A revolution in research as AI searches masses of open data maybe the next big thing, but researchers are not ready for an imminent arrival.
The ’23 edition of Digital Science and partners survey of researchers attitudes to open data finds they still do not believe they are appropriately credited for sharing, pointing to low citation rates.
Having the time and knowledge to publish are also continuing issues. And grant conditions that mandate open data are not helping, they argue. “We may also be seeing some fatigue in enthusiasm for open research in general as open data policies come into place and researchers find themselves with even less time to comply with the directives of their funders,” the authors suggest.
Certainly, altruism is up there, with “public benefit” as the second most common reason why data should be public. But not everybody is sold on open data as a research principal, 10 per cent of responders are uncomfortable with theirs’ being re-used in any format.
This reflects a broader issue – ambivalence about what AI can do with small data sets. “Only a small proportion” of researchers have considered using it to collect, analyse and annotate their data in previously machine un-readable formats – figures, tables, supplementary files.
This is despite the potential of large language models that can write code. When they work, they can be a researcher’s personal data scientist, exploring data sets, proposing research questions, designing statistical analysis and presenting results graphically, all within minutes.
“AI raises the tantalising prospect of data finally taking its place at the heart of our scientific articles, with integrated transparency and reproducibility as to how the data were processed and analysed,” the report states.
Which may be what makes some researchers nervous.
