Justifying our Decisions: Great for Plausible Deniability, not so Great for Medical Diagnosis

Sadly, Larry had always approached from the side that wasn't posted and a natural treasure was destroyed before anyone could react

Sadly, Larry had always approached from the side that wasn’t posted and a natural treasure was destroyed before anyone could react

I am looking at the idea of “justification” as discussed separately by Kenneth R Hammond, Jonathan Haidt, and Steve Catty and Jamin Halberstadt.

Ken Hammond created the JDM metatheory dichotomy of coherence and correspondence.  Coherence tests decisions on rationality while correspondence tests decisions on empirical accuracy.  Coherence advocates start with the mind of the decision maker.  In examining rationality of judgment, the main criterion is consistency.  Bayes’ theorem is the model for mathematical coherence of decision making. Coherence focuses on justification.  It describes departures from “ought” coherence.

In a 1996 article, Hammond discussed the work of D M Eddy in examining medical diagnosis.  Eddy demonstrated the incompetence of physicians in the justification process. (base rate error, etc)  Hammond considered Eddy’s work to be a landmark that made clear that formal reasoning demanded for justification does have to be taught, and when it is, it can be applied.  However, he noted that Eddy focused on the process of justification that follows the inference by the physician that the patient has breast cancer.  To examine the actual inference, Hammond suggested that the correspondence metatheory is needed.  Correspondence would start with the focus on nature, the patient and mind.  Correspondence theorists assume that the natural world has tangible indicators that in the medical context are known as signs and symptoms.  These indicators are uncertain in that not many signs or symptoms can be tied to one and only one underlying cause.  Thus, the physicians must rely on multiple fallible indicators.

Hammond concludes that coherence and correspondence should be used in a complementary manner since both perform best at different problems and appeal to difference criteria for evaluating judgment and decision making.  With regard to justification, Hammond concludes that physicians were not good at providing rational justification of their original inferences, but that correspondence tests were needed to evaluate the original inferences.

indexJonathan Haidt is the author of  The Righteous Mind  Why Good People Are Divided by Politics and Religion.  The book is directed toward the lay reader and has proved very popular.  Haidt introduces justification in two chapters revealingly titled:  “The intuitive dog and the its Rational Tail” and  “Elephants Rule”.  He quotes Howard Margolis:

 Given the judgments (themselves produced by the non-conscious cognitive machinery in the brain, sometimes correctly, sometimes not so), human beings produce rationales they believe account for their judgments.  But the rationales (on this argument) are only ex post rationalizations.

    Haidt  explains that judgment and justification are separate processes.  This certainly matches with Hammond’s inference and justification in medical decision making.  Haidt concentrates on moral reasoning, but gives it broad application.  Haidt states that:  “We do moral reasoning not to reconstruct the actual reasons why we ourselves come to a judgment; we reason to find the best possible reasons why somebody else ought to join us in our judgment.

Haidt creates the useful metaphor of a rider on an elephant for our minds.  The rider is rationality and the elephant is intuition and instinct.  The rider evolved to serve the elephant.  Intuitions come first and strategic reasoning second.  As Haidt says, if you want to change someone’s mind, “talk to the elephant first”.  People are good at finding ways to not violate their intuitions.  They only need one reason to doubt your argument or conclusion if it violates their intuitions. The rider’s main job is to justify what the elephant does.  The rider is like the elephant’s lawyer.  The lawyer is more concerned with his client’s reputation than the truth.

Haidt quotes Phil Tetlock, an accountability researcher:

A central function of thought is to making sure that one acts in ways that can be  persuasively justified or excused to others.  Indeed, the process of considering the justifiability of one’s choices may be so prevalent that decision makers not only search for convincing reasons to make a choice when they must explain that choice to others, they search for reasons to convince themselves that they have made the right choice.

 So our conscious reasoning tries to persuade rather than discover.  We are more like politicians looking for votes than scientists looking for the truth.  When our motivated reasoning wants to believe something, it needs just one piece of evidence, even pseudo-evidence, to give us permission to believe.  We have a justification if someone asks.  But, when we do not want to believe, we ask if we must believe.  And again, we only need one bit of contrary evidence to dismiss the belief.

Haidt seems to illuminate why Hammond’s physicians might not be so good at coherent justification.  Bayes theorem is only a couple of hundred years old so the evolved mind is unimpacted.  The rationality add-on package was not designed for truth but instead to protect reputation.  This makes sense if we believe that human cooperation is adaptive and that we play “tit for tat” to achieve it.  Haidt might benefit from discussing how Hammond’s correspondence theory and Brunswik’s lens model impact the intuitive elephant.

Another sort of justification exercise is to tell people that they will have to provide reasons for an attitude or prediction along with that attitude or prediction.  This has been shown to change the information on which an attitude or prediction is based.  They provide reasons they believe justify the attitude, but not having access to the intuitive information, they can only provide reasons that are easiest to put into words or that are most accessible or plausible.  As discussed by Catty and Halberstadt, they then report and temporarily adopt the attitude or prediction justified by this biased set of reasons.  This can reduce performance.  It may reflect disruption of simple heuristic processes.  One task referenced in a study was predicting the results of  college basketball games.  The familiarity or availability heuristic can provide good performance for such tasks.  To use Haidt’s metaphor, the rider may be told the best answer by the elephant, but if the issue does not have a strong social component (moral, political, etc),  the rider when forced to provide justification may change his answer to a wrong one.

To summarize, humans are often justifying.  If the subject is social, moral, or political, the justification typically does not impact the decision. If the subject requires rational justification that is consistent and coherent, humans need rigorous training or they will get it wrong.  If the subject is not that important and not moral or political, the justification process may actually get us to change our minds to make the justification work.

References

Catty, S. and Halberstadt J (2007) The Use and Disruption of Familiarity in Intuitive Judgments, in Intuition in Judgment and Decision Making.  Edited by Plessner & Betsch.  New York:  Lawrence Erlbaum Associates.

Haidt J.(2012) The Righteous Mind  Why Good People are Divided by Politics and Religion. New York:  Pantheon Books.

Hammond, K.R. (1996) How Convergence of Research Paradigms Can Improve Research on Diagnostic Judgment. Med Decis Making 1996;10:281-287.

4 thoughts on “Justifying our Decisions: Great for Plausible Deniability, not so Great for Medical Diagnosis

  1. Pingback: The Invisible Gorilla | Judgment and Decision Making

  2. Pingback: Everyone else is a hypocrite | Judgment and Decision Making

  3. Pingback: Simultaneous feeling and deciding | Judgment and Decision Making

  4. Pingback: Future of the Mind - Judgment and Decision Making

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.