Commitments to the future of Applied Behavioral Science

A few days ago, Michael Hallsworth published an academic article in Nature setting forth what he saw as the future direction for applied behavioral science. It was rightfully well-received as an overview but also criticized for its ironic lack of behavioral focus: it tended towards statements of principle over directly recommended actions.

That’s a reasonable critique, just as Michael’s response was reasonable: that providing an actionable guide would have been too unwieldy and beyond the scope of Nature. So as someone aligned with Michael in a desire to expand the field, I’m going to take his statements and go one step deeper — not by recommending actions but evaluating my own efforts to expand the field of applied behavioral science aligned against his principles. I’m not suggesting anyone should choose the same path as me and I think there are smart evolutions to be had, but if we’re going to center behaviors, here are the commitments I can be held accountable to.

1) Use behavioral science as a lens.

Discussion: To me, this lens is clear: centering behavior as outcome and using a science-informed method. In order to make behavioral science a lens at scale, it needs to be a process that anyone can learn and apply and I have written extensively about and continued to evolve the Intervention Design Process to that end. I have translated applied behavioral science into a clear 12-week sprint, with distinct phases, roles, and deliverables, and handoffs from business model (how we know what behaviors to change) to operational model (how we scale behavioral interventions). And we need more of those; fewer cognitive models (COM-B, B-MAT, competing pressures, etc.) and more process models that people can actually pick up and use.

Commitment: Continue to evolve the IDP as an applied process model.

2) Build behavioral science into organizations.

Discussion: I literally started a company that gathered together people who have done this at scale to teach other organizations how to evolve from their current processes to a behaviorally-informed organization (BeSci.io is Behavioral Science IN ORGANIZATIONS). There are many consultancies running trainings as an augment to their project work; we refer out project work to do only capability building.

Commitment: Make sure BeSci.io stays alive as an organization with its original mission intact.

3) See the system..

Discussion: This one may be a place where I need to change my behavior; either that or I disagree with the principle as expressed. I think focusing on tightly defined behaviors is actually fine for behavioral models — it is the job of the preceding step (business model, since I generally work in for-profit organizations) to coordinate the overall interaction of behaviors. My concession might be that by insisting on piloting in situ, we’re taking the system into account — if a behavior cannibalizes gains in another behavior, that is something we’re monitoring for.

Commitment: Enrich the “business model” process to compliment the “behavioral model” process.

4) Put RCTs in their place.

Discussion: This seems to be simply an emphasis on good, practical pilots, coupled with a strong dose of ethics. Michael uses the example of telling parents to talk to their kids about the science curriculum inadvertently causing them not to talk about other things that are also important. To me, this is back to the business model case: we need those who set out what behavioral statements we need interventions for to focus on the systemic importance and interrelation of those behaviors. I frequently draw this as a waterfall diagram of behavioral statements: we have an overall outcome we care about, then a series of other behaviors that lead to that outcome, with an infinite nest of sub-behaviors. By understanding the causal relationship between sub-behaviors and the behaviors they lever up to, we can balance across the system.

Commitment: Make a separate “behavioral science for the business model” training.

5) Replication, variation and adaptation.

Discussion: This entire section is best encapsulated by this sentence: “These challenges mean that applied behavioral scientists need to set a much higher bar for claiming that an effect holds true across many unspecified settings.” My method for avoiding this is simple: I don’t claim that anything I have designed generalizes beyond the context in which it was designed. At the beginning of the process, a behavioral statement is set that identifies population, limitations, motivation, behavior, and measurement — context identification is inherently part of the work. Every intervention is thus viewed as context dependent and while we can look to other work as a potential book of ideas for possible interventions, nobody should be claiming that something generalizes to any individual circumstance unless they have measured that individual circumstance and satisfied themselves with the evidence that makes it appealing for scale (we use an outcome statement for that).

Commitment: Continue to insist on behavioral statements before projects and outcome statements after projects.

6) Beyond lists of biases.

Discussion: Michael wrote “we are lacking good explanations for why findings vary so much”; I’m certainly not lacking. The explanation is actually identified earlier in his piece: the system is extraordinarily complicated and variable and depends on interactions between systemic pressures and the literal biology of each individual person. I don’t pursue universal theories because I believe that’s the domain of academic behavioral science; the whole point of applied behavioral science is that our outcome is behavior, not theory. I can imagine some future world where we are good at figuring out the generalizability of practices but the business sphere already has a way to do that: best practices, competition, and evolution. Put differently, I have no intention to contribute to any theory of generalized human behavior change as part of my applied behavioral science work (my academic work is clearly ringfenced), don’t believe other applied behavioral scientists should do that as part of their work, and instead think they should be working on the theory of applied behavioral science; that is, what generalizable practices get people to centralize behavior and use science-informed methods.

Commitment: To keep my applied and academic work separate.

7) Predict and adjust.

Discussion: Hindsight bias is absolutely a massive drag on the field of applied behavioral science. When stakeholders believe they “know” something without evidence and then view the process as an expensive way of “knowing what was already known,” then those processes are systematically defunded. This is where behavioral statements and outcome statements become so critical; because they explicitly establish both the measurement upfront and then create a throughline of evidence (this outcome came from intervention, this intervention changed this pressure, this pressure was identified through these quant and qual methods), they explicitly reject the “we knew” argument by codifying how an observation goes through experimentation to reach scale.

Commitment: To continue insisting on both behavioral and outcome statements.

8) Be humble, explore and enable.

Discussion: I reject humility as fundamentally valuable; it has long been the tool of the oppressor to subjugate and marginalize, because it relies on the fact that others will promote you if you don’t. Instead, we should strive to be accurate and limit our statements to what evidence supports; I am listing my current and future activities here simply because they are what I am doing and what I intended to do. And I think accuracy is what Michael is actually calling for, alongside diversity and democratization, which I think are deeply interlinked. To reach the highest form of applied behavioral science, we need to make it something that anyone can feel comfortable doing; in the same way we teach biology or chemistry or physics, we need to teach applied science and scaffold its use. That’s why I distribute so many free materials, teach a donate-what-you-can version of the class, and do public office hours and report on their diversity.

Commitment: Continue to use behavioral statements and outcome statements; create models that are increasingly comprehensive and accessible (driving down, not up); continue office hours and diversity reporting.

9) Data science for equity.

Discussion: The data science here is a red herring and I actually think there are two points here: outcomes and methods. The second is actually somewhat easier: assuming an outcome is ethical, it is relatively straightforward to use ethical review procedures to determine if the method of creating that outcome is ethical. I wrote an entire chapter about this in the book, teach it explicitly in our course material, etc. The more complicated one is outcome ethics and I actually think this isn’t about ethics but values. An individual behavioral outcome is not ethical or unethical, but rather equitable or non-equitable: they either widen gaps or close them. And that’s a value question. As someone who values lifting up those who are at the bottom of the pyramid, whether economic or otherwise, it is the explicit use of 100% of my non-economic applied behavioral activity. From GetRaised to IAskedHer to HowSureAreYou, literally every applied behavioral science intervention I create is aimed at increasing equity.

Commitment: Keep doing exactly what I’m doing: every non-economic project is explicitly equity aimed. Remind students that if they pursue outcomes that are contrary to my values, I will fight their work with mine.

9) No “view from nowhere”.

Discussion: This principle is, to me, one of the most important on the list. Earlier points addressed diversity in who is doing applied behavioral science and what we are doing it on, but focusing on our own position of privilege and the ways our culture tinges our intuition is just as important and indeed may be necessary in order to enable the other two. And because humans are notoriously bad at introspection (because our lens is built by the same environment that formed what we’re examining), I firmly believe that has to happen externally. That’s why I have an anonymous feedback form, a dedicated way of checking in with people I’ve worked with previously and public accountability around their ratings, and of course the diversity report for my outreach activities.

Commitment: Provide venues for anonymous feedback; explicitly seek out anonymous feedback and analyze it according to gender, ethnicity, etc. to monitor for gaps; track and publish transparent accountability metrics for outreach.