AI is changing the future of research, but not in the way we think

Artificial intelligence continues to affect all walks of life. In the research sector, one probable outcome is the increasing importance of ethnography, as researchers migrate to the methods that give them an advantage over computers. Ethnography , a way of collecting data that involves observing life as it happens instead of trying to manipulate it in a lab, has traditionally been downplayed in academic disciplines such as economics. But its possible adoption might require a wholesale change in the way graduate students are taught. In the pre-AI age, the scientific method for doing research would typically start with observation and reflection by a human researcher, resulting in the formulation of a novel hypothesis. The researcher would then design a method for collecting data, and proceed to gather and analyse it. After forming conclusions, the researcher would then advance to the final stage, which is writing up their findings and communicating them with the rest of the research community, typically in the form of academic papers. AI has disrupted every link in this chain. Whereas the observation and reflection stage used to take weeks, months or years – as a researcher reads and absorbs decades of scientific findings – today, with the assistance of ChatGPT and comparable tools, this step can be compressed to a handful of hours. AI instruments can easily formulate novel hypotheses and propose a suitable research design. Some – but not all – of the data gathering can also be done rapidly by a computer, such as an AI-powered bot scraping data from the internet or seamlessly cataloguing hours of video. Synthesising the data and presenting it in the form of a scholarly paper can also be performed in a few minutes by well-programmed software, before the cycle resumes. As AI tools continue to improve, the rough edges around this process will be further smoothened, making scientific discoveries executable in minutes. As these advancements inevitably arrive, will PhD-holding scholars eventually go the way of the film projectionists and lift operators, being made obsolete by the wheels of technological progress? This may well be the case, but in the meantime, traditional researchers will maintain relevance by adapting to AI and focusing on filling the gaps in its armoury. As mentioned above, one of the areas in which AI struggles most is gathering data, especially if doing so requires communicating with humans, showing empathy and gaining trust. Anyone who has engaged in a discussion with ChatGPT knows that it can be an excellent conversationalist, but only conditional on the fact that we humans are the ones initiating the dialogue and deciding on the topic. Few people would be amenable to the idea of a chatbot asking them questions that it formulated as part of its own data-gathering efforts in some obscure scientific subfield. Most also feel very uncomfortable with the idea of AI-powered cameras and microphones observing their daily routines, whatever the context. In other words, for the time being, AI is a poor ethnographer, especially when compared to a well-trained human. At the same time, ethnography makes only a small contribution to most scientific disciplines, with the exception of anthropology and sociology. Ironically, one of the reasons is the post-Second World War computer revolution, as this made the process of analysing quantitative data quicker and cheaper than at any time in history, spawning generations of researchers with an affinity for applying statistical methods. A further reason is that some disciplines – most notably economics – tend to look down on ethnography as lacking in rigour and relying too much on a researcher’s subjective impressions. Whether the prevailing aversion to ethnography is down to an obsession with numbers or methodological sneering, AI is likely to force a change. Researchers are like all other professionals – they worry about losing their jobs to AI. They are willing to adapt both by learning to use AI as a productivity-enhancer, and by gravitating towards the activities that AI has yet to master, such as ethnography. Given the deep-seated nature of the ignorance of, and antipathy towards, ethnography within a number of disciplines, graduate training programmes will probably require significant reforms to get the new generation of scholars up-to-date on the method. This is especially true given that ethnography itself can be enhanced by AI, as in the case of a computer program assisting an anthropologist in the analysis of interview data gathered in the field. In sum, rather than rendering researchers obsolete, the rise of AI may re-orient the research ecosystem in ways that elevate the value of human presence and intuition. Ethnography – long sidelined by disciplines enamoured with quantification – is poised for a renaissance, not despite AI, but because of it. As machines increasingly dominate the realms of abstraction, synthesis and computation, the irreplaceable human capacity for empathy, contextual sensitivity and interpersonal trust will come to the fore. For disciplines like economics, this moment presents both a challenge and an opportunity: resist the shift and risk obsolescence, or embrace it and evolve.