Bad logic, but good game theoory

John D. Cook writes:

Ad hominem arguments are bad logic, but good (Bayesian) statistics. A statement isn’t necessarily false because it comes from an unreliable source, though it is more likely to be false. (...) Some people are much more likely to know what they’re talking about than others, depending on context. You’re more likely to get good medical advice from a doctor than from an accountant, though the former may be wrong and the latter may be right.

This is true, but I don't think it's the most important reason why ad hominem arguments persist. After all, they are used also (and perhaps mostly) against reliable sources of information. The reason is game theory. Most interactions we face are signaling games of partial conflict. We talk not just to transmit information, but also to influence other people's beliefs and actions in ways that are beneficial to us. We talk to persuade more than to inform. (Of course, communicating information is a necessary condition of persuasion. You cannot persuade anyone by talking to them if they do not understand what you're saying. Language could not evolve if all human interactions were zero-sum--there wouldn't be enough coordination to establish a common understanding of messages.) Persuasion consists of sending messages of the form "given your preferences, you should do X because Y." The logical validity of such message is of course completely independent of the sender's identity; however, the truth value of the conclusion (that doing X is good for you) is not. If someone gives you a logical argument for why it's good for you to do X, it's perfectly rational for you to wonder why they want you to believe that. Perhaps it's good for them, not so much for you. In other words, it matters who it is that's telling you this, for strategic reasons. Their preferences are relevant information in terms of assessing the likelihood of the conclusion being true.

What separates ad hominem as good game theory from ad hominem as fallacy is whether you're using information about sender's identity to update your beliefs about truth value of the conclusion or about validity of the argument. Here's an example. Suppose you are a (benevolent and completely ignorant) dictator of a medium-sized country. You wonder if establishing minimum wage requirements would help low-skilled workers and, in the process of your wondering, you ask the opinion of an expert economist. The expert comes back to you with the standard microeconomic theory argument that enforcing a minimum wage higher than the market rate increases unemployment among low-skilled workers. You then find out that your expert owns a whole bunch of enterprises that depend on low-skilled labor. It's logically valid and perfectly rational for you to use that information to postpone your decision with respect to the minimum wage (for example, until you get more opinions from experts who have no stakes in the conclusion). But, it's a fallacy if that information causes you to conclude that microeconomic theory is wrong.