The Facts Speak For Themselves

I recently went to an ACOG-endorsed educational activity in Washington DC. The speaker, from a nearby local university, fancied himself an expert on the subject matter. The subject matter itself is quite controversial, so his opinions about it are not the opinions of all experts. At one point, he presented two papers which disagreed with his main narrative, both published very recently and both containing more empiric data than any of the previous papers on the subject which had established the main dogma that he advocated for and was a proponent of. He quickly brushed off the first paper with the following quote: “I’m always worried about authors’ own interpretation of their data.”

This remark is simply striking and it embodies the cognitive mistakes that pseudo-intellectuals often make while attempting to dress up opinion as science. These pseudo-intellectuals believe that data is king. They will invoke the authority of data at every turn as if there is only one interpretation of data and as if all data speak some self-evident truth. The same speaker later glibly stated that “these data support the fact that…”, and in so stating he revealed his true predisposition. He believes that he already knows what the facts are, and he must find data that can be presented in a manner which supports those facts.

Consider again his first nonsensical statement. Do not all authors publish an interpretation of their data? Does he believe that scientific papers merely contain raw data, stripped of context and interpretation? Of course not. Every scientific study ever published contains an interpretation of the data provided by authors. I will admit, that most of the time, either the data or the interpretation is wrong. We have good evidence that up to 80% of published scientific findings are incorrect. But he was not questioning the paper merely on those grounds, he was upset that the interpretation of the data did not support the facts that he already believed to be true and therefore cognitive dissonance occurred. Had he been the author of the paper, the same data would have been presented with opposite conclusions. At some level, perhaps subconsciously, he understands that he could have and would have written the paper in a different way; at the same time, however, he believes that the facts speak for themselves because he has lived his entire adult life restating this argument-ender over and over again.

But if the facts speak for themselves, then how is it possible that he would have written the paper differently than those authors? This contradiction is at the root of why most published research papers are incorrect. The truth is, facts or data do not speak for themselves because all facts or data must be provided in context. That context forms the narrative. That narrative is what we take away from observational studies. The context that underlies such a narrative is the product of our own cognitive biases, past experiences, and personal hypotheses. So two people can look at the same data and draw opposite conclusions.

When we do a scientific experiment, we tend to disprove the null hypothesis. With the null hypothesis disproven, we often make the mistake of assuming that our alternative hypothesis is then proved. Worse, we make the mistake of believing that there is only one alternative hypothesis. The pseudo-intellectual speaker believes that his alternative hypothesis has more validity than the other guy’s alternative hypothesis, when in reality it is likely the case that neither hypothesis is correct, but only that the null hypothesis is likely rejected.

Consider these facts:

  1. The police arrest a man with a long criminal history, including imprisonment for breaking and entering, larceny, assault with a deadly weapon, and attempted murder.
  2. At the time of the arrest, an officer on patrol ran into a pawn shop after hearing gunshots. He found the man holding a gun in his hand and the owner of the pawnshop was shot dead on the floor. The man was standing by the cash register which had been busted open with the gun.
  3. The gun had the serial number filed off and the man’s fingerprints were on the gun and trigger. There was also gunshot powder residue on his hand and sleeve showing that he had recently fired again.

So these are the facts, and no doubt the prosecuting attorney would tell the jury that these facts speak for themselves. He would provide the context and narrative for the jury to interpret these facts. It might go something like this,

This man, a hardened criminal with a long track record, stormed into the pawn shop demanding money with a gun that he purchased illegally. When the pawn shop owner stood up to him, the man fired two shots. One shot just missed him and was found an inch deep in the wall by the side door. The second shot hit the man cleanly in the heart and killed him dead. The patrol officer ran in just after hearing the second shot fired and found the defendant standing over the cash register which had just been busted open with the butt of the gun.

The prosecutor will call for the man to be strung up on the nearest lightpost. Then the defense attorney will explain the same facts but in a different context and with a different narrative. Her speech might go something like this,

Yes it’s true that this man has lived a hard life and done some unforgivable things. He lives with the guilt of his past every day. After he served 15 years in prison for nearly killing a man over $800, he entered society a changed man. While in prison, he had a religious conversion. He earned his GED. He was committed to never being the man he was again. When he got out, he lived in a halfway house in the worst part of town. He walks to work every day to the grocery store a block away from this pawn shop. On this day, as he walked by he heard a shot and saw that the glass of the door had been broken. He looked in saw the pawn shop owner lying on the floor gasping for breath and immediately rushed in to help. As he did, he saw another man busting the cash register open with again. Luckily, as the man hit the register with the butt of the gun, he dropped the gun to the floor and then the defendant rushed at the man and hit him. The man quickly ran for the side door and my client picked up the gun and fired a shot as he fled. Just then, the patrol officer stormed into the shop.

So which story do you believe? If you are a more clever writer than I am, you can probably come up with a dozen different stories that are supported by the three facts that we all believe to be true. How do you know which one is true? The truth is you do not. The best we can hope for is determining which scenario is most probable. There’s also a possibility that some of the facts as presented are not accurate, but even if they are, it’s not the facts that matter, it’s the narrative.

Where does narrative come from? Narrative is pieced together by our brains and your narrative might be slightly different than mine because your biases and experiences are different than mine. If you are an experienced police officer or district attorney, you are used to the excuses and hard luck stories of every criminal you’ve ever seen. You’ve heard every excuse and you will be convinced that justice was not served if the man is acquitted regardless of what evidence is presented in court. If you are the defendant’s mother, you are just as certain that he is innocent and that he is a truly changed man who is morally and physically incapable of killing someone.

If you are a scientist who has spent his entire career trying to find evidence that proves your view of a particular area of science, then you are just as blind and biased as the mother and the police officer. Such bias is inescapable and must be accounted for when determining both who’s innocent and guilty in court and which papers present valid hypotheses and which do not.

I was excited to see the cover of the Winter, 2017 edition of The PharosThe Pharos is the magazine of the Alpha Omega Alpha Honor Medical Society. You can read it for free here.

If you’re a regular reader of howardisms, you can imagine my excitement when I saw that the lead article was going to deal with the most important issue facing diagnostic error and subpar medical practice today. There has been a lot of interest in and focus on the subject since the Institute of Medicine published a follow-up to To Err is Human entitled Improving Diagnosis in Healthcare in 2015 which highlighted congnitive bias a leading cause of medical mistakes.

But then I actually read the article written by Pharos editor Richard Byyny. The article isn’t about cognitive bias and how it affects our clinical reasoning processes at all; instead, the article is about racial, gender, and other similar types of discrimination. While the author starts out strong with a discussion of the neuropsychology of cognitive bias and an explanation of Dual Process Theory, he quickly leaves these solid footings and ventures out into a discussion of implicit bias as it refers only to the above mentioned discrimination topics.

This article, which does not mention that at least 10% of all patients have a diagnostic error attributable to cognitive bias, does significant harm to efforts to educate physicians about cognitive bias and to help them overcome it. Why? Because it continues to perpetuate myths about cognitive bias and System 1 thinking. The author condemns and vilifies implicit bias and the System 1 thinking that relies upon it as inherently bad, nearly always resulting in bad outcomes, and also claims that we can overcome them by being conscious of them and changing them. He repeats the tired canard that bias in our brains is produced by fear.

These main beliefs about implicit cognitive bias are always espoused whenever someone focuses on the incredibly narrow application of bias that relates to racial or gender discrimination. But they are fundamentally not true. The author of the article has a bias about bias. His lack of understanding about what cognitive bias really is and what it’s importance really is will blind him from being able to do anything about it.

Implicit bias is essential for our survival. Most of the time our biases serve us well. Yes, many of them protect us from danger; but most of them save us valuable time so that we can actually live and survive in a world that would be overwhelming without them. Yes, they are shortcuts, but almost always these shortcuts produce the right result. System 1 thinking is the cognitive framework which utilizes these time-saving and life-saving heuristics. We cannot live our lives without using implicit bias in almost every decision and thought that we have. A person who thinks that he is free of implicit bias or that he has used metacognition or other tools to purge his brilliant mind of it is the most dangerous of all because he ceases to recognize how pervasive his bias truly is. He is the guy at the meeting who thinks that the other author is biased in his interpretation of facts but that he himself is unbiased.

Most System 1 decisions produce good outcomes. Most biases and stereotypes are true. The criminal holding a smoking gun over a dead body five seconds after the shot  is probably guilty and the cop who draws his firearm to subdue a suspect in that situation does not have time for System 2 thinking. That’s how you get dead cops.

If this is all true, then why are there almost 100,000 words on howardisms talking about how bad our cognitive biases are and why they result in medical error? Because we can always do better. We don’t need to get rid of cognitive bias and System 1 thinking, but rather we need to realize that that’s how our minds work and look for situations where such thinking might produce a bad outcome. Sometimes we need to use System 2 thinking, but it’s probably less than 5% of the time in clinical medicine. We also need to do a better job of producing high-quality and evidence-based heuristics; in other words, we need to make sure we have the right biases.

Sometimes a guy standing over a dead body with the smoking gun is innocent, but the exception doesn’t disprove the rule. We shouldn’t just convict every defendant at face value, but about 95% of the time the face value is correct. This secret is understanding when that other 5% happens and how to recognize it. This is true in surgery and it’s true in life.

The facts don’t speak for themselves, we speak for them. When we do, we do so with bias and preconceived ideas of what the truth is. I’m not asking you to change, I’m asking you to be open-minded and recognize that you might be just as wrong as the next guy.