New Delhi, Jan 15 (PTI) Researchers using artificial intelligence publish three times more papers, get five times the citations, and progress faster in careers. However, papers that did not use AI covered
a wider range of scientific topics and revealed a higher engagement among scientists, according to a new analysis.
Researchers from the US' University of Chicago and Tsinghua University in China said the use of AI could be prompting scientists to "converge on the same solutions to known problems rather than create new ones", thereby narrowing scientific exploration and engagement.
'AI-augmented research' makes use of AI to quicken and enhance human research -- not replace -- by automating tasks involving vast amounts of data and analysis, enabling scientists to focus on higher-level thinking, creativity and final decision-making.
The analysis, published in the journal Nature, looked at 41.3 million research papers and found that AI-adoption shrank the number of topics studied by nearly five per cent and brought down engagement between scientists by 22 per cent.
The researchers explained that scientists using AI tend to migrate toward fields having abundant data, where AI tools demonstrate measurable advances on scientifically legible benchmarks.
Rather than expanding exploration across science, AI concentrates attention on data-rich domains while leaving a growing number of potentially fruitful areas unexplored, they said.
AI-augmented research creates "lonely crowds", or popular topics attracting a concentrated attention but with a reduced interaction among papers citing the same work, the team said.
This leads to more overlapping research and a contraction in knowledge extent, with scientists converging on the same solutions to known problems rather than generating new ones.
"Scientists who engage in AI-augmented research publish 3.02 times more papers, receive 4.84 times more citations and become research project leaders 1.37 years earlier than those who do not," the authors wrote.
"By contrast, AI adoption shrinks the collective volume of scientific topics studied by 4.63 per cent and decreases scientists' engagement with one another by 22 per cent," they said.
The authors said policy interventions are needed that actively promote gathering new data and alternative uses of AI that expand rather than contract science.
They suggested incentivising research in data-poor areas and encouraging the use of AI systems designed for exploration rather than optimisation.
The team noted that the very models that can generate highly probable outputs are also uniquely situated to recognise the character of surprising data, artifacts, and their scientific consequences.
AI systems will need to be reimagined to expand the sensory and experimental capacity of scientists -- not only cognitive -- enabling and incentivising scientists to search, select and gather new types of data from previously inaccessible domains rather than merely optimizing analysis of standing data, the authors said. PTI KRS
KRS MAH
MAH
