18 May 2016

Unlearning in Behavioral Science

Recently I came across various materials* which state that many of the things I learned in grad-school at Erasmus University aren’t as I’ve learned them. Among them are the bodies of knowledge on choice overload, priming, ego-depletion etc. (Just as a note, Erasmus Research Institute of Management is in top 3 European research institutes in its field, so I got top level education from highly qualified professors.)

* This piece by Jason Collins I found particularly interesting. 

Nowadays, there is serious doubt casted upon ego-depletion, studies on priming have been challenged on a wide scale, apparently, choice overload doesn’t manifest each and every time and, it seems that, the endowment effect doesn’t manifest in isolated groups of hunter-gatherers, which means that it is not an innate (evolved) human feature.

It looks like I have to unlearn many of the things I have learned.

That’s interesting.

Yet, I can’t simply delete from memory knowledge (information) on ego-depletion (just as an example). Even if I could delete information from my memory, in the case of priming there is no clear list or criterion on which results hold and which don’t.

The situation isn’t as grim as it seems at first glance, particularly for the applied part of behavioral science.

The fact that not every result found in a scientific study (lab or field) doesn’t hold in different contexts is not exactly news. Moreover, people who profoundly understand behavioral science know that there are some strong and some weak phenomena. For example, we know that loss aversion and mental accounting are strong phenomena and we also know / knew that priming is a weak one.

When doing applied behavioral science work, you want to rely on the strong phenomena. Sure, you can use weaker ones such as priming, but you can’t rely on them.

For example, when designing an intervention intended to keep a space clean, it is essential to rely on convenience of trash-cans and social proof (social imitation – more broadly). Yes, you can dispense a citrus scent (olfactory priming) to promote cleanness, but that is more of an add-on and not the core of the intervention.

We knew that priming studies aren’t exactly at the top of reliability and replicability, but most behavioral science specialists have learned about ego-depletion and many (most?) took it as a given.

Apparently it is not.

However, this doesn’t bring huge changes to applied behavioral science work. The most important practical learning from the entire body of knowledge on self-control and ego-depletion is:

Design is more powerful than self-control

And it holds true.

Maybe the self-control issues that we face are not due to ego-depletion; they could be due to fatigue, forgetfulness, lack of (availability of) attention etc. Yet, how we solve them remains unchanged.

The fact that choice overload doesn’t manifest every time and failures of choice architecture aren’t exactly news, either.

Choice architecture works ONLY when there is no clear pre-existing preference (i.e. the chooser faces ambiguity).

I have been saying this in my training on choice architecture since 2013 (when I joined the field).

If in shop A (men) shoes come in sizes 40-45 and in shop B they come in sizes 41-49, I will buy 43 in either shop, simply because that is my shoe size.
For more on this topic read To Be Clear on Ambiguity 
Fewer options to choose from is simply easier for the consumer (chooser), but choice is about a lot more than being simple to choose.

Take the example of Total Wine shop(s). This is what they say on their “about” page:

Our typical store carries more than 8,000 different wines from every wine-producing region in the world, including more than 2,000 wines not available in any other store. (source)

Through the lenses of choice overload 8000 different wines sounds crazy. According to the choice overload principle they should have went out of business a long time ago. Here are a few explanations why they’re still doing well.

First, when wanting to be “the place to go to for wines”, you need to have lots of wines.

Second, many wine enthusiasts seek variety and want to explore. (Others simply want to get drunk and probably have an existing preference for the wine with the best alcohol/ price ratio).

Third, a wine bottle is a low-impact purchase. If you buy something you don’t like it’s just a few (more) dollars spent for an interesting* experience.

*interesting is a word used when you don’t like something and either you don’t admit it or you don’t want to say it out-loud.

For full fairness, Total wine uses choice architecture and offers a lot of structured choice.

Just as a rhetorical question:

Why do we put choices about houses, retirement plans in the same category with ones about jams, wine and coffee?

This, however, is another story.

Coming back to the applied part of behavioral science: We have to test if our interventions work or not, thus we can assess in each particular case how the number of choices offered influences the outcome.

When doing applied behavioral science we have to take into account the bigger picture.

For example, if in a coffee shop reframes the “bring your own cup” discount into a surcharge for paper-cups behavior shifts towards the desired direction: fewer paper cups used. However, it may also lead to a decrease in clients who simply don’t like the idea of paying explicitly for paper cups, even if the total price is the same as at the coffee shop across the street where the paper cup is included.

Moreover, clients who intended, but forgot to bring their own cup will feel bad (due to loss aversion) for having to pay for a paper cup. Moreover, they will experience regret, which is one of those emotions you don’t want to have associated with your business.   

One of the beauties of science is that it evolves and if we are to be professionals in a field of (applied) science we need to keep up with it and, occasionally, unlearn things.