A Distortion of the Self: Brain-Computer Interfaces

A Vormund for Augmentation

In the realm of brain-computer interfaces (BCIs), the introduction of a digital "Vormund" addresses a pivotal challenge: processing the inherently continuous nature of human behavior, particularly language and movement, which cannot be neatly reduced to a limited set of discrete categories. This translation process, while enabling significant advancements in human-computer interaction, particularly in medical and augmentation applications, also raises questions about the preservation of individual autonomy and the integrity of personal expression.

Derived from German, "Vor" meaning "in front of" or "before" and "Mund" meaning "mouth," a Vormund represents someone who speaks and makes decisions on behalf of another, akin to a legal guardian's role for a minor. This concept, while not inherently negative, suggests a diminished level of personal agency for the individual under the Vormund's care. In the context of BCIs, this introduces a nuanced perspective on the technology's role in mediating human-computer interaction, where the BCI could act as a Vormund by translating and discretizing the continuous nature of human thought and motion into a form comprehensible to machines.

Exploring the functionality of invasive BCI systems reveals the complexity involved in interpreting neural activity. The data derived from brain recordings is influenced by numerous factors, including the unique anatomy of each brain, the specific region being monitored, and the nature of the task being performed. This variability underscores the challenge of developing universal standards for interpreting neural patterns, necessitating sophisticated machine learning models to correlate specific neural activities with intended actions or communications. This involves identifying a correlation where an abstract pattern of neural firing, as recorded by the implant, aligns with a specific function, such as a particular arm movement. This approach has proven to be effective: by identifying a limited set of actions (e.g., “open/close hand”), we can train a model to predict with remarkable accuracy the action the cortex is attempting to execute. Similarly, with a constrained set of words or phrases, we can deduce which is being articulated based on the neural patterns observed in the brain's Broca’s area.

This scenario where a technological system needs to impose some categorisation on our movement and speech marks the initial instance where the concept of a Vormund is pivotal: As human activity is inherently continuous and not discrete, our movements and the words we employ to convey meaning are imbued with nuance and are not confined to a predetermined, narrow selection of actions or words. For motor control, actions span a spectrum that is poorly represented by rigid, predefined categories. Unspecified and new movements are essential for performing daily tasks, such as repairs. Consider speech production: the phrase “do you mean that?” can be articulated with different emphases (e.g. “do you mean that?”), altering its meaning without changing the words themselves. Relying solely on a BCI to decode speech into a fixed set of discrete outputs is inadequate due to the rich, nuanced nature of human language. Also, for patients who have already lost certain functions, starting to train a model on complex movements or nuanced speech may not be feasible. Initially, this might not seem overly concerning, but it merely scratches the surface.

Advancement in the pursuit of full function restoration necessitates predictive and pre-trained models. Our understanding of human cognition is often framed by theories that involve feedback loops, where action and reaction are governed by the anticipation of and response to discrepancies between predictions and actual experiences. The brain economizes resources and enhances efficiency by reacting primarily to stimuli that deviate from these predictions. This is evident when reading text, as we tend to gloss over familiar sequences until confronted with an unexpected elephant. In BCI applications for language restoration, efficiency can be significantly improved by relying on a limited, contextually predicted vocabulary instead of an extensive one. The integration of predictive language models, already trained on vast datasets, substantially augments the system's ability to faithfully reproduce human speech nuances. This leap forward, however, introduces a nuanced issue: such models are typically trained on a dataset not reflective of how humans naturally acquire language, encompassing vast amounts of text unfamiliar to the patient. Consequently, this method of predictive decoding relies on principles vastly different from natural human communication, potentially distorting the patient's means of expression in ways that can be perceptible to them. This distortion in language may lead to alterations in personality, bringing forth unpredictable outcomes and essentially assigning our digital persona a digital Vormund.

Before diving deeper into potential solutions, it's crucial to understand the core issue presented by this digital guardianship. The artificiality inherent in AI-generated text—its tendency to feel insubstantial and generic—highlights a fundamental problem: it risks creating a replacement personality under the stewardship of a digital Vormund. This situation risks turning genuine communication into a semblance of interaction, where the exchange seems meaningful on the surface but lacks authentic significance. Yet, this does not lead us to dismiss technological advances outright. Limitations in linguistic expression, when navigated creatively, can sometimes foster more genuine connections, as seen with alternative communication methods like sign language.

A potentially positive development could be the advancement of a more frictionless communication: human communication, as we all know, is rarely perfectly smooth but riddled with misunderstandings, clarifications and genuine incomprehensibility, even when all participants are speaking the same language. The augmentation of speech decoding by means of a digital Vormund can remove such obstacles as communication would be “standardised” to some degree.

The main task in dealing with these issues is to find the sweet spot where we are genuinely restoring communication to otherwise impaired patients. A too frictionless and smooth exchange is most likely not what should be the goal, as we want to avoid a Vormund engaging in a pseudo-conversational monologue. Some of these points have been raised by patients who share personal anecdotes of feeling patronized by their Vormund. It is crucial to understand the significance of such complaints as physicians may immediately strive to remove such sentiments in the patient without actually treating the root cause of the feeling. The prospect of a patient actually getting accustomed to this new kind of patronage is more dangerous than if the problem is still painfully obvious.  

As BCI technology becomes more accessible, it will likely remain the preferred choice for many, despite its limitations. Acknowledging these challenges is the first step towards innovation. Personalizing predictive language models offers one solution, tailoring them to individual users' histories or even utilizing data from personal archives like social media. Moreover, developing hybrid communication methods that blend traditional and technological means could reintroduce the subtleties lost in digital translation, perhaps through enhanced gestures or novel forms of sign language. Ultimately, the individuals using these technologies are best positioned to identify the most effective communication strategies.

Something we finally need to understand about technology is that tool usage is essentially bi-directional, facilitating certain capacities whilst diminishing others, engaging in a constant feedback between the tool and the user. Language and its various augmentations (written language, the printing press, recorded speech, mass media) have a profound feedback effect on our personalities and our understanding of what it means to be human. The structure of our semantic web, meaning the vast complexity of meanings and significations between the words and phrases in our vocabulary directly correlates with our lived experience and the way we perceive the world around us. The prospect of algorithmically generating our language according to statistically modeled syntactic structures and directly implementing that model into our biological tissue represents a profound breach of traditional barriers. As of now, the transformation of human personality that is initiated by such a fundamental invasion into our cognitive system cannot be overestimated. We simply lack the sufficient engagement of researchers on a fundamental philosophical level about the questions outlined here. 

In navigating the complex interplay between brain-computer interfaces (BCIs) and human cognition, we are confronted with the delicate balance of enhancing communication capabilities without compromising the essence of the self. The concept of a digital "Vormund," while serving as a bridge between continuous human thought and digital interpretation, also prompts a critical examination of the technology's impact on personal autonomy and expression. As we push the boundaries of what is technologically possible, the feedback effects of such tools on our personalities and the essence of human experience come into sharper focus. The integration of algorithmically generated language models into our cognitive processes represents not just a technological advancement but a philosophical challenge that demands careful consideration. Ensuring that future developments in BCI technology enrich rather than diminish the human experience requires a nuanced understanding of the relationship between tool and user, alongside a commitment to addressing the profound questions these technologies raise about identity, agency, and the nature of human interaction.