Dana stopped drinking in February and her daughter was getting married in June. By April she had built the case for one glass of champagne at the toast, and the case was good. Champagne had never been her drug. One glass at a toast is not a return to anything. Refusing it in front of two hundred people makes her the story on a day that belongs to her daughter. Her counselor had said in March that all-or-nothing thinking carries a risk of its own, and some people do moderate, and the research on that is not nothing.
Dana is invented.
Good here means it holds up when you go looking for the weak joint. It does not mean taking the glass is wise.
This post is about motivated reasoning, which most people have backwards. Motivated reasoning does not produce flimsy excuses. It produces arguments that hold up under inspection, because an argument that fell apart would not persuade the person making it.
What she was doing has a name
Ziva Kunda wrote the paper on this in 1990, and the one-line summary of it that everybody repeats is wrong in a way that matters.
The summary is that people believe what they want to believe. What she actually described is a search. Real memory, real evidence, a genuine reasoning process, running with a thumb on the scale about which results come back. She called the two versions accuracy goals and directional goals, and the difference is not whether you reason but which reasons you happen to find.
Then she put a fence around it, and the fence is why the paper is still the one people cite. People come to believe what they want to believe, she wrote, only to the extent that reason permits.
The fence is the interesting half
To satisfy that constraint, the argument has to hold up. Kunda's account has people constructing a justification that would persuade a dispassionate observer, and, crucially, not knowing they are doing it. She calls the sense of having been objective an illusion, produced by the same process it is supposed to be evaluating.
Run that forward one step. The fence cuts two ways, and the first way is the one people miss. Sometimes reason does not permit, the search comes up empty, and the person does not reach the conclusion they wanted. Kunda's claim is not that wanting a conclusion delivers it. Her claim is that wanting a conclusion delivers it only when memory and evidence can be made to support it.
Which means that when the search does succeed, what comes out the other end is a good argument. It has to be. A weak one would not clear the fence.
So a bad argument is a genuine signal. If you hear one, motivation probably was not what produced it, because motivation cannot produce one. The reverse does not hold, and this is where the instinct runs backwards. A good argument does not clear the person who made it, and it does not convict him either. Quality is what the process delivers when it works, so quality settles nothing in either direction. Every premise Dana had was true. The wedding is one afternoon. Champagne was not her drug. Her counselor did say that. Some people do moderate and there is a literature on it.
There is a second-order version of this that is worse. Someone who reasons well has more material. A larger stock of true premises is a larger stock of premises a one-sided search can come back with.
Why rationalization is the wrong word for it
The word people reach for here is rationalizing, and it carries the wrong picture. Rationalizing sounds thin. It sounds like a story somebody waves a hand through, the kind a friend could puncture over coffee.
What Kunda described does not produce thin arguments. It cannot, because a thin argument fails the test the process is running.
I would go further and say the reasoning that ends a stretch of somebody's recovery is typically tight, specific, and already rehearsed against a skeptic who could not knock it down. That is my own read and I mark it as one, because nobody has measured the distribution of reasoning quality in that situation.
Anybody waiting for a bad argument to show up is waiting for the wrong thing.
The thing that is actually visible
The felt sense of having thought it all the way through is what the process produces. So checking your own reasoning harder does not help much, because the checking runs on the same machinery.
What is visible is the shape. Dana had built a case and had not built the other one. That asymmetry can be seen from outside the argument by anyone who bothers to look, in a way the merits cannot, because evaluating the merits means entering the argument and the argument is good.
The signal is not clean. Plenty of people build one case and no other because they were right the first time, or because they built the other one months ago, or because the question did not warrant the work. One-sidedness is common in people who are not fooling themselves at all. What it does is narrow the field to cases worth a second look, which is different from catching anybody.
It is also visible to the person themselves, occasionally, and not by introspection. It shows up as a factual question rather than an act of self-examination. When did I last spend twenty minutes assembling the case against this. That question has a yes or no answer, it does not require deciding whether any particular premise was true, and it can be answered honestly by somebody who is in no condition to evaluate their own reasoning.
What that leaves exposed
The claim that a bad argument is a signal does not hold. It requires motivated reasoning never to produce a weak argument, and Kunda does not say that. Her constraint concerns when a search for a desired conclusion succeeds. It says nothing about what people do when the search fails. They often hold the conclusion anyway on an argument that would not survive an afternoon.
Motivation can produce a bad argument, so the signal I offered is not available. One-sidedness remains, and I have already conceded that it is weak. The remedy has its own problem. Noticing that you built one case and not the other sits close to the procedure William Miller withdrew from his method after finding that it reduced commitment to change in ambivalent people. I distinguish noticing from building. Nobody has tested whether that distinction holds in the week that matters.
The obvious remedy, and why it may be the wrong one
The move that suggests itself is to build the other case with the same effort. Sit down and construct the strongest version of the argument against, the way you constructed the one you kept.
That is a real technique with a name and it does not do what you would expect. Weighing the pros and cons of a change is called decisional balance, and William Miller, who built motivational interviewing around it, published a paper in 2015 arguing that running it with an ambivalent person tends to decrease their commitment to change. He took it out of his own definition of the method. His reason is worth sitting with: a carefully counterbalanced list of pros and cons is not a cure for ambivalence, it is an operational definition of ambivalence.
So I am not going to end by telling anybody to argue both sides with themselves, because the person who knows this best found that it backfires in exactly the situation this post describes.
What survives is smaller. Noticing that you have built one case and not the other is information. Building the case against, alone, in the week the case for arrives fully assembled, may be the worst available response. This is where a post stops being useful and a person is needed.