evaluate impacts of neural devices on users’ agency. Schӧnau et al. (Schӧnau Andreas et al. 2021) introduce the Q-ACT to help assess changes in agency when a person uses a neural device, arguing that agency sits at the intersection of ethical issues rarely explored together in the neuroethics literature. Based on our team’s previous conceptual and empirical work (Muñoz et al. 2020; Zuk and Lázaro-Muñoz 2019; Zuk, McGuire, and Lázaro-Muñoz 2018; Zuk et al. 2020), we offer three suggestions to help guide further development of the Q-ACT.

First, the Q-ACT is the kind of tool whose development we have advocated for in the past (Zuk and Lázaro-Muñoz 2019), particularly given its incorporation of elements highlighted by experiential and relational conceptions of autonomy. Such a tool will help operationalize and measure the impacts that neurotechnologies can have on the lives of patients and their families, including impacts on personality, identity, agency, authenticity, autonomy, and self (PIAAAS) (Gilbert, Viaña, and Ineichen 2018). However, actualization of this kind of tool has been limited by the sometimes dramatically different ways of talking about key constructs of interest and how they overlap. Through the Q-ACT, Schӧnau et al. position agency as a key aspect of the experience of living with a neural device and implicate autonomy as a unifying concept for holistically understanding impacts on PIAAAS-related factors, particularly responsibility, privacy, authenticity, and trust.

We have argued (Zuk and Lázaro-Muñoz 2019) that three conceptions of autonomy appear in the neuroethics literature: the experiential and relational conceptions already mentioned, as well as a more traditional conception of autonomous agency as solely consisting in a set of basic capacities. This traditional, capacity-based conception of autonomy refers to an individual’s capacity to act intentionally, with understanding, and without internal or external controlling influences (Beauchamp and Childress 2001). This conception typically informs clinical understandings of autonomy used to assess an individual’s capacity to consent to medical interventions in ways that are informed by and congruent with their attitudes, desires and preferences. In the context of neural devices, a person may experience compromised autonomy in the traditional sense if they (1) feel no viable treatment options exist apart from acceptance of the neural device, 92) if they suffer from a psychiatric condition or some form of cognitive impairment that potentially affects their ability to choose a treatment option with sufficient understanding or appreciation, or (3) if the effects of the neurotechnology itself alter an individual’s mental state in a way that impacts their ability to effectively consent or withdraw consent (Cabrera, Evans, and Hamilton 2014). To these basic capacities, experiential conceptions of autonomy add a focus on agentive experience (Bayne 2008), that is, the experience of intentional and deliberate causation as well as a sense of initiation and control over ones thoughts and actions (related in important ways to Schӧnau et al.’s dimensions of responsibility, trust, and authenticity). These conceptions of autonomy in turn differ from relational autonomy, which include as part of an individual’s autonomous agency the agency of others as well as the individual’s circumstances of life more generally (Zuk and Lázaro-Muñoz 2019).

These three conceptions of autonomy offer distinct but complementary ways to understand individual experiences with neural devices and likely cross-cut the dimensions outlined by the Q-ACT model. These different conceptions may help not only to describe but also to explain observed disruptions in any of the four dimensions in the Q-ACT, including whether and why disruptions might appear to be “inherently connected and highly influential on each other” (Schӧnau Andreas et al. 2021). For example, if traditional autonomy is somehow compromised by a neural device, one might anticipate effects on multiple dimensions (e.g., trust, authenticity, responsibility). Any discovery of such patterns depends on a tool’s ability to capture ways in which features of agency might converge or diverge on the basis of these fine-grained conceptual distinctions. Envisioning an ideal tool for evaluating subjective impacts of neural devices, we expected that each dimension of measurement would include items specific to these three conceptions of autonomy (capacity-based, experiential, relational) in order to examine various permutations and potential patterns among them. In one permutation, an individual may retain basic agentive capacities (e.g., ability to form and enact intentions) despite impaired experience of agency or relational agency. Conversely, she might experience strong relational agency (e.g., ability to negotiate others’ access to one’s own information) but still experience detriments in traditional agency (e.g., inability to control her own thoughts or emotions).

Delineating between and among these conceptions of autonomy and exploring patterns in their relationship to the four Q-ACT dimensions may help researchers and clinicians to discover, anticipate, and respond to potential disruptions in autonomous agency. If one’s basic capacities were impaired (traditional autonomy), a tool such as the Q-ACT may help to identify ways in which assistive technologies and other agents themselves can support the agency of those with impaired basic capacities (relational autonomy). We thus encourage the authors to construct the Q-ACT in such a way as to account for (and allow further research into) the forms of autonomy that likely cross-cut the four outlined ethical dimensions, perhaps in patterned ways.

Second, we would like to comment on the concept of authenticity in Schӧnau et al.’s application of Q-ACT to a particular neural device: closed-loop DBS for major depressive disorder (MDD). We base our suggestions on our previous work examining the relationship between DBS, MDD, and authenticity’s opposite, alienation (Zuk, McGuire, and Lázaro-Muñoz 2018), experienced as not feeling like oneself. In that piece, we distinguish between two forms of alienation: mere lack of identification and active disavowal. In the first case, someone may experience alienation from some aspect of oneself (or one’s conditions of life) when one ceases to identify with that aspect. This scenario corresponds to the hypothetical case of Cora, described by Schӧnau et al., a DBS patient who fails to feel sadness at the funeral of a friend due to her neural device over-correcting for the presence of negative emotions without taking into account social context. In this case, Cora’s experience of alienation is due to lack of identification with the flat affect that has replaced what would otherwise (i.e., “normally”) be an emotional experience of sadness over losing a friend. This scenario is qualitatively different than a second one in which Cora actively disavows her identification with the object of her emotional experience—her deceased friend—as the result of new (post-device) negative appraisals of their friendship. This sort of active disavowal—while constituting a form of alienation—maybe be a positive attempt toward restorative (as opposed to deteriorative) self-estrangement from certain people, places, conditions, objects or even mental states perceived as detrimental to an individual’s mental or physical well-being (Gilbert 2018). A careful delineation of passive versus active scenarios suggests that not all types of alienation are negative, and that certain types of alienation can indeed be representations of agency and autonomy. A framework that seeks to operationalize the role of agency, autonomy and authenticity in experiences with neural devices may benefit from a granular understanding of the different possible permutations of alienation: passive versus active, positive versus negative, and the different kinds of objects from which an individual can feel alienated.

This granularity is important because, as Gilbert et al. have argued (Gilbert 2018), it is only through sophisticated measures of these complex phenomena that further advancements can be achieved in neuroethical reflection on these technologies. The Q-ACT provides a promising preliminary armature for deeper investigation into how different aspects of agency—including but not limited to (in our view) dimensions of trust, responsibility, authenticity and privacy—structure individuals’ diverse experiences with neural devices. In our own exploration of ethical concerns among DBS researchers (Muñoz et al. 2020; Zuk et al. 2020), clinicians, patients and caregivers, we have uncovered other factors—such as data privacy, security, risk and safety concerns, financial and post-trial access issues, etc.—that impact patients’ current and future experiences with DBS—which may not fit neatly into the Q-ACT framework as it currently stands. While the range of pressing concerns cited by our respondents may have conceptual links with the agency-focused concerns in Q-ACT, it is not clear whether they all map directly or indirectly onto this framework, despite consensus that they shape patient experiences. We encourage Schӧnau et al. (our third suggestion) to consider which of these further concerns relate sufficiently to agency to be measured by Q-ACT, and which will instead require the development of further measures to investigate impacts beyond agency.

We applaud the authors’ contribution to a field in need of evaluation tools capable of keeping pace with the rapid innovations in neurotechnology and its impacts on individuals and society. We look forward to empirical explorations using this or future iterations of the Q-ACT tool.