On Code and Consequence: The Troubling Case of A Technosolutionist Utopia and Opportunities for Responsible AI

Imagine a world where Artificial Intelligence (AI) - powered health coaches help you eliminate your most problematic health behaviours, from smoking to sedentary behaviour, by delivering real-time behavioural nudges tailored to your unique medical, biometric, and lab data. These personalised AI health coaches can deliver real-time behavioural nudges that promise to incrementally improve your health from the comfort of your own home. Sam Altman, CEO of Open AI, and Arianna Huffington, CEO of Thrive Global, forward this vision in an op-ed titled AI-Driven Behaviour Change Could Transform Health Care, penned for Time magazine, arguing that this technology is closer than we think. 

While the promises of AI health coaches remain enticing, this response article critically examines how their vision is fraught with subtle framing techniques that obscure key ethical concerns, perpetuate problematic normative assumptions, and reinforce a technosolutionist rhetoric that overlooks the complexity of health and society. In response, I argue that ‘good AI’ usage in this context is not simply innovative or efficient, but must be responsibly designed and deployed, with consideration for its societal, environmental, and ethical impacts. 

The Problem with Technosolutionism and the Wicked Nature of Health

Altman and Huffington (22025) present an intuitively appealing account of AI. As healthcare trends towards precision medicine, they highlight AI’s ability to engineer personalised and adaptable health interventions in real-time. Their strategic partnership exemplifies a shared belief in AI’s transformative potential to revolutionise healthcare delivery. Yet, at the core lies a central, unspoken ideological assumption: that AI can and should be used to address societal problems. This assumption sidelines political reform, investment in public infrastructure, and structural change in favour of algorithmic solutions. This is a textbook example of technosolutionism, a term coined by Evgeny Morozov (2013) to describe the belief that technology can solve complex problems without first engaging with the underlying economic, political, or historical forces. 

Technosolutionism is not simply a technological solution. Rather, technosolutionism occurs when we grossly underestimate the complexity of a problem and overestimate technology’s ability to adequately address this problem (Morozov, 2013). In Altman and Huffington’s article, they distill problematic smoking behaviour into a consequence of individual failings; as such, AI-powered targeted behavioural nudges are positioned as being able to change smoking behaviour. This account, however, does not ask why somebody may experience problematic smoking behaviours to begin with. The determinants of their smoking are unaddressed, and AI nudges are positioned as a one-size-fits-all solution to stop people from smoking. The technosolutionist approach runs counter to the ability to critically appraise why complex problems are so enduring in the first place, with contemporary critical scholars arguing that this thinking further reinforces the inequities it aims to address (Allen, 2024; Benayoun, 2025; Byrum & Benjamin, 2022). 

In healthcare, a wicked problem refers to issues that are difficult to solve, usually because of multiple interconnected causes and no clear causal pathways (Rittel & Webber, 1973). Wicked problems exist within a larger web of intersecting forces, including social determinants like income, built environment, housing, racism, and systemic disinvestment. The complexity of wicked problems means that AI is not always sufficiently advantaged to compel the change it seeks to promote. For example, AI can nudge somebody to take their insulin, but cannot ensure that an individual has access to it in the first place. 

Altman and Huffington (2025) describe how AI health coaches may aid in reducing sedentary behaviour by framing sedentary behaviour as a straightforward problem with an obvious solution. In actuality, sedentary behaviour is a wicked problem shaped by a constellation of social forces, including ability/disability, socioeconomic status, education, occupation, and access to green space (Ju et al., 2025; Park et al., 2020; Ryan et al., 2022). One person may be sedentary because of poor health literacy. Another may live in a neighbourhood with no public parks and recreational facilities. Yet another may be juggling several jobs and lacks the time for focused physical activity. 

Collectively, these examples highlight how the ‘one size fits all’ solution offered by AI erases the diversity of determinants that shape a person’s experience with a health problem. Addressing sedentary behaviour involves creating environments that enable healthy choices. These environments must account for people’s positioning within broader social systems and institutions. Failing to account for these determinants runs the risk of amplifying inequities. Unlike Altman and Huffington (2025), I do not believe that reminders, nudges, and virtual health coaching are enough to solve complex health problems. 

What Problem is AI Solving, and What are the Costs?

Altman and Huffington (2025) gloss over a critical question: what kind of problems is AI addressing, and what are the costs and risks of its implementation? Without engaging with ethical, social, political and environmental critiques, blanket deployment of AI technologies runs the risk of enforcing inequities. Health inequity is not a function of technical oversight in the creation of AI; rather, it is a systemic reality shaped by structural ‘isms’ including racism, sexism, heterosexism, ableism, and capitalism (Krieger, 2020). These isms exist across the AI pipeline, and without intentional design and accountability that acknowledges these systems, AI risks becoming a tool that codifies bias, displaces responsibility, and obscures harm in the name of optimisation (Green et al., 2024). 

Innovation cannot replace intention, and AI is not designed or deployed in a vacuum. Good AI, therefore, is responsibly developed, thoughtfully implemented, and deeply accountable to the social, ethical, and environmental contexts in which it operates. AI has the potential to augment the way people interface with the health system; however, it must first respond to the structural conditions that drive poor health outcomes. 

This framing of good AI as necessarily responsible runs counter to Altman and Huffington’s (2025) framing, which assumes that emerging technologies can offset decades of health fragmentation, exclusion, and underinvestment that drive health inequities. Neither Altman nor Huffington contend with the root causes of health inequity, and instead sidestep these considerations with vague promises of ethicality. Altman and Huffington further perpetuate a normative assumption that, while acknowledging that problems exist within the health ecosystem, places the primary responsibility of good health squarely on individuals rather than on institutions. In their view, individual failure, rather than systemic failure, is the biggest barrier to good health. In propagating this idea, Altman and Huffington (2025) adopt a political position congruent with a capitalistic agenda that seeks to profit from a ‘solution’ to systemic failures without attempting to address the root causes. 

The Case for AI: Promise Coupled with Caution 

Despite these critiques of the Altman-Huffington (2025) article, I remain cautiously optimistic about the future of AI in health. I agree with the idea that AI has the potential to streamline system inefficiencies, improve access to care, enhance diagnostic and prognostic accuracy, and personalise treatment plans (Khalifa & Albadawy, 2024; Topol, 2019). In these arenas, AI tools can supplement and enhance traditional healthcare (Khalifa & Albadawy, 2024). In contexts of workforce shortages and system strain, AI may alleviate pressure and enhance access (Jiang et al., 2017). 

Despite this, deploying AI at scale for health raises several risks. Bias and exclusion are among the biggest risks of scalable health AI, in which incomplete or biased data systematises inequities. For example, an AI model trained and optimised with data derived from cisgender White men may misrepresent, misdiagnose, or otherwise underserve transgender women of colour (Buolamwini & Gebru, 2018). Gaps in datasets and data repositories used in the training of AI models persist in the deployment context; they don’t simply disappear. We must attend to this to address inequities from the onset, adopting proactive rather than reactive approaches. 

Obermeyer and colleagues (2019) illustrated how biases in AI can cause real harm in healthcare systems. They found that the healthcare needs of Black patients were systematically underestimated by a risk prediction algorithm. This algorithm used cost as a proxy for healthcare needs without first acknowledging that racial and economic disparities shape access to care and treatment. In underestimating the healthcare needs of Black patients, the number of Black patients who were identified for additional care and treatment was reduced by half. Moreover, at the same algorithmically assigned risk levels, Black patients were sicker than their White counterparts. This example highlights how, despite being data-driven and optimised for efficiency, the model replicated structural inequities and biases. This further illustrates what happens when AI is designed to optimise outcomes without consideration for the forces that shape those outcomes. In these scenarios, AI becomes a breeding ground for inequity. 

Data privacy and surveillance are another important concern; if AI tools require continuous tracking and monitoring of biometrics, speech, habits and preferences, what mechanisms are in place to safeguard this data against misuse (Sharon, 2020)? How is that data collected and stored? Are there mechanisms in place to mitigate data exploitation? Environmentalism is another issue of relevance. Training AI models requires a significant investment of resources and carries a substantial carbon footprint (Strubell et al., 2019). This reality is part of ongoing dialogues around the long-term sustainability of AI tools given their impact on planetary and environmental health. 

Outside of these issues, I will continue to emphasise that technosolutionism obscures the root causes of systemic issues that may benefit from collective action, advocacy, and policy reform. AI should not distract from the additional work needed to create more equitable systems. Interventions must be upstream; irresponsible AI may serve as a glamorised band-aid solution while inadvertently diverting resources away from the means of implementing change. 

Good AI  and Human-Centred Design: The Only Way Forward for Health Technology 

If technosolutionism offers ill-suited “solutions,” then good AI represents a deliberate, ethical, and responsible approach to technology that is purpose-driven, tradeoff aware, and accountable. Human-centred design thinking offers a useful framework for fostering responsible innovation central to good AI. In the context of AI, human-centred design thinking is rooted in humanism, empathy, and creativity, and applies these understandings across the AI development pipeline (Saeidnia & Ausloos, 2024). This approach foregrounds the problem identification process and emphasises the importance of understanding the users of AI tools. In understanding user experiences, struggles, and desires, developers can begin to identify the problem and determine if AI is the right tool to address it (Dym et al., 2013). 

Community-engaged pedagogy may enrich the problem identification process by centring community voices. Community-engaged scholarship highlights how problem identification is ideally done in direct collaboration (or at a minimum, consultation) with the diverse interest holders who may be impacted by the issue (Collins et al., 2018). This process helps to ensure that problems are grounded in the structural realities identified by interest holders (Collins et al., 2018). Rather than masking the need for systemic reforms by presenting band-aid solutions, good AI acknowledges its limitations and meaningfully serves its intended end-users. As part of this process, good AI tools should also be compared and contrasted against other emerging technologies and health interventions to determine their unique value proposition. This process enables the creation of intentionally designed tools that resonate with those most impacted by the technology.  

Good AI usage must be cognizant of the environmental, ethical, and societal tradeoffs associated with AI as a whole. AI cannot be readily adopted without first conducting a cost-benefit analysis. Is this technology adding enough value to offset its costs? Are there populations at risk of being unserved by this technology? Are there populations shouldering the disproportionate risks of this technology? AI is not developed nor deployed in a vacuum. A robust analysis of what this technology stands to offer in the wake of other interventions helps build the case for its existence. If an AI technology cannot justify its existence over other forms of technology, perhaps alternative interventions are ultimately favourable. This is perhaps most significantly overlooked by Huffington and Altman (2025); in adopting a technosolutionist lens, they fail to ask if investing in AI health coaches is the best use of finite resources, and if these innovations stand to truly improve health. They fail to ask why wicked problems in health continue to endure over time, and present AI as a ‘catch-all’ to these issues. 

Finally, AI must be accountable to those it intends to serve. Responsible AI is transparent AI that is subjected to ongoing scrutiny and independent audit. Responsible AI shares the features of its model and does not conceal the demographic breakdown of the individuals present within the training data. These auditing mechanisms help build a case for trustworthy AI by assessing a technology’s ability to safeguard sensitive data and information. Trustworthy AI involves creating mechanisms for continuous performance monitoring and creating space for course correction. 

Conclusion 

As AI becomes more ubiquitous, we approach an important juncture: to what extent do we balance the efficiency of AI with its trade-offs? While Altman and Huffington’s (2025) Time article offers a compelling vision into what AI may look like in healthcare, this vision is incomplete and requires further interrogation of AI’s suitability, risk, and broader societal context. Good AI embraces these challenges, fostering accountability to people and the planet. AI is not an autonomous actor; the onus rests on AI developers, institutions, and policy makers to prioritise responsible design, equity, and social justice. In a world that races to build, it becomes even more important to slow down, ask harder questions, and build responsibly only what is truly worth building.