When scientists get information overload
Illustration: Sarah Grillo/Axios
Science is moving at a dizzying pace: around 2.5 million scientific journal articles are published a year around the world, and still the volume keeps climbing. But rather than propel science at an increasing clip, the flood has created information overload — and threatens to hold back progress.
The big picture: We have written about how artificial intelligence and faster computing are allowing scientists to go after much bigger problems. But part of the problem is also the more mundane task of simply keeping up with their field in an age of too much data.
Scientists spend a lot of time reading each others’ work. In order to do useful research, they need to know what else is going on: the day’s trends, techniques, and outstanding questions.
- But the sheer volume means no one can read all the relevant work.
- Nor can anyone realistically find only the best papers.
This sets up an impossible choice, says Doug Raymond of the Allen Institute for Artificial Intelligence: A scientist can take the time to understand where her field stands, and risk getting scooped, or publish fast with possibly only incremental results.
So several new projects are using AI to cast a wide net, searching the internet for interesting scientific papers that may have otherwise been buried.
- Mikey Fischer, a Stanford PhD student, created Assert, a site that shows 10 papers at a time, scored by how much they are discussed by influential Twitter accounts and whether their authors published their computer code.
- Andrew Mauboussin, a Twitter data scientist, wrote PCA News, a feed that searches for tweets about AI-related papers, scoring them based on likes and retweets, the quality of replies, and the influence of the account.
- The Allen Institute yesterday announced improvements to Semantic Scholar, a popular search engine for academic research. Now, the site shows tweets, videos, presentations, news stories, and computer code next to research.
By turning away from the traditional peer-review system, these new projects reward research that sparks online buzz, and is verifiable by other scientists.
- "I want a kid in their garage to publish a paper and for it to have impact," says Fischer.
- Raymond, who manages Semantic Scholar, said the new approaches are "making it easier for new scientists to break through with high-impact and high-risk research in fields where there is just an overwhelming amount of publications to sift through."
But, but but: The rigors of peer review can’t be replaced by 240 characters, and social media can be as much of an echo chamber as the ivory tower.
- Just because a paper is trending on Twitter “does not mean it is high quality science,” said Mauboussin.
- A focus on Twitter chatter means researchers with a large following will see a disproportionate boost.
- Assert aims for a middle ground by soliciting simple feedback: Readers can rate papers on their quality, leave comments, and ask researchers questions.