Consider one of the more recent, potentially transformative technologies to have emerged: large language models (LLMs). These have become extremely widely adopted in the past few years, and their capabilities are impressive. They have also been aggressively promoted as enhancing, or even replacing, human cognition. This premise has been accepted by many in government and industry, as well as by a huge number of individual users – and yet it remains unclear if it is well-founded. It is also unclear how LLMs are affecting the ways we think.

The picture is gradually becoming clearer. A series of studies at the MIT Media Lab explored how LLM use affects students writing an essay. A key finding, widely publicised, was that functional brain connectivity — essentially, which parts of the brain talk to each other, and how much — dropped significantly when ChatGPT was in the picture. This is not that surprising: if the tool provides immediate, fluent answers, there is little need to formulate questions or process the responses. More striking, however, was that participants showed a dramatic decline in ability to remember the contents of their essays within just seconds of “writing” them, and a sizeable minority – 16% – did not feel any ownership over what they had co-created.

This cognitive debt is the cost of using the tool. It is not unique to LLMs: GPS, for example, has made it unnecessary to give our brains a workout with navigational chores. We have largely accepted the cognitive debt in this latter case, trading the inconvenience of being lost for decreased awareness of our surroundings. We have also accepted the replacement of a specific skill – deciphering maps – by on-screen instructions. Should we do the same with LLMs? What are the specific use cases for which we should be happy to shed the cognitive load? How can we identify these uses, and pinpoint what we stand to gain and lose if we embrace them?

Beyond LLMs, another set of devices is getting closer to our brains. Advances in miniaturisation and data processing mean that brain-sensing technologies like electroencephalography (EEG) and fNIRS (functional near-infrared spectroscopy) can now be packaged into devices designed for everyday use – most obviously, “smart glasses”, which are on the roadmap for almost every major technology manufacturer. Eye-tracking and other inputs (such as electrodermal activity), allow various states — cognitive load, stress, performance, intent — to be monitored and plugged into different use cases. It is also now feasible to use a cheap and simple device to reconstruct, from brain activity alone, the images and text someone is looking at. So far, the reconstructions are rough and preliminary, but their quality is improving quickly.

These advances have undoubted utility in assistive contexts. Brain activity can be used to direct helper robots; smart glasses can help those with vision impairment to understand what is around them; image reconstruction could help those without speech to communicate. But there are immediate questions around privacy and data retention, and subtler ones about how cognition and behaviour would change for both the individual user and those they interact with. Advances now on the horizon promise more radical challenges still, bringing scenarios from science fiction into our daily lives: potential erasure and manipulation of memories, for example, or the ability to tell precisely what someone else is thinking. These are no longer idle or fanciful speculations. We need to think now about what we should and should not do with these emerging technologies.

Mind, body and machine

The relationship between minds and tools is complex and bi-directional. It takes cognitive ability to devise and fashion a tool; but the tool, once made and used, creates new opportunities for cognitive flourishing. The study of how this has played out in our history and prehistory is called cognitive archaeology, a field which examines the various processes, socio-material conditions, and transformations that have shaped human intelligence over the course of human history and evolution. There is a strong anthropological dimension to this, as well as a special concern with understanding the changing relationship between cognition and material culture.

Return to the example of a prehistoric stone tool — the handaxe. You can ask what kind of mind is needed to produce that tool, and its antecedents and successors, and thus reconstruct how human cognition might have evolved. A handaxe certainly transforms the capabilities of a hominin body; the question that remains is whether it also transforms the capabilities of our minds. Does designing, making and using a tool change the ways we think? Is it as a form of cognitive enhancement?

One answer to this question is rooted in cognitivism – what is broadly known as the classical computational or representational view of mind. If we take this perspective, then cognition is what the brain does, and what the brain does is internalise and process information about the world like a super-powerful computer. Enhancing cognition then means developing technologies or adopting behaviours that make you more efficient in performing specific computational tasks – which is what we identify as being “smart.” The objects of material culture – tools, artefacts, artworks – are simply the passive by-products of human intelligence.

This is the premise implicitly held by many who are developing the distant digital descendants of those stone tools, and a promise often made explicitly by their marketing. If adopted, long-established forms of slow thinking, bodily intelligence, and creative material engagement give way to a form of cognitive capitalism, continually seeking to maximise computational profit, but losing sight of some of the most fundamental aspects of human cognitive life.

However, this view pays little heed to alternative forms of wisdom and often neglects the social and emotional aspects of human intelligence and well-being. It becomes particularly hard to sustain when considering our experience of technologies – dark-patterned interfaces, compulsive smartphones, surveillant smart-glasses – which have been designed to affect and sometimes exploit how we think, and about whose effects there is considerable anxiety. We sense that something is being lost in the drive for efficiency and revenue, and that we are not necessarily acting in our own best interests by using these technologies.

A different approach to cognitive archaeology is then to see the tool and the practices surrounding it – its design, manufacture, application and perhaps decorative or symbolic or uses – as making up a kind of cognitive ecology, playing an active role in developing the capacities that make the human mind what it is. Humans evolve through constantly creating things – in the broadest sense of technologies and techniques. These, in turn, transform our minds, our experience of selfhood, and the ways we relate to one another and make sense of the world. This metaplasticity, is one of the distinctive features of our species’ ability to evolve: brains, bodies, and the material environment jointly and interactively create the conditions of their adaptation, and this applies to the modern-day forager of digital information just as much as it applied to the Palaeolithic hunter-gatherer and tool-maker. This view suggests that the nature of human intelligence is continually open to deep reorganisation and reconstitution – thus, in an important and generative sense, always incomplete.

Applied to the present, this view illuminates a common concern: that recent growth of neurotechnologies and brain-machine interfaces challenges our understanding of what it means to be human – and some sceptics would even claim that such radical alterations raise a threat to “human nature.” But if human cognition has always been prosthetic and provisional, then there is no universal or permanent human nature to be threatened by alteration. A capitalist, neoliberal view of cognitive efficiency will be very different from one based on the participatory logic and cognitive ecology of, say, a small, animistic society. We must accept that human becoming is and has always been based on the possibility of change – be it through conventional technologies and stone tools, or the more recent bio- and neuro-technological interventions and interfaces.

Nature, incentives and design

Inasmuch as there is a potential threat, it has very little to do with the fact that enhancement is changing our minds, no matter how great the degree of alteration. That would simply be a continuation of a process that has been playing out for two million years. Instead, the danger primarily relates to who decides and controls what alterations are needed, and what consequences that has for our cognitive lives. So, instead of asking whether cognitive enhancement is “unnatural,” the question we need to be asking is, first and foremost: Who is trying to shape our understanding of enhancement, as well as its modes of implementation, and to what ends?

In some cases, the answer is relatively straightforward. Few would object to a blind person using a stick to navigate the world around them, or, to repeat the earlier example, using smart glasses to better perceive it. In other cases, however, a cognitive technology might be positioned as a corrective for a level of incapacity that was not perceived as problematic before the technology existed. Would it be a good idea to use smart glasses to prompt us through our feelings of social awkwardness, for example? We have seen in other spheres how perceived inadequacies or dissatisfactions can be re-contextualised as health conditions requiring and deserving of treatment. The danger is thus that forms of cognitive enhancement may be disguised as therapy for conditions that only appear problematic from the point of view of those invested in promoting the enhancement itself.

The metaplasticity of the human mind should not be license to naturalise or normalise any kind of cognitive enhancement, which comes in many different forms and conditions. Many are little understood and should therefore be carefully studied and regulated before being fully adopted. We might think of cognitive enhancement in the Platonic sense of the technological pharmakon: both cure and poison, a supplement or substitute that can both support and unsettle — a blessing or a curse.

In this light, the topic of cognitive enhancement forces us to specify what we consider precious about being human, what we want our species to become in the future, and what interests we want to shape that pathway. While the fact of technology-driven or -enabled cognitive change is not necessarily in itself problematic, the assumed, declared and implicit motivations for it may be. To consider this, we need to think about how technology mediates intimate dimensions of who we are: that is, how it affects our sense of self and agency, our identity, bodily habits, temporal rhythms, emotions and so on. We can also study how our aspirations and anxieties are programmed into or addressed by the design of technologies, in order to make those mediations both more effective and more ethically palatable.

The need for this kind of perspective, drawn from cultural anthropology and the sociology of science, is not exclusive to digital technologies. Precursors can be found in studies of consumer technologies such as the clock, automobile, or washing machine, or more recent studies of pre-internet devices such as television. Consider the slot machine, every aspect of which is designed to neuromodulate gamblers’ experiences: the payout mathematics, the buttons, lights, sounds and even their ergonomic seats. Heavy users describe using the machine to gain entry to what they call the machine zone: a dissociative state where they are so focused on playing that their daily worries, social demands, sense of time, space, monetary value, and in some cases even bodily awareness drop away. For these gamblers, the machines have become a vital appendage, one they depend on for the escapism they seek. There are obvious comparisons to be made with the attention technologies which seek to consume our time and focus today: smartphones, gaming, social media, and so on.

One case study for this kind of evolution when it comes to neuromodulating technologies are “wearables” – devices that are nominally used for staying on track with health and fitness goals, and increasingly for time and productivity management. When these devices emerged, around 2010–2012, they were heavily associated with the quantified self (QS) movement, which emerged from groups of technologically-savvy individuals coming together to reflect on what they might learn from data-gathering devices and analytic software, often of their own making. This was an organic and somewhat artisanal movement, which relied on creative analyses of data, sometimes provided by devices and sometimes tracked manually. The expected reward for this often laborious effort was insight into the mysteries, dynamics, and challenges of day-to-day life – drug side-effects, relationship dynamics, associations between diet and productivity. QS was a kind of curiosity-driven quest for self-knowledge through numbers.

In due course, however, the QS movement began to attract those who saw the potential to commoditise and commercialize their approach and thus bring it to a mass market. Typically, the outcome was a product that automated much of the data-gathering and analysis, launched at the Consumer Electronics Show (CES) before making its way to tech retailers: body-fat-measuring bathroom scales, heart-rate-detecting earbuds, focus-enhancing headphones, and so on. As time went on, the devices became capable of gathering multiple inputs – like fitness wristbands and, more recently, smart rings. Various functionalities were also incorporated into phones, removing the need for to buy, wear, and use a dedicated device.

This shift in hardware coincided with a transformation in the use case. Many consumers turned out to be overwhelmed and discouraged by the resulting avalanche of often opaque data and wary of adding to an already mounting list of self-regulatory behaviours (diet, screen time, exercise regimes and so on). The wearables manufacturers reacted by shifting from providing “comprehensive data” to “meaningful insights”. Rather than the wearer using the wearable to record and understand their states and behaviours, the wearable began monitoring, “understanding” and prompting the wearer to take algorithmically-prescribed actions. Today you can buy devices that will gently push your back to prompt you to sit up straight if you are slouching, or that will tell you to take a pause if your breathing becomes rapid or ragged. The emphasis shifted from tracking to coaching: rather than getting your data, you get nudges, notifications, and suggestions for behaviour change.

This is the context for the current proliferation of consumer neurotech. Non-invasive, direct-to-consumer devices intended to monitor, modulate, or interface with brain and nervous system activity – such as learning, attention, and affective states – are becoming ever more common. Computational curiosity and deliberation, as seen in QS, has given way to desires and designs for more automated forms of bodily, emotional, and behavioural self-regulation, where self-recording becomes passive rather than active, self-reflection becomes unnecessary rather than essential, and self-regulation becomes a function of algorithms rather than intentional action. The emphasis is now moving again, from devices that make recommendations about your physical health to your mental well-being: for example, by claiming to stimulate the vagus nerve in order to reduce stress. In some cases, the use case is passive: you simply wear the device and go about your life. Leaving aside the question of whether such devices actually work as intended, there remains the question of whether passively modulating stress would in fact be a good idea. Stress, after all, is a reaction to circumstances; acting on it to change those circumstances might be a more constructive approach – and sometimes, evidence suggests, it is critical to our ability to perform at a given standard.

The slot machine was one response to feelings of pressure, being overwhelmed, and a sense of volatility and unpredictability in one’s life. Seeking out this kind of technology, putting it on, and wearing it consistently is another kind of response. While responsibility is typically placed on users to decide how and when to this utilise these devices and regulate the consequent behaviours, the manufacturers of the technology and other stakeholders are implicated as well. How these devices could or should be governed, and the respects in which our lives are enriched or impoverished as a result, are open to debate.

It is important to note that while we can generalise and extrapolate about trends and directionality, this is also a moment of flux, with different attention-seeking models being trialled in the nascent consumer neurotechnology market, even within specific niches. Those seeking improved focus, for example, can find a technology that purportedly trains you to be more disciplined, like a coach; another which surveils you and alerts you to your shortcomings, like a watchful boss; and a third which recommends “doses” of sustained attention, like a pharmaceutical prescription. Each of these interventions entails a quite different conception of attention, is designed to capture attention in a different way, and expects and requires the user to act in a different way. Each has its own specific logic, and those logics can be quite different from one another. Debates over cognitive enhancement in the attention economy should be mindful of these differences.

Use, misuse and abuse

Devices like the ones above provoke many questions. One of the most intriguing is: Where is the line between enhancement and therapy? In recent decades, bioethicists and many others have tried to identify this boundary. Yet, in a range of wealthy societies, we are increasingly seeing a discourse of “wellness” within which interventions that would have been considered enhancements a decade ago are now positioned as essential to one’s therapeutic well-being. Some advances in neurotechnology are already categorized and marketed this way, such as EEG wearables, and others are likely to follow suit.

This creates several challenges. One is that of imbalanced access: not everyone will be able to get technologies they may feel they need. This is particularly pertinent in systems with socialized healthcare, where budgets are finite, but where patient demand often exceeds clinical views on need and efficacy, sometimes resulting in legal challenges. The complexities of national healthcare structures and needs will have an enormous influence on how these technologies are deployed and implemented. If money is going to be invested with a view to improved clinical care or wider societal impact, then it needs to be clear that there is a pathway to that improvement and that this is money well spent.

Another challenge is the likely use of neurotechnology in other areas that blur the lines between therapy, enhancement, reformation and punishment. An example might be the way that hormones, psychosurgery and pharmaceuticals have been used, historically and currently, to manage the behaviour of prisoners, and those with serious mental illness. It is plausible that neurotechnologies, if they prove effective, might be similarly used in ways that do not aim to treat an ailment or improve well-being, but to make such individuals more tractable to control.

A third challenge speaks to the question of efficacy. Neurotechnology is the subject of considerable hype, with manufacturers making aggressive marketing claims. The resulting high expectations do not match the complexity of clinical, social, and economic realities; actual experience with supposedly therapeutic devices may prove disappointing or even demonstrate harm. Public disengagement or suspicion with neurotechnology could be one result – particularly if this sector is allowed to develop without due regard to issues of privacy, consent, and confidentiality. The commercialisation of neurotechnologies also needs to be free of the mentality, sometimes espoused by governments that are keen to foster innovation, that seeks solely to identify “barriers to implementation.” In some places, regulatory approaches are being signalled that would confine scrutiny to questions of product safety, rather than considering need and efficacy. This framing accepts that these technologies have their places and uses without asking more fundamental questions about the desirability of the resulting changes in our cognitive landscape, in the way we regard well-being, and in our very conceptions of humanity. To address this, social scientific, philosophical and ethical expertise must be incorporated into innovation processes, to ensure that research and development involves a view towards social responsibility, not just regulatory compliance.

A final challenge is the social dimension of these technologies. Design research on cognitive enhancement technologies is usually framed in terms of product design of wearable sensors and body-worn interfaces, and consideration of how these are used by an individual. But there is a wider picture: those individuals navigate a society, interacting with others who may or may not be using similar or different technologies. That means that the uses and effects of such technologies can deviate considerably from the blueprints and business plans laid out at their inception – a situation that becomes acute as cognitive enhancement devices become used to explicitly manage those interactions, rather than inform their users. Technology framed as novel and transformative needs to be introduced in a way that enables properly grounded research from scientists, clinicians, and ethicists, rather than over-investing in hype and perhaps misplaced expectations.

The need to address these four challenges – imbalanced access, potential misuse, unproven efficacy and social dimensions – makes it essential to secure meaningful stakeholder engagement across patient groups, clinicians, industry, academia, and policy. The job of the participatory designer is to uncover the strategies needed to ensure that the ways these technologies are deployed are ethical, safe, and inclusive: it is to ask how we engage various publics in the transition to cognitive enhancement; how we look at the social, political, and ecological implications; and how we ensure that human and planetary values drive the innovation processes. Advances in cognitive enhancement might be amazing for some, but they have the potential to reinforce underlying inequalities; this is one of the critical issues that participatory design needs to explore.

There is a strong market pull for cognitive enhancement innovations, but this needs to occur alongside the development of ethical structures, principles, and practices, which can be a creative activity in itself. Social research with affected communities, including empirical bioethics work, will be needed to identify the modalities that are deemed acceptable and desirable. This work needs to be funded independently of industry, to ensure objectivity and avoid bias. Participatory design approaches to co-designing, prototyping, iterating, and testing can help to ensure that society is ready when cognitive enhancement technologies are fully deployed. People who will be affected by these should be included in the development of frameworks to govern their creation, development, manufacturing, and implementation. Citizens’ juries can play a critical role in establishing what people want, need, and expect from neurotechnologies.

Developing cognitive enhancement technologies in this way will help to ensure that concrete actions are taken toward equitable distribution of power. Awareness-building should begin in primary school, with play-based approaches to help children in exploring cognitive enhancement (among other technologies) in creative ways, and to facilitate discussion of ethical and social aspects and implications in age-appropriate ways and within supportive environments. Regulation must be flexible and anticipatory, able to respond to technology advances rather than lagging behind them, as has been widely regarded as an issue with AI regulation. Global practice networks should be required to design and implement best-practice approaches. At the multilateral level, implementation guidance will help countries to adapt these best practices to their specific contexts and cultures.

In this process, the role of convener and bridger is critical – bringing together scientists, technologists, industry, government, and society and “translating” their observations and concerns in order to hash out the complexities of future regulation. We will need to assemble different parts of society to make sure that implementation pathways, governance structures, and design systems for cognitive enhancement technologies are working for everybody – and thus make sure that they are ethical, safe, inclusive and suitable for all.