The era of big data has overturned the classical model of how science works, and budding researchers right down to the school level need to be taught how it has upended discovery, a number of senior scientists have argued.
Instead of collecting information through experimentation to test a theory, scientists are now able to harvest data generated in unprecedented quantities through myriad digital sources, delegates to Berlin Science Week, this year held largely online, were told.
“What data science does is overturn the paradigm of the scientific method,” said Fosca Giannatti, a director of research at Italy’s Information Science and Technology Institute in Pisa.
Now, instead of coming almost exclusively from experimentation, scientifically useful data were “sprouting everywhere, like mushrooms”, she said.
“This is a revolution for many sciences – agriculture, medicine, astronomy, humanities, social science,” she argued during a panel discussion of the scientific method in the 21st century.
Scientists have long been debating the extent to which a tsunami of data would change how they work. Ms Giannatti’s research into human mobility has been transformed in the past 20 years by vast quantities of data from smartphones, for example.
The challenge for scientists now is to explain to “the youngest generation…what data is about, and what it can tell us”, said Edith Heard, director general of the European Molecular Biology Laboratory.
“I think outreach is incredibly important,” she said. “The new generation of scientists will be data scientists, naturally data scientists,” added Ms Giannatti.
But both pushed back at the notion, debated over the past decade or so, that a deluge of data and the computing power to process it would make the traditional, hypothesis-driven scientific method obsolete, with machines discovering new insights in vast piles of information, and human scientists taking more of a back seat.
“Life sciences have really been transformed by big data – big data on a scale and of a quality as never before,” said Professor Heard.
“Big data alone really does not help us completely understand the complexity of living organisms, or an ecosystem…where organisms exist, interact and survive,” she argued. “Big data from the biologist’s perspective provides the information but not the knowledge, and for this we still need conceptual approaches.”
Bias in datasets remains a risk, she also warned. “I work on a process that only occurs in females. I was appalled to realise when I started my studies that many of the big cohorts of data that were being generated for disease studies – cancer or immune diseases – were only being generated in males,” Professor Heard said.
This “chronic bias” means that when women are treated by doctors with medicine, “we’re treated as little men – we’re given the doses according to our size and weight relative” to men, leading to side-effects that impacted only women and to the withdrawal of some drugs, she said.
