Where Our Thinking Goes Wrong

The recent and still unsolved murders at the University of Idaho are horrifying and as much as I’m generally loath to use current, and tragic, examples, the investigation and public discussion highlight well cognitive biases that can slide unnoticed into our thinking and skew how we view a situation. Perhaps in this instance, where we are not the primary investigators, it doesn’t so much matter. But in the case of the investigators, it does, and in situations you face every day in your business and life it can matter a great deal when cognitive biases negatively impact your decision processes.

We are faced every day with information of uncertain quality, and we must make decisions with only partial and often ambiguous information. That’s not a problem, it’s just a fact of life. The problem comes when we decide to use this information and we fail to appropriately incorporate any uncertainty judgment into our thinking. Richards Heuer, in his seminal work The Psychology of Intelligence Analysis, discusses how we often fall prey to the “best guess” strategy, in which we fail to appropriately weigh uncertainty.

Take the dog that was in the apartment where the murders happened. I’ve heard and read several times that because the dog didn’t bark the murderer must have been someone known to the victims and the dog. The problem is, we don’t know the dog didn’t bark. We can infer the dog didn’t bark, but we should assign some uncertainty to that, say a 60 percent chance the dog didn’t bark. Unfortunately, even when we initially assign a level of uncertainty, once we incorporate the information into our analysis we very often, unconsciously, adjust the probability to around 100 percent. That is, we become certain the dog did not bark and we make assessments as if that is a definite situation rather than just a possible situation.

I’ve also seen most of the investigative focus on one victim, Kaylee Goncalves. Why is this? There were, after all, three other victims. Shortly after the murders became public, Kaylee’s sister, through a bit of good investigative effort, announced that her sister had tried several times to call her ex-boyfriend shortly before the time the murders likely occurred. This information may or may not be relevant, but because it was the first piece of the investigation we heard, we anchor on it and decide that the murders must somehow be related to Kaylee and not any of the other victims. This is anchoring bias and it gets in our way all of the time.

We then look at new information through this lens and we run the risk of discounting or discarding any new information that does not focus on Kaylee as the primary target. In addition to anchoring bias, we also have here a form of confirmation bias. We seek information that confirms our initial theory, e.g. someone was stalking Kalyee, and disregard information that may suggest another victim was the primary target or that they all were equal or equally random targets.

I just read a headline, one of many on the same topic, stating that the attacks were definitely personal. It is true that stabbing someone is a much different act than shooting them from afar, and it is true that some stabbings are personal and borne out of rage. It also is true that murders in which handguns or other firearms were used (in the United States) are about five times more prevalent than stabbing deaths. But that doesn’t mean Americans don’t stab each other to death on a fairly regular basis (at least three times a day) and worldwide the number is much, much higher because guns are harder to come by.

Based on our commonly held stereotypes of how stabbings work, and because most of us rightfully would struggle to imagine stabbing someone, we fall prey here to illusory correlation bias. We look at stabbings and see that some of them are personal, but we fail to look at all those stabbings in which a knife was either just the most available or the most convenient weapon (e.g. where one is going to murder someone in a residential area and does not want the attention a gunshot would draw). We also don’t consider all the shootings in which there was personal anger. A true to life correlation table would probably look something like this:

   Stabbing Not Stabbing

Personal 25 25

Not Personal 25 25

 In other words, there is no true correlation. But because we can easily call to mind instances in which knife attacks were personal, and instances in which shootings were seemingly random (availability bias), we interpret the data more like this:

  Stabbing Not Stabbing

Personal 100 0

Not Personal 0 100

And thus we see correlation where there isn’t any and move forward in our investigative or decision-making process based on a potentially flawed understanding of the actual probability.

This now brings us back to the problem of confirmation bias. Because we, incorrectly, believe there is a high correlation between stabbings and personally motivated murder, we could fail to look at other equally probable options. The victims may well have been chosen at random, and we run the risk of missing this when we see stabbings as “crimes of passion” and more so when this theory is announced to the public.

There have been many missteps in this investigation, but to be clear, I am not faulting the investigators. I’ve been in very similar situations, with no idea who committed a brutal crime and grasping to find any sensible theory. Making decisions and deciding on best courses of action in critical situations is always difficult; however, if we can become more aware of the things that get in the way of our decision making – which they do every day – then we can substantially improve our thought processes and analysis of information and intelligence as we work toward resolution and understanding. In other words, enhanced awareness of our biases allows us to see more true connections rather than just those we hope to see.