Seeing is believing: Information visualization and the debate over global warming
Table of Contents
Visualizing consensus
Because it is possible to find at least one scientist willing to dispute almost any claim, it is important not just to know what some scientists think, but to know what most scientists think. But how does one measure consensus in the scientific community? The traditional method is to send out a poll and then publish the results. This method has several drawbacks: it can take a long time; the results can be skewed depending on which scientists were polled; and, even if you know what most scientists think, you still don’t necessarily know why they think it. There’s no way to know what data they are relying on to support their claims.
Figure 8: Interactive map based on a search of articles published between 1981-2003 that relate to the topic “mass extinction.” The image shows the connections between articles as well as the relative importance of particular articles. By Chaomei Chen.
Dr. Chaomei Chen, professor at Drexel University and editor of the journal Information Visualization, is trying to map science—or at least scientific literature—in order to understand what scientists are thinking about and why. Using articles that appear in peer-reviewed journals as his source, Dr. Chen has developed software that maps the relationships between scientific articles and the sources they cite. Every article not only presents new material, but also cites previous articles that dealt with the same or a similar subject or that the author feels are influential or relevant. And, because the journals are peer reviewed, at least a few other scientists must think that the article is worth publishing. Chen’s software maps the relationships between of articles that address that topic and displays those relationships as lines of different colors. For instance, if an article about global warming cites another article that studies carbon dioxide emissions, both articles will appear as dots on the screen with a green line connecting them. The more often an article is cited by other articles, the larger the dot that represents it. This makes it easy to see which articles are most often cited. This interactive map allows you to click on the dot and pull up a bibliographic reference so you can look up the article.
Articles on similar subjects are also clustered together, so that you might see, for example, that hundreds of articles have been written on global warming but only a few have been written about the idea that the earth is getting cooler. “You can think of this like a voting system,” says Chen. “Each citation is a vote for that article. When you start to put them together you can see a network begin to emerge and you can see which contribution or piece of knowledge is most often cited and therefore considered most important. This network you can think of this as a snapshot of the scientific communities thinking on that subject at that particular time as recorded in the scientific literature.”
If you map these articles to a timeline, you can start to see patterns in the way that science in general progresses. Not only can you see where the scientific community is focusing its attentions—what Chen calls a “research front”—but you can also see what areas are receiving less attention, and what areas were once popular but have now fizzled out. Depending on the x and y axis the user selects, science resembles a staircase—progressing through a series of breakthroughs—rather than the slow, steady accumulation of knowledge one might expect. Sometimes this shift in interest occurs because new data becomes available, sometimes it is due to external events. The destructive fury of hurricane Katrina, for example, inspired a flurry of articles about the relationship between rising temperatures and extreme weather. These articles may cite recent studies on hurricanes or they may pull up an article that was written years ago but whose findings have gained new importance in light of the event.
Figure 9: Map based on articles published between 2000-2006 that relate to the term “climate change.” By Chaomei Chen.
Most articles are published and then soon superceded by other research or methodologies. A few, however, become classics and are cited over and over again, which, in the graph above, is what causes the long strands looping back into the past.
These visualizations tell us what scientists are talking about, but they don’t necessarily indicate what they are saying. We can’t yet tell whether most scientists think global warming is real just by looking at the images—we would have to drill down by clicking on the dots and then actually read the significant articles. “We don’t know that everyone necessarily agrees with the paper they are citing, but just that it has value as a reference or discussion point,” says Chen.
Chen’s current research involves the analysis of conflicting opinions. His goal is to find a way to map what a specific piece of research has to say on an issue—for example, which papers support the idea that humans are responsible for climate change and which hold that climate change is part of the natural fluctuation. “You will be able to see the substructure of the whole matter of debate. In addition to knowing how many articles support an opinion, you can see what kind of evidence your opponent consistently draws on to make their argument.” Scientists tend to be cagey sorts and rarely come right out and bluntly state their conclusions. Even after reading an entire article it can be difficult to figure out just where a scientist stands on the issue. For this reason, Chen decided to work with issue that people clearly have strong positive or negative opinions about. He decided to start with the debate over the bestselling book, The Da Vinci Code, using the comments posted to Amazon.com.
There main reasons he made this surprising decision is that the book drew unusually polarized responses: the 3000 people who reviewed the book tended to either love it or hate it with very few in between. Just as importantly, Amazon.com not only allows people to provide a written review of the book, but also to rank the book by assigning it a certain number of stars. Five stars means they loved the book and one star means they hated it. This ranking is important because it gave Chen and his team a means of correlating the opinions expressed in the text. Chen set out to discover which terms people who assigned the book one or five stars most often included in their descriptions.
Figure 10: A view of the cluster “Leonardo da Vinci art” and surrounding clusters in positive reviews. By Chaomei Chen.
This research is significant because eventually it will allow Chen to be able to map not just the topics of scientific articles, but also whether those articles support or refute a specific claim. We will be able to compare the number of scientists that support a particular theory at a single glance and be able to access the research they use to support their opinion with just a click. With luck, this software should effectively end the debate over whether or not the majority of scientists support a theory, making it much more difficult for opponents of scientific theories to keep the public confused and, therefore, inert.
