Thomas Z. Ramsøy

Brain noise and bad decisions

The effect of “brain noise” on decisions is now becoming the new rage in behavioral economics. In Daniel Kahneman’s recent book, Noise: a Flaw in Human Judgement, it is clearly described how different types of noise can affect our judgement and decision-making. Let’s take an excerpt from the description of the book:

Imagine that two doctors in the same city give different diagnoses to identical patients—or that two judges in the same courthouse give markedly different sentences to people who have committed the same crime. Suppose that different interviewers at the same firm make different decisions about indistinguishable job applicants—or that when a company is handling customer complaints, the resolution depends on who happens to answer the phone. Now imagine that the same doctor, the same judge, the same interviewer, or the same customer service agent makes different decisions depending on whether it is morning or afternoon, or Monday rather than Wednesday. These are examples of noise: variability in judgments that should be identical.

Book description at Amazon

In other words, noise seems to affect our decisions in often detrimental ways! Noise and its effect may even seem random, but let’s see if that holds up.

So what is noise actually?

The view taken by Kahneman and authors tend to focus on the way in which we humans are not rigid and automated decision-making machines. From this perspective, two distinct noise factors come into play:

  • Inter-individual noise: this is a term for the “random” variance that we see between people. You and I may differ in our ability to remember things, to be willing to take risks, and how much we are scared in a horror movie.
  • Intra-individual noise: people can vary over time, and our preferences can shift from time to time. Over the course of our lives, we change. Even within a single day we can change

In an interview, Kahneman states this as follows:

I’ve been studying bias all my life, but a few years ago encountered an instance of noise, and I was very impressed both by how much noise there was (among underwriters judging exactly the same thing) and mostly I was impressed by how little people knew about it.

Daniel Kahneman, at the Behavioral Scientist

Is it noise, though?

The main claim here seems to be that the variance in behavior is, to some extent, “random” and unaccounted for. But is it really noise? Perhaps what is perceived as noise merely happens because we don’t have a proper model understanding of the human decision-making apparatus, both within and between people.

A bit more than a decade ago, I coauthored a paper entitled “How genes make up your mind.” As the title shows, our genetic makeup can actually account for some of the variation between people. For example, if you are more scared than me at the movies, it might be that you have one particular type of genetic makeup of your serotonin transporter system (for the curious, it’s the short-allele carriers of the serotonin-transporter length 5-HTTLPR polymorphism).

The same happens for risk taking. Here, studies have shown that people who have dopamine DRD4 7-repeat allele carriers take 25% more risk, while people with the serotonin-related 5-HTTLPR s/s allele carriers take 28% less risk.

In other words, the genetic makeup of a population can account for some of the variation we see between individuals. This is very different from being “noise” and suggests that when we have a more sophisticated view of the human brain and mind, we can explain things that would otherwise be seen as noise.

I’m still me, but I choose differently

How, then, about changes within the same individual? In our paper, we also showed how psychoactive drugs and medications can change behavior. This happens through the regulation of the previously mentioned neurotransmitters such as dopamine and serotonin.

For example, one typical treatment of Parkinson’s Disease is to increase the level of dopamine. Here, one of the often seen side effects is behaviors and experiences that are often seen in high-dopamine sufferers such as schizophrenia and pathological gamblers, ranging from delusions and hallucinations, to abnormal risk taking. By contrast, treating schizophrenics can have the side effect of showing Parkinson-like symptoms (e.g., mental and bodily rigidity).

Within-subject variance can also be seen when we’re hungry, have too much coffee, and other changes in our state and homeostasis. But to say that this is noise is really not warranted. Instead, we should treat this variance as something we need to understand better, so we can come up with solutions for knowing when to decide, and when to take a break with a meal or a coffee.

And what ever happened to updating our preferences? In my 2012 coauthored paper “Branding the brain” we suggested a model that explained how consumers change over time as a result of their experience with branded products. Think now about combining this model with the aforementioned genetic variance: due to their genetic makeup, some consumers may only require one-off learning, while others are slower to learn brand-product contingencies (and does this make them more loyal customers?).

Branding the brain: A critical review and outlook - ScienceDirect
The Plassmann-Ramsøy-Milosavljevic model of customers’ brand preference learning.

It’s not noise, it’s your model that is wrong

With a single publication, Kahneman and his team are now throwing us at a new and exciting problem. And it is indeed interesting, since it’s such an overlooked concept and problem.

But let’s grab this rope where it has been left off, and not from the beginning. We know a bit more about what this “noise” thing is. We can already explain a good chunk of the otherwise unfathomable variance we see in our data. There’s a meaning to the dissonance in our data.

As any good statistician will tell you, the more things you know, the more you can use to correct your model. A corrected model is a better model. Crucially, what we otherwise see as noise becomes explanable.

So what is all the fuzz about? I know, it’s about explainable variance 🙂