AI ACADEMIA
Failed experiments getting attention?
The UK should use AI to improve the quality of its research, according to a new report from the Centre for British Progress.
The report says the government should focus on areas where AI can be most useful, especially in metascience, the study of how research is done and how it can be improved.
One key point is that AI could help fix ongoing problems in the research system.
It could make evidence reviews faster, help researchers share code and data more easily, and reduce the cost of checking whether studies can be repeated with the same results.
Here’s what happened:
AI could help make research reviews faster and improve reproducibility.
The report wants the UK to increase the share of world-leading research in REF 2029.
A proposed Dark Data Prize would reward researchers for publishing useful results that usually go unseen.
The lab clean-up
The report says the UK should aim for 30 per cent of research across three REF 2029 panels to be rated as world-leading. In REF 2021, that figure was nearly 20 per cent.
The report also highlights the importance of “dark data”, things like null results, failed experiments, and other findings that often never get published.
It says this information could help researchers avoid wasting time, improve the quality of wider research reviews, and build better AI tools for science.
To support that, the report suggests creating a UKRI Dark Data Prize to reward researchers who publish useful null results.
AI helping scientists sort through chaos? Love that for them. Also where was this when I was doing my masters!? 😭 - MG