Agile science: creating useful products for behavior change in the real world

TLDR

Evidence‑based practice is crucial for behavioral interventions, yet how to best support real‑world behavior change remains debated. This paper defines a set of products and a preliminary process for efficiently and adaptively creating and curating a knowledge base for real‑world behavior change. Drawing parallels to software development, the authors propose an agile science process with generation and evaluation phases that produce and assess operational definitions of three products—self‑contained modules, computational models, and personalization algorithms—while emphasizing early and frequent sharing. If successful, agile science could yield a more robust knowledge base for behavior change.

Abstract

Evidence-based practice is important for behavioral interventions but there is debate on how best to support real-world behavior change. The purpose of this paper is to define products and a preliminary process for efficiently and adaptively creating and curating a knowledge base for behavior change for real-world implementation. We look to evidence-based practice suggestions and draw parallels to software development. We argue to target three products: (1) the smallest, meaningful, self-contained, and repurposable behavior change modules of an intervention; (2) "computational models" that define the interaction between modules, individuals, and context; and (3) "personalization" algorithms, which are decision rules for intervention adaptation. The "agile science" process includes a generation phase whereby contender operational definitions and constructs of the three products are created and assessed for feasibility and an evaluation phase, whereby effect size estimates/casual inferences are created. The process emphasizes early-and-often sharing. If correct, agile science could enable a more robust knowledge base for behavior change.