I have long considered all my co-editors of this special issue to be good friends. That is, until they asked me to write an editorial on the topic of “what is next?” When a bunch of experts in judgment and decision-making ask you to do something they know to be impossible, you should be suspicious, right? Do they think I don’t know that predicting the future of science is impossible?
They slyly assigned Katy Milkman the job of luring me into agreeing. The first request came via email with what had to be a deliberately impenetrable subject heading: “Ask for OBHDP Special Issue You're Co-Editing: 1–3 Paragraphs on the Future of Nudge.” The other three co-editors were copied, the message was long and complicated, and, to top it off, the first word of the subject was “Ask.” Katy surely knew there was no chance I would read that email, which of course was part of her cunning strategy. She figured that when she sent the inevitable follow-up email I would feel guilty about not responding to the first one. Guilt is a powerful nudge.
The expected second email came three days later, this time with a catchier one-word subject line: “Noodge.” (Have I mentioned that these emails arrived in the early days of the COVID-19 lockdown?) This new email began by acknowledging that the first one had been too long and poorly timed, lulling me into a false sense of security that I was being excused and off the hook. But then, Katy launched the heavy artillery. She framed her request in a way that made my acceptance the default option: “Hope you’re up for writing 1–3 paragraphs, but let me know if not and we’ll manage. :)” We all know that defaults are powerful, but did she really think this was going to work on me? Although I was mildly miffed at the brazen noodging, I find it hard to say "no" to Katy, so I stuck to my usual strategy of lying low and ignored this email as well, foolishly hoping she would give up.
That hope was dashed a week later when the third email arrived with the subject line: “pretty please with sugar on top. :)” Plus, she pulled out another trick she had up her sleeve: a deadline! “The introduction is due in just a few days!” She was telling me that this assignment, which I had never agreed to do, was almost overdue. Of course, she also knew I was trapped in my home with very few excuses. Seeing no plausible escape route at this point, I capitulated and agreed to her request.
Conclusion: nudging works! Even on me.
Recall her request was that I write one to three paragraphs. This is already the sixth paragraph so by all rights I should already be done. Certainly, I will not be lured into making any forecasts. Phil Tetlock is her colleague! But since the word processor is already open, I will instead offer a few thoughts about my hopes and dreams for this enterprise.
My first hope is that the range of “nudges” expands. We know a lot about the effect of the kinds of strategies Katy used in her emails to me such as defaults, reminders, deadlines, guilt, salience, and norms. Come to think of it, I am surprised Katy didn’t try “90 percent of all recipients of my emails agree to do what I ask.” While I concede that these ploys often (though not always) work, it can't be that they span the entire behavioral science repertoire. So I am hoping to see studies using a different set of behavioral insights. I am sure there are good ones out there.
A different problem I see lies in the difficuly of distinguishing the efficacy of behavioral interventions from the degree of difficulty of the task being undertaken. Often, behavioral scientists are asked to help change a particular behavior but are severely limited in the ways they can alter the underlying environment. Rarely, if ever, are researchers given the opportunity to design the entire choice architecture. We get to remodel the kitchen, but not design the entire home, let alone pick the lot on which it is built.
Consider, for example, the interesting paper in this volume by Barnes et al. The authors intelligently and diligently investigate various ways to help people make better choices among health care plans in the Obamacare marketplace. But naturally they must work within the existing structure.
What if instead a group of behavioral scientists were able to design such a system from scratch? They could solve the problem of poor decision making by simply eliminating all choices and offering a single-payer plan, but let's assume that is not an option. Instead, let's stipulate that the researchers must design a system to do the original job of the Affordable Care Act (ACA), which was to fill in the gaps of our messy healthcare system. If that is the job, then there are many aspects of the ACA choice architecture that might be improved beyond what Barnes et al. had to play with — especially if we could provide the U.S. government with a well-functioning information technology system.
For example, if you were designing this system from scratch, would you tie the subsidy a household is given for 2021 healthcare needs to the income they earned in 2019? Probably not. You would sensibly want to tie the healthcare subsidy to current income rather than earnings from 11 to 23 months ago. But to do this, you would need permission to renovate the existing infrastructure. Although one part of the government receives up-to-date payroll information from employers, the only data available to the ACA administrators is the previous year’s income tax return. Ideal choice architecture in this case would come close to a full gut rehab, especially of the information processing capabilities. (Hint: it would not require COBOL programmers.)
Another aspect of the choice architecture that you might want to consider is whether and how to group plans into categories. But once again, as a researcher, you must deal with the fact that these choices were made long ago, and a bit oddly. Plans are grouped primarily by the total out-of-pocket costs policymakers might have to pay. Strangely, the names of the categories are metals (like Olympic medals): Bronze, Silver, and Gold, with an additional super category Platinum. Furthermore, very high deductible plans do not get assigned a metal (or a medal!); instead they are called catastrophic plans. Why these names were picked and whether the names matter is not something you or any other researcher can now manipulate.
My basic point here is that behavioral science researchers are almost always trying to nudge in the context of complex systems in which they can at best tweak behavior at the margin. It might be more productive to shift from creating nudges to reducing sludge, that is, eliminating the barriers that make otherwise good decisions difficult. Or even more ambitiously, identifying projects where behavioral scientists can be involved at the very start, helping to create the blueprints of a program before ground has ever been broken.
So dear friends and co-authors, rather than answer your entirely impossible question, I am instead offering a wish: I hope the future involves embracing choice architecture in the fullest sense of the word, moving from the questions of interior design to the tougher, upstream questions of plumbing, carpentry, masonry, electricity, zoning...to really see what responsible choice architecture can do for the world.