Biased Production

One of the easiest ways that an interest group can mislead the public and policymakers is to fund or produce research that is congenial to their cause and fail to publish unfavorable results. This is called biased production. The most famous historical example comes from the tobacco industry which from 1955 until 1986

spent more than 130 million on sponsored research, resulting in twenty-six hundred published articles. […] This research was then shared, along with selected independent research, in industry newsletters and pamphlets; in press releases sent to journalists, politicians, and medical professionals; and even in testimony in Congress (The Misinformation Age, p.104).

Biased production is most common when an industry or interest group has a vested interest in a particular conclusion being perceived as true. More on this in a moment. First, let’s look at the ways biased production can occur.

Biased Production with In-House Research

The most straightforward way occurs with in-house research; i.e., when a particular firm or an industry group conducts research in their own facilities by their own employees. Biases in this model can occur a few ways–not all of them nefarious. For example, sometimes the mere desire for a research team to come up with a marketable product can induce bias in how data is collected and interpreted. For more on these sorts of biases follow the link to my critical thinking course. These unintended biases aren’t biased production.

Genuine biased production occurs most commonly through selective publication when in-house research teams fail to publish unfavorable results. There are many clever techniques to do this some of which I will cover here. The most common is to run many small independent trials, then only publish the favorable ones. Rather than run a trial with 100 individuals you run 10 trials with 10 people. Because most data is messy, studies with small sample sizes will not be sufficient to capture genuine patterns. This means that there’s a greater likelihood of studies with false positives or false negatives. This method allows firms to generate favorable studies without engaging in outright fraud. Publishing only the favorable trials while scrapping the unfavorable ones biases the public research record. (This effect also occurs in normal science and is called the file-drawer effect. The difference is that it’s not premeditated or deliberate)

Fishing is another method. With any multi-variable data set you can keep looking for correlations until you find one–and you always will. Good science commits to a hypothesis and statistical method before the data is collected and the best science has a third independent party evaluate the data. By looking for correlations after data has been gathered, groups can find false positives which in turn allows them to publish studies the with appearance of favorable results.

Sponsoring Third Party Researchers

In-house science can’t be avoided. Most industries require in-house science for new product development. But in-house science makes the public skeptical (or at least it should). To dispel some of that skepticism, firms or industries can fund studies at universities or independent research groups. Although, researchers are required to disclose funding sources, the public rarely reads the primary research study and the disclosure may not be reported by the media (especially if it’s ideologically friendly media). Researchers may have received the funding because they are known to be friendly to the industry/interest group or may simply feel pressure to produce positive results to secure future funding for their lab. All this can cause biases–intended or not–to creep into the outcomes. Also, depending on the agreement, the group that funded the research may have the power to decline to publish a study if it yields unfavorable results.

Detecting Biased Production

Biased production is one major reason why you should never draw strong conclusions from a single study. First, even if all researches were angels, unintended biases can creep into studies. Second, most studies are not designed to be comprehensive evaluations of an issue. They are designed to illuminate a subset of an area of inquiry. This is why it’s absolutely critical to look for trends in the literature. Ask: What direction are most studies pointed? What direction do the largest and best controlled studies point? What do the meta-analyses and systematic reviews say?

For example, the tobacco industry, even up into the early 2000s, was able to produce studies casting doubt on whether second-hand smoke was a cancer risk. But if you looked at the trend it the literature, these studies ran against it.

But this isn’t the worst of it. What makes biased production so pernicious is that it can distort the trends in the scientific record. If a firm or industry group funds or conducts large numbers of studies with negative results and don’t publish them while publishing only positive results, positive studies will be over-represented in the trend. The opposite occurs if they don’t publish positive results. Biased production creates an illusion of scientific disagreement.

It’s impossible for a layperson on their own to detect biased production. If you think you can, you’re vastly overestimating your knowledge and competence in that domain (see Dunning-Kruger effect). To detect it, one has to be deeply familiar with the entire literature in that domain and know which labs are producing what, and who is funded by whom. This involves being a reviewer for journals, attending specialist conferences, producing independent research, and studying the domain for many years–not just googling what other people have said.

Fortunately, there’s a short cut. We don’t have to be experts in all things nor can we be. Identifying whether there is a consensus of experts on an issue gives the layperson strong reasons to adopt their view. Of course, experts can sometimes get it wrong, but a consensus among them is more likely to get it right and less likely to be fooled by biased production compared to the likelihood of a non-expert being fooled or getting it right.

To summarize: Propagandists use biased production in two main ways. The first is to generate congenial research. If they produce a study with congenial results you better believe they’re gonna promote it like heck to the media. Press releases all day! The public will hear a lot about a particular study but nothing about the general trend in the literature. This distorts how the public perceives the issue if the study is incongruent with trends in the literature. They will overweigh the single study. (see availability heuristic)

The second way is to distort the actual trends in the literature. By suppressing results in one direction, the other direction looks much stronger than it actually is. In this way, biased production can fool even the most scientifically literate since it’s extremely difficult to draw conclusions from literature that is hidden from you!