I read an article in The American claiming that science is losing its credibility because it has adopted an “authoritarian tone”. The author, Kenneth P. Green, asserts that science used to be in the business of simply stating facts neutrally; but now it has become increasingly assertive about drawing conclusions from the facts, that is, telling us what we ought to be doing.

For those unfamiliar with the magazine, The American is published by the American Enterprise Institute, a right wing think tank. While I have no idea how scholarly they pretend to be, this particular article is so full of errors and strawmen that it demolishes any credibility they may have as a serious think tank.

Here is the gist of it. They performed an “experiment”, in which they searched for certain phrases in the Lexis/Nexis database. The phrases were:

• science says we must
• science says we should
• science tells us we must
• science tells us we should
• science commands
• science requires
• science dictates
• science compels

They narrowed the search by date, and reported cumulative results for each year, from 1980 to 2009, and graphed them. Here is the graph:

The steep green line at the top is the cumulative total for all phrases. The next two lines, which show a significant increase over the years are for “science tells us we should” in purple, and “science requires” in brown. The rest of the lines are scrunched up at the bottom and do not show any sharp increases in the frequency of those phrases.

From this molehill of “experiment”, the author derives far reaching conclusions. He says that the graph shows that the occurrence of these phrases has increased sharply over the years since 1980, which reflects an increasingly “authoritarian” slant to science. He says:

“In other words, around the end of the 1980s, science (at least science reporting) took on a distinctly authoritarian tone. Whether because of funding availability or a desire by some senior academics for greater relevance, or just the spread of activism through the university, scientists stopped speaking objectively and started telling people what to do.”

Now consider how laughably unscientific this experiment is. If you have children in middle school, consider if your average 10-14 years old could have designed a better experiment. Then understand that Mr. Kenneth P. Green, resident scholar at American Enterprise Institute failed.

I’ve described some problems with their “experiment” below.

The Graph is Meaningless Unless Normalized

It’s a fair bet that between the period covered by the graph (1980-2009), the size of the Lexis/Nexis database grew tremendously. Recall that back in 1980, people used 300 baud modems, and hard drives in gigabyte sizes arrived from IBM on a palette, and cost thousands of dollars. The web as such barely existed. Now consider the situation today, when 2 terabyte drives are available by mail order for a couple hundred bucks. Obviously, databases have grown. A lot more is being stored in the Lexis/Nexis database today than used to be the case in 1980. Here’s a brief history of Lexis/Nexis, showing how they have grown by incorporating more and more publications into their database.

Anyone interested can do their own search, but 2 minutes on Google turned up these facts:

• By 1983, the LEXIS database had 12.5 million pages, including the full text of federal and state laws, court decisions, and much of British and French law.
• Today, to serve its user population of about 5 million subscribers, LexisNexis hosts over 100 terabytes of content on its 11 mainframes (supported by over 300 midrange UNIX servers and nearly 1,000 Windows NT servers) at its main datacenter in Miamisburg, Ohio.

This is not even going as far back as 1980, the date the author uses, when the database would have been even smaller. In 1983, it was 12.5 million pages. Even at a generous 100 kilobyte per document (100 kilobytes is a good sized novella), the size of the database was about 1 terabyte in 1983. Realistically, it was probably much smaller than 1 terabyte. Terabyte size databases were exceedingly rare in 1983. Today it is 100 times larger. This is because Lexis/Nexis is indexing a vast number of magazines, journals, legal documents and other texts today, than it was in 1980.

With a much larger number of publications being indexed today, it’s no surprise that any given search phrase produces more hits. This is why graphs such as these are completely and utterly useless. If the author had even a little bit of common sense, he would have taken the trouble to contact Lexis/Nexis, and ask them specifically “how many gigabytes per year do you add to your database today? How many gigabytes per year were you adding to your database in 1980?” If the difference between 1980 and today is 10 fold, then simply divide today’s numbers by 10 to obtain a normalized result for today. If you want to plot a point for every year between 1980 and 2009, then you need to ask the same question for each year – how many gigabytes of data per year were you adding in 1989? In 1990? In 1991?”

If this was too much trouble, I can suggest a simpler test, which is not as accurate but better than nothing. Pick a phrase that has nothing to do with science, such as “I like cookies”. Do a search for it in the same way, year by year. I am willing to bet cookies to peanuts that he will find the exact same result – that the frequency of occurrence of this phrase increases yearly. This is simply a result of the increasing number of resources indexed by the database, and has nothing to do with whether cookies are really more popular today than in 1980.

So pick half a dozen such non-science phrases. “I like cookies”. “Cars are fun”. Whatever. Get some numbers for how those phrases have changed in frequency, then normalize to those numbers. Better than nothing.

Results Show the Opposite of What Author Claims

Mr. Green makes fleeting reference to the increasing size of the database:

“Some of this may simply reflect the general growth of media output and the growth of new media, but if that were the case, we would expect all of the terms to have shown similar growth, which they do not.”

He gets zero points from me for this. No, we wouldn’t expect all phrases to show similar “growth”. First, as I explained above, no “growth” was demonstrated. You cannot demonstrate growth unless you normalize the numbers, which he failed to do. However, if he had normalized the numbers, even then, any growth (or shrinkage) does not need to be even. Language is an evolving thing. Over time, some phrases become popular. Others become archaic or obsolete. This graph stretches 30 years, over a generation long. That’s plenty of time to see statistical effects in the popularity of phrases.

But the funny thing is that even if you grant him his point, it shows exactly the opposite of what he claims. What are the phrases that are becoming more popular, according to his graph? They are:

• science tells us we should
• science requires

Compare that to which phrases are at the bottom, that did not become more popular:

• science commands
• science dictates
• science compels

Which is more authoritarian? “Science commands”? Or “science tells us we should”? The fact is that the most authoritarian phrases (commands, compels, dictates) are the ones that have grown the least in popularity. If anything this is a sign of decreasing authoritarianism in science. If he had bothered to normalize his numbers, these phrases would probably all have negative growth. But somehow he misses all that and just merrily goes on his way.

Some of the phrases are particularly poorly chosen, such as “science requires”. This could easily be part of a statement such as “… credit in science requires that you take three 101 levels courses in physics, chemistry and biology …” which doesn’t have a darn thing to do with the “authoritarianism” of science, just some school listing its requirements. Or it could be “what science requires of time“, meaning what are the scientific constraints on our understanding of time that need to be taken into account. Again, not an “authoritarian” directive telling people they better not smoke or they’ll get lung cancer, or they better watch the greenhouse gases (Mr. Green’s pet peeve) or the Earth will get hot. Or “Mercury Mission Shows Science Requires Patience“.  This is precisely the content that’s getting indexed in Lexis/Nexis, and is showing up on the graph.

One click on Google turns up a hundred thousand hits, and from what I can see precious few have anything to do with Mr. Green’s thesis about authoritarianism. If you pick such a commonly used phrase, of course you’ll see its use spike as more material is indexed. But it says nothing about authoritarianism in science.

The Lexis/Nexis Database Doesn’t Represent What Scientists Say

The Lexis/Nexis database consists of popular magazines, TV reports, business journals, legal documents, and other texts of this nature. What it does not include are scientific journals. In other words, the material in the Lexis/Nexis database represents the words of journalists, not so much scientists. If you want to see what scientists actually say, better databases would be those which index scientific journals.

So really what he’s claiming is that journalists are using these phrases more often than they used to, that journalists are becoming more “authoritarian” about science. Perhaps in some cases the journalists are actually quoting scientists, but certainly not in all. Journalists also editorialize the words of those they interview. They also present their own viewpoints. Without any effort to differentiate between what the scientists said and what the journalists said, how could you draw any conclusions about scientists? You wouldn’t, if you cared about the truth. Obviously, Mr. Green is not so burdened. He has an agenda to push, and he gets busy pushing it.

The Agenda

So what is Mr. Green trying to do? He works for the American Enterprise Institute, a right wing think tank. They produce reports that are cited by right wing politicians, to support right wing agendas. The science he particularly hates is climate science, specifically global warming. He mentions it specifically:

“The climate community is probably the biggest user of the authoritarian voice, with frequent pronouncements that “the science says we must limit atmospheric carbon dioxide concentrations to 350 parts per million,” or some dire outcome will eventuate.”

This is what he’s fighting against. Apparently, he’s not happy with the gl0bal warming reports, and he doesn’t want any legislative actions taken. So what is he really saying?

When he says “science should be neutral”, what it amounts to is that scientists should just state the facts as they see them, and then shut up. In particular, they should never make any suggestions about what ought to be done. They should have no political voice.

Who then has the political voice? Who makes the decisions? If the scientists shut up, then it’ll be the non-scientists making the decisions. In other words, in the most complex technical matters, when it really helps to have an understanding of science to know what you’re talking about, he wants to silence the most technically qualified people. He wants the only people who are allowed to make “ought” statements to be the most clueless – hacks like himself, politicians of all stripes, whatever. So long as they’re not scientists.

This sort of viewpoint, overwhelmingly silly though it may be, comes from a very real resentment that people like Mr. Green have. Science is outside their understanding, specially highly technical matters such as climate science, where you need a technical understanding of an immense body of data to even sound half-intelligent. Being unable to use science himself, Mr. Green wants to deny it to his opposition as well. He wants to have a shouting match between people as clueless as himself, with the scientists all locked out of the discourse, because if they participate, they’re “tainting” science, don’t you know.