Argumentum Ad Hominem: Bias and the Case Against Industry-Funded Research
Here are some contentious questions guaranteed to start an argument:
- Do oil companies fund research that is more likely to minimize the threat of climate change or downplay the risks of fracking?
- Do tobacco companies fund research that is more likely to minimize the health risks of smoking?
- Do pharmaceutical companies fund research that is more likely to minimize the health risks of a novel drug?
For some people, these are rhetorical questions. Their intuition in such cases cries “yes!” Do maxims like “follow the money” and “don’t bite the hand that feeds you” spring to mind as relevant under these circumstances? For others, these questions are tantamount to ad hominem fallacies driven by a knee-jerk anti-corporatist mentality. Who is right?
Argumentum Ad Hominem
For many years, any ad hominem (“against the man”) argument was regarded simply as a logical fallacy. More recently, however, some argumentation theorists have distinguished both fallacious and non-fallacious instances of ad hominem. In other words, ad hominem should be regarded as an argumentation scheme and whether a particular argument following this scheme is fallacious (or otherwise ineffective) depends on how the argument fares in terms of a set of critical questions.
Suppose that, as a matter of simple statistics, the vast majority of arguments ad hominem are indeed fallacious. Even if we grant this statistical assumption, that does not warrant assuming that a particular argument under consideration commits a logical fallacy. For any given argument, due diligence requires that we honestly address the relevant critical questions, and ascertain how the argument answers them.
The ad hominem argument has the following general form:
X asserts that P.
X is C.
Therefore, presumably, not-P.
The argument has a variety of subtypes depending on how C is characterized. In each case of ad hominem, C is supposed to be relevant to the warrant that X’s assertion that P confers on P itself.
- In the abusive subtype, C impugns X’s character for veracity or rationality. If X is a liar or pathologically irrational, one is led to presume that the opposite of what X says is (likely to be) true.
- In the circumstantial subtype, C impugns X’s pragmatic consistency. If X acts (or has acted) in a manner inconsistent with X’s assertion that P, one is led to presume that X’s practice reflects the truth of the matter, rather than X’s explicit assertions.
- In the bias subtype, C impugns X’s impartiality or fairness. If X is unwilling to entertain any evidence or reasoning in support of not-P, one is led to presume that X is incapable of mounting any rational defense of P or refutation of not-P.
- In the poisoning the well subtype, C impugns X’s intellectual independence. If X is committed to a position on an issue in virtue of being in a role within an organization (e.g., an adherent of some cause), X’s assertion that P is presumed to be a “talking point” or an assertion “on message” rather than a judgment reached and validated independently. Asserting P becomes a function of X’s self-identification as a member of a group.
(Additional subtypes could well be explored for other attributes C. This list just accounts for the most commonly identified variants.) The ad hominem argument can now be unpacked a bit further as follows:
X asserts that P.
X is C.
If X is C, then X’s assertion that P makes P less plausible.
Therefore, presumably, not-P.
As stated, the conclusion may well be too strong, depending on the details of the case. A weaker, more nuanced conclusion might be: “Presumably, P should be regarded as less plausible on the basis of X’s assertion of it.”
It’s readily apparent that this way of representing an argumentation scheme is highly abstract. The analysis of a particular ad hominem argument would be made easier by fleshing out what the respective argumentation schemes are for each of the subtypes. The instantiation of the argumentation scheme for a subtype still may not be a specific argument, for the assertion that “X is C” may cover a range of importantly related cases, such as the bias alleged to be inherent in industry-funded research. In order to map a specific argument onto the argumentation scheme for ad hominem at a tractable level of abstraction, let’s first look at a stereotypical argument.
The Argument from Bias: The Tobacco Example
Let’s take the tobacco argument as an example, and try to make the tacit premises as explicit as possible. The argument might be made as follows:
ABC profits from the sales of smoking products. If the public thinks that smoking is harmful to health, then they will purchase fewer smoking products, and ABC’s profits will decline. Clearly, ABC doesn’t want that, since it is in business to maximize its profits and its returns to shareholders. ABC will therefore aim to suppress any characterization of the effects of using its products that, if publicly known, would lead to lower sales and profits. Among the scientific research initiatives into the effects of smoking, ABC would have the most control over the way in which the results are presented if it conducted its own internal research, or outsourced the research to an independent lab that it funded directly. Since ABC has an interest in promoting the results of research that downplay the negative effects of smoking, ABC executives can be expected to exert their influence over the scientists whose research they fund, leading to a bias among the ABC researchers conducting the investigations. It is therefore highly likely that ABC tobacco company funds research that is more likely to minimize the health risks of smoking than would the independent investigations of a third-party research team far removed from the influence and pressure of ABC to downplay the negative effects of its products. Whatever ABC cites as scientific research concerning the health effects of its smoking products should therefore be taken with a greater than usual degree of skepticism.
The “big oil” and “big pharma” arguments could be spelled out along similar lines, each argument implicitly relying upon a particular way of instantiating the bias-type ad hominem argumentation scheme. Abstracting from the details of the various industry groups, we can see a common pattern:
- The financial prospects or market reputation of a company X is such that they would be negatively impacted if claim P were true.
- C sponsored research R that is conducted by T, who is charged with ascertaining whether P is true.
- T concluded on the basis of R that P is untrue, unlikely, or unsupported by the data.
- Because of the circumstances of T with respect to X, T will be biased towards not-P.
- Therefore, presumably, P should be regarded as more plausible on the basis of T’s research.
I think the cogency of this overall argumentation scheme is what accounts for the appeal of dismissing industry-supported research as irremediably self-serving. Before we reach a verdict however, we still need to consider the critical questions for any argument based on the ad hominem scheme.
Critical Questions for Ad Hominem Arguments
If an ad hominem argument is acceptable (or cogent), then affirmative answers to the following critical questions should be forthcoming:
- Is it true that X asserts that P?
- What evidence is there that X is C?
- Is the fact that X is C relevant in the context of a critical discussion related to P?
- Is the weight of the presumption claimed in the conclusion strongly enough warranted by the reasons given?
If negative answers are not forthcoming to one or more of these questions, then the argument is a bad one – that is, it would not command the assent of an ideally rational listener. In particular, if there is no evidence that X is C, or it is irrelevant that X is C, then the argument commits the ad hominem fallacy.
(Some may disagree that my ad hominem argument is an example of the bias subtype, and insist instead that it is a case of poisoning the well. I am open to being convinced on this point, but I’m not so sure that for present purposes this is a distinction that makes a difference. If poisoning the well is itself a subtype of bias, then the tension is resolved – both attributions are legitimate.)
For the bias subtype, the critical questions are similar:
- Are the premises of the argument true?
- What evidence is there that T will be biased towards not-P?
- Is the fact that T is biased relevant?
- Do the reasons given provide sufficient warrant for the presumption that P should be regarded as more plausible?
The Presumption of Bias
Premise (4) in the argument pattern above reflects the cultural cliches “follow the money,” and “don’t bite the hand that feeds.” A bias that reflects the interests of the corporation, if present, might manifest unconsciously among researchers who are otherwise committed to being objective. Let’s assume that scientists by and large intend to be objective, i.e., that deliberate attempts to suppress evidence supporting unfavorable outcomes for a firm’s products are a rare occurrence. (Doubtless this does indeed sometimes happen, however.)
If the answer to our contentious questions is “yes,” it would suggest that (a) it is legitimate to presume in certain cases that unconscious bias is present (or at least likely) and (b) the bias that is supposed to exist would justify rejecting the claims made by those who are presumed biased (or in lowering our assessment of the likelihood of their assertions).
Is it legitimate to presume that bias exists, solely on the basis of cultural cliches? Recall that one of our critical questions implies that we ought to ask what evidence exists that the research team in such cases will be biased. If the requirement of actual evidence is taken seriously, it is difficult to see how it is legitimate to presume that bias is present (or at least likely) in such circumstances. If the requirement for evidence is disregarded, and a presumption of bias substituted for it, that would constitute a critical part of an unwarranted ad hominem case against the corporation and its interests.
How might we demonstrate the likelihood of bias in the case of ABC? We could review the published research coming out of ABC’s labs and compare it to the results of other studies done by different groups. Such comparisons might expose flaws in ABC’s methodology. The demonstrable presence of methodological flaws in a scientific study is grounds for skepticism, to the extent that those flaws undermine the logical basis for the conclusions of the study. For example, the authors of the study may use “novel” methods to manipulate the data in order to extract the desired “signal” or trend. They may throw out inconvenient data sets or be aggressive in purging outliers from the data. All of this might be evidence of bias, but whether those methodological flaws are themselves evidence of bias is, in a sense, beside the point. What warrants our withholding assent to the conclusions of a study is knowing that the research was flawed in particular ways, regardless of the etiology of those flaws.
While bias can show up as inappropriate data manipulation, it is just as often true in cases of outright bias that there is a discrepancy between the empirical data within the study, and the interpretation of those data in the text of the abstract, summary or conclusion. For example, if the authors of the study claim that an effect was insignificant when the data clearly indicate the effect was significant, that is evidence of bias. Similarly, if the authors exaggerate the significance of a negligible effect, that too is evidence of bias.
Suppose we were successful in establishing that ABC was biased in some way. Would that justify disregarding the conclusions of industry-funded research in general? Hardly – that could be a fallacy in itself: hasty generalization. We would need to do an extensive, systematic meta-analysis of the research on the health effects of smoking (or the efficacy and safety of a new drug) to determine to what extent industry-sponsored research manifests the marks of bias – data manipulation, data misrepresentation, etc. Obviously that meta-analysis would be difficult to do.
Since proof of systematic bias is elusive, isn’t it legitimate, on inductive grounds, to suppose that industry-sponsored research is more or less liable to be biased? No! That is exactly the issue that requires proof. Ultimately, this is an empirical question about the general quality of scientific research.
Of course, there are some studies that are flawed and that are industry-funded. But to suspect that industry-funding is a contributing factor to such flaws, we would need to know the base rate of flawed studies in a particular field, and then compare the proportion of flawed industry-supported studies to studies overall. If there is a statistically significant difference in the two proportions, that is evidence that industry-funded research is more likely to be substandard. Once that has been demonstrated, only then would it be fruitful to ponder whether bias explains the observed quality difference. (To my knowledge, no such research has been attempted. If it exists, I would welcome hearing about it.)
In the absence of inductive evidence about the overall quality of industry-funded research, or in the absence of a methodological critique of a particular study, we have no business being suspicious of some piece of industry-funded research. But this is not a license to accept whatever comes out of corporate-funded labs either. We need to subject the evidence and reasoning of industry-funded researchers to the same level of scrutiny and with the same standards as we would any more “impartial” researcher.
The key critical question for ad hominem (bias-type) arguments is the one that asks whether we have evidence of bias. If there is only a presumption of bias, and no evidence for it, then we have an instance of fallacious argumentum ad hominem.