Beyond the Techno-solutionist Reductionism of AI Applications
By Ayush Bhardwaj
Techno-solutionism in artificial intelligence (AI) or AI solutionism is increasingly becoming an issue as AI is naively suggested as a necessary and unquestioned solution for a range of societal problems. While the recent monumental advancement in Large Language Models (LLMs) is responsible for the wide adoption and popularity of LLM-based AI applications, it is barely comparable to the first or the second industrial revolution that led to a tremendous rise in production, including food production. The following blog article focuses on the ineffectiveness of AI to solve the complex socio-economic problems that it is unthinkingly and repeatedly recommended for.
The Problem of Techno-solutionism
While some scholars differentiate between techno-fixes and techno-solutions (see Sætra and Selinger, 2024), for the ambit of this blog piece, I will be using the terms interchangeably. The idea of a “technological fix” dates back to the 1960s when Alvin Weinberg suggested that the introduction of a suitable technology to address a social problem can eliminate the need for a change in individual social behaviours. Since then, techno-fixes have been employed over a range of issues: education, public health, mental health, poverty, riots, illiteracy, and to address online child safety concerns. Technology-mediated solutions, proposes Weinberg, address the symptoms of a problem rather than the problem itself. In so doing, they adopt a “band-aid approach,” which tends to a problem through quick and easy fixes.
The myriad techno-solutions that have been proposed to date fall into the category of either contested and controversial or overtly foolish. The idea of nuclear weapons as minimally sufficient nuclear deterrence was supported by the father of the nuclear bomb, J. Robert Oppenheimer, as a technological fix to war. Weinberg himself believed that birth control technologies could fix social problems related to gender inequality and discrimination. Techno-feminist scholar Judy Wajcman (2004) criticises the birth-control Pill for effectively standardising women’s reproductive functions on a mass scale. The design of the Pill, which required a regimen of medication for twenty days each month, was influenced by moral considerations and ideas about the natural body. Gregory Pincus, the American biologist leading the research, chose to design a pill that replicated the ‘normal’ menstrual cycle. As a result, all users of the Pill now experience a standardised four-week cycle, reducing the natural variation in menstrual cycles among women. The history of technology is no short of examples of how proposed techno-fixes often led to a cascade of new social problems.
When AI tech is sold off as a technical fix
The tech billionaires and tech firms funding the research in AI propose it as a solution for seemingly ‘all’ our problems. AI has already been deployed to automate decision-making processes: in healthcare, it is being used for predictive medicine, patient data and diagnostics, and clinical decision-making; in higher education, for offloading of academics’ and universities’ work; and in modern warfare, AI supports lethal autonomous weapons, information warfare, and AI-informed control systems. AI-based solutions are also being proposed for climate change adaptation and mitigation action, optimising agri-food production as well as philanthropic activities.
Stuart Russell, a Professor of Computer Science at the University of California, defined AI solutionism as the attitude that ML-based algorithms, once hitched to a vast array of training data, can offer a solution to all our problems. Much like all tech that preceded AI tech, it would seem unlikely that the complex issues of climate change, poverty, food insecurity, global terrorism, recurrent wars in the international arena, etc., would lead to an immediate resolution. Even when AI is offered as a solution to many, if not all, of these issues, there is no guarantee that the use of advanced AI tech can get us any closer to their resolution.
Currently, many issues within the AI tech itself require a solution. At least three internal challenges complicate the use of AI-based applications to social, environmental and international problems:
(a) AI biases: AI systems often have inherent biases. They carry their developers’ Eurocentric perspectives, historical injustices, and therefore, replicate systemic inequities. LLM models are trained on flawed datasets that embed the historical bias of our social and political institutions. A 2023 test conducted on generative AI (GenAI) tools revealed that they amplify both racial and gender stereotypes.
(b) AI is based on an extractive revenue model: AI systems are built on an extractive revenue model that facilitates the accumulation of private data of individuals so that it can be utilised for monetary benefits.
(c) GenAI is deeply unreliable: All GenAI tools (e.g., ChatGPT, Copilot, and Gemini) have been found to present fabricated data or misinformation along with factual data as authentic. The phenomenon is referred to as “AI hallucinations.” Where GenAI has been used to classify and predict outcomes, it has produced several inaccuracies. An Australian government trial revealed that AI-produced summaries are only half as good as human ones, and that it almost always misses the subtle nuances and implicit meanings. Likewise, a 2025 medical survey in Spain revealed that AI-based systems missed 31% of actual cases of potentially lethal melanomas (cancers).
Falling into the habit of Techno-fixes in India
The revolution in AI technologies was bound to be harnessed by India to fulfil its developmental goals. India envisions that the AI ecosystem required for its own indigenous AI revolution would be built on top of its Digital Public Infrastructure (DPI). India’s DPI, which majorly constitutes the India Stack, is an array of interoperable platforms that enable mass delivery of public services. India’s Viksit Bharat 2047 vision lays out plans for the convergence of AI and DPI. The convergence of AI and DPI is believed to unlock the potential for faster adoption of predictive governance, predictive service delivery, dynamic resource allocation, and linguistically-specialised AI in India.
However, such convergence also seems to signify the continuation of the digitisation process heralded by the Digital India Mission (2015) and continued in the form of India Stack. The push for faster adoption of AI tools skips the debates regarding AI risks and safety, and rhetorically constructs AI as absolutely necessary for the development of India. AI adoption in India has already been scaled up through affordable skilling programs, the launch of IndiaAI Mission with an outlay of INR10,300 crore, the establishment of a cloud computing platform called AIRAWAT, new semiconductor plants and other requisite public infrastructure.
There is a lingering belief in India’s top bureaucrats that AI will “solve” complex problems in crucial sites of postcolonial development: agriculture, health, and education. Nandan Nilekani, the architect of India’s Aadhar biometric system, said that “To [unlock] India’s potential with AI, the trick is not to look too hard at the technology but to look at the problems people face that existing technology has been unable to solve.” This resembles Weinberg’s techno-solutionist approach of concentrating on the symptoms of the problem.
In a similar vein, Evgeny Morozov, in To Save Everything, Click Here (2013), argued that in order to provide technological solutions for social problems, the adherents of the belief incorrectly frame social problems. Techno-solutionists myopically turn their heads to things that they can achieve: eliminating inefficiencies, reducing ambiguities, and decreasing opacity. When they find themselves face-to-face with unsolvable social problems, they invent fake problems that their technical fixes can fix.
The focus on the new possibilities created by AI tends to shift attention away from the provision of public goods like elementary education and primary healthcare to reducing inefficiencies in the process of delivery of these goods by the privatisation of delivery systems. By spearheading the AI revolution in India, the incumbent government attempts to project an image of benevolent, technocratic developmentalism, while some questions regarding the delivery of core public services–health, education, nutrition–are still unsettled.
The quick technological fixes offered first, by the digitisation, and now, by the automation of service delivery using AI, brush aside, amongst many others, the concerns regarding the emergence of a competitive welfarism in India. Yamini Aiyar (2023) argues that competitive welfarism reduces welfare offerings as “a cynical electoral strategy that risks draining the exchequer.” Aiyar believes that the emergent form of competitive welfarism is characteristic of a techno-patrimonial state: a state which discursively reconstructs citizens as passive recipients (labharthis) of the state rather than active, rights-bearing actors.
The core problem remains: who are the rightful beneficiaries of welfare benefits? What new welfare schemes should the government launch to protect the worst sufferers of India’s economic policy spanning many decades since its independence? AI solutions may, at best, offer optimisation of a process already in place. What they do not do is allow thinking of the problem from all possible angles. Such problems call for a multimodal, multi-institutional, and trans-disciplinary approach for resolving the issue.
Way Forward
At a time when AI safety is being discussed and debated at all important global platforms, India has ranked at the top in a survey on the willingness to trust AI systems, with 75 per cent of all Indian respondents willing to put their trust in AI technologies. This reflects the misplaced techno-optimism of the nation. In general, the BRICS nations were found to be more techno-optimistic than their Western counterparts. Techno-solutions create the illusion that the problem has been taken care of. Such solutions best serve the interests of the governments whose tenures last a few electoral cycles, and technical fixes appear to be more viable compared to long-term, tedious solutions often involving a behavioural change on the part of the masses.
Resisting AI solutionism would entail questioning the relevance of a new AI technology or a particular AI application to the socio-economic and (naturally, to a large extent) political problem. Contextualising AI use and AI behaviour to the specific demographic, community, nationality, and the problem the technology is supposed to grapple with would throw light on settled patterns of behaviours that AI solutions build on top of. Besides, the contextualisation would also assist in mapping out the probable changes the AI-based solutions can introduce in the community-specific.
Technological solutions are claimed to be politically neutral. But once their adherents reframe social problems in for-profit techno-solutionist parlance, the privatization and the marketisation of public concerns appear as the next logical step. Resisting AI solutionism thus also requires disintegrating settled patterns and age-old processes that legitimise the transfer of social concerns from civil society to the state, and post-neo-liberalisation, from the state to the private sector.