Sunday, February 15, 2004
Update: I've been floating this argument with my colleagues and the consensus is (a) that I'm nuts; and (b) that I'm not totally nuts. In light of those discussions, I want to offer the following clarification. While I maintain that ad hominem attacks sometimes legitimately allow for testimony to be excluded from consideration, one would be guilty of the ad hominem fallacy if one held that the legitimacy of this exclusion was tantamount to a showing that the testimony was itself flawed. I didn't mean to be saying anything stronger than this, but I think I did in a couple of places.
Over at .jasonblog. a couple of posts in the catastrophe corner have generated a tremendous amount of discussion. The general topic of the posts is whether the concerns raised by the PEAK Oil site give us reason to think that an inevitable spike in the price of oil will bring down human civilization. One argument for the claim that they don't provide such a reason turns on the assertion that the folks behind PEAK Oil are nutjobs. It has been objected that this argument is unsound because it commits the ad hominem fallacy.
In this post I will argue that the argument in question doesn't commit the ad hominem fallacy as that fallacy is properly understood. In addition, I'll float an unrelated speculative argument in favor of ad hominem reasoning. I begin with a short exposition of the fallacy.
It will be helpful to have a definition. The following definition is typical of those found in introductory texts:
The ad hominem (Latin for "to the man") fallacy directs attention away from evidence for the claim to a person making the claim. The social, political, religious, or other views of the person or the person's character, personal background, or self-interest is focused on in order to support or discredit the statement the person has made. "You shouldn't listen to what she says, because she's a Communist (or something else)" is sometimes used to discredit a person's statement. But this works only if we assume that all Communists (or other group) are mistaken about all matters, which is, of course, false. [source: A Preface to Philosophy, Mark Woodhouse]
For convenience, let's introduce the following terminological conventions. Let's call the person against whom an ad hominem attack is leveled the 'agent.' We'll refer to the conclusions advanced by the agent as 'claims' and to the reasoning given in support of those claims as 'testimony.' And we'll use the term 'trait' to refer to whatever it is that is said of the agent. Then we can say that when an ad hominem fallacy is committed an agent's claims and the testimony given in support of them are undermined on the basis of traits ascribed to the agent.
The ad hominem fallacy is an informal, rather than a formal fallacy. This means that its status as a fallacy isn't related to the logical form of the argument, but is tied instead to the substance of the reasoning. Because the fallacy isn't one of logical form, it is possible for two arguments to have the same formal structure but for one to commit the fallacy while the other doesn't. So, in order to decide whether a given argument commits the ad hominem fallacy a judgment must be made as to whether the trait ascribed is relevant to the conclusion.
With this understanding in mind, I want now to introduce a logical form which some fallacious ad hominem attacks seem to me to share with certain arguments that don't commit the fallacy. That form is:
Let C be a claim about some subject matter S, and let A be some agent and T be some trait. Further, suppose that the only evidence for C is the testimony of A.
(1) A has T
(2) Having T undermines the credibility of testimony regarding S
(3) All evidence for C has been undermined
hence, (4) C ought not to be accepted.
There is a suppressed premise, namely that one ought not to accept claims for which all evidence has been undermined. We'll leave that aside.
Whether or not the inclusion of ad hominem reasoning is legitimate turns on the status of premise 2. Typically, this is an empirical question. An example of what might be considered legitimate ad hominem reasoning is: "The only evidence we have that Jake served in the Alabama National Guard is his word, but Jake is an inveterate liar." My claim here is that the empirical fact that Jake is an inveterate liar, if true, would serve to undermine all of his testimony regardless of content.
This doesn't mean that a document referred to by Jake must be entirely disregarded. My claim is that the reference to the document, though not the document itself, is part of Jake's testimony. So Jake's claim that a document exonerated him, as testimony, would not give us a strong reason to search for the document. But if the document were provided and it did, in fact, exonerate Jake, then its testimony wouldn't be undermined by facts about him.
The point is that a trait ascribed to Jake legitimately undermines the status of his testimony. Knowing that Jake is a liar, we may legitimately disregard the whole of his testimony. We are not obligated to go through it in an attempt to discover precisely where it diverges from the truth.
I want to note that the liar example is not uncontroversial. The typical examples of legitimate ad hominem reasoning involve a logical structure that doesn't directly concern the evidential standing of testimony. For example, consider Woodhouse's discussion of legitimate uses of ad hominem reasoning:
Whether a nominee for the president's cabinet is an alcoholic, for example, may be a very important issue insofar as it pertains to job performance. And if someone states that she is morally opposed to gambling, it would support a charge of hypocrisy to note that she had a financial interest in a casino. What would be fallacious would be to assert that an alcoholic's views concerning beverage taxes must be false, or that a gambler's arguments concerning the legalization of parimutuel betting must be unsound because of their personal histories.
Woodhouse's cases of legitimate ad hominem reasoning are cases where traits of the agent are relevant to the truth of the claim they are making. What I have been arguing, however, is that sometimes traits of the agent are relevant to the status of their testimony. As such, it might seem that I am committed to the following textbook example of the fallacy:
"What she says about Johannes Kepler's astronomy of the 1600's must be just so much garbage. Do you realize she's only fourteen years old?" This attack may undermine the arguer's credibility as a scientific authority, but it does not undermine her reasoning. That reasoning should stand or fall on the scientific evidence, not on the arguer's age or anything else about her personally. [source: Internet Encyclopedia of Philosophy]
On my understanding this is an example of the fallacy because the age of the agent is not relevant to the status of her testimony. My claim is that there can be features of the agent which are relevant in this way because they undermine the status of the testimony as reasoning. Denial of this claim would constitute a rejection of my liar example. But such a denial would also mean that every instance of the logical form given in three is an instance of the fallacy. If that were so, then the failure of the example would teach us an important lesson -- it would teach us that the fallacy is formal rather than informal. This strikes me as an extremely unlikely result. However, if it were accepted the speculative argument I give later would remain unaffected.
Consider this argument: (1') the author of the PEAK Oil page is a nutjob conspiracy theorist; (2') the testimony of nutjob conspiracy theorists does not constitute evidence for any claim; (3') the only evidence for the PEAK Oil catastrophe claim is the testimony of the nutjob conspiracy theorist; hence, (4') we ought not to accept the PEAK Oil catastrophe claim.
I claim that this argument does not commit the ad hominem fallacy. Those who have, on .jasonblog., disagreed with this assessment have granted 1' and would, I think, grant 3' as well. The accusation of fallacy rests on the claim that their testimony cannot be judged based on the trait ascribed in 1'. This is what 2' denies. And I have argued that the evaluation of premises like 2' is an empirical matter. If 2' is false then the argument is fallacious.
In this section I want to float a speculative argument in support of ad hominem reasoning. The argument relies crucially on a distinction between reasoning and reasoners. That is, we can distinguish between the criticism appropriate of fallacious reasoning and the criticism appropriate to the epistemic agents who engage in it. I make the general claim that reasoners are not always properly criticized even when their reasoning is fallacious. In particular, I claim that there are many instances where exemplary epistemic agents engage in fallacious ad hominem reasoning.
The first move in the argument is to notice that epistemic agents are finite. As such, they are limited in the expertise they bring to an inquiry and in the time that they can devote to furthering that expertise. Even if all agents had equal cognitive capacities, they would, owing to their different histories, diverge in their degrees of expertise about particular matters of fact.
Because epistemic agents vary in their expertise about particular matters of fact, they vary in their ability to evaluate one another's testimony. For example, a particle physicist may be a fairly competent evaluator of the testimony of a chemist, but might not be competent to evaluate legal arguments. Similarly, a lawyer might do a good job with legal arguments while being unable to parse articles published in Nature.
Because epistemic agents have finite time, they can't always develop the expertise necessary to evaluate particular arguments. Moreover, even if they have the requisite expertise, or could easily acquire it, they must consider whether investigating the argument is a worthwhile investment of their time.
What emerges from these considerations is the necessity for standards which agents can use when deciding whether the testimony of others is worth attending to. This is a need which is based in fundamental facts about human reasoners, facts which have been true throughout human history and, thus, throughout the period during which our canons of reasoning were formed.
The application of such standards will sometimes involve commission of the ad hominem fallacy. The practice is nevertheless, I claim, embedded in our accepted norms of reason, and necessarily so. Hence, agents who engage in ad hominem reasoning sometimes exemplify our best epistemic practice.