Some problems are so large that you cannot see the root cause.
The perfect example is climate change. The basic facts are that the climate is warming up because of fossil fuel use. The details are so complex and intricate that we still don't know the whole story. It's almost impossible to keep up with the rapid pace of change.
"Since 1990's first assessment report (AR), of the Intergovernmental Panel on Climate Change, (IPCC), we estimate that there have been more studies related to climate impacts published each year," scientists write in a paper led by Max Callaghan, a quantitative data researcher at the Mercator Research Institute on Global Commons (MCC) in Germany.
"This rapid growth in peer-reviewed scientific publications about climate change is already pushing manual expert assessment to their limits."
This is a problem in and of itself. How can we ever understand the problem of climate changes if it is so large that it is impossible to analyze, measure, or even comprehend it objectively?
Even meta-analysis studies done by humans are restricted to "dozens to hundreds" of studies.
A different type of entity could be used to read the vast amount of climate science published.
Callaghan and colleagues added another study to their list. They used a deep-learning AI tool called BERT, which uses deep-learning language analysis AI tools to classify and identify over 100,000 scientific studies that detail the effects of climate change.
Although automated analyses such as these are not a substitute for human experts' careful assessments, the researchers admit that their method can still do things that human experts simply cannot.
This meant analyzing large amounts of data, identifying many different types of climate impacts, mapping them across every continent and understanding them in the context anthropogenic contributions of historical temperature and precipitation trends.
Researchers warn that machine-learning analyses at this scale, especially at such a large scale, can produce false positives and other uncertainties.
"While traditional assessments can offer relatively precise but incomplete pictures of the evidence, our machine-learning-assisted approach generates an expansive preliminary but quantifiably uncertain map," the researchers write.
However, before that, some troubling statistics were already generated by the AI analysis.
The study found that 80 percent of the global land area, excluding Antarctica, already exhibits trends in temperature/or precipitation. These climate impacts can already be attributed to an estimated 85 percent population.
We didn't need artificial superrain to know that climate change was a huge problem. But what is telling is the geographic focus of studies on climate impacts.
Nearly half (48%) of the planet's land hosts three quarters (74%) of the global population. High levels of evidence of impacts upon human and natural systems were found in areas that had attributable precipitation or temperature trends.
This means that there is a lot overlap between research on human-caused climate change and impacts on the natural world in countries like North America, Europe, and South Asia.
However, these links may not be as strong in other locations. This could be because there isn't enough climate science to study those regions.
Researchers suggest that the lack of evidence in individual studies may be due to the fact that these locations have not been extensively studied. This could also be due to economic factors (low-income countries are significantly less researched) or geographic characteristics (inhospitable areas are sparsely populated).
The team concluded that "Ultimately, we hope our global, living and automated database will help us to jumpstart a host of reviews on climate impacts on specific topics or particular geographical regions."
"Science advances when it stands on the shoulders giants. In times of expanding scientific literature, giants' shoulders are becoming harder to reach. Computer-assisted evidence mapping can give you a leg up.
These findings were published in Nature Climate Change.
