External validity and assignment of experimental vs. control treatment providers within small work groups: a research note

Abstract

Abstract Objectives When offenders or victims are randomly assigned to receive experimental vs. current treatments, the external validity of results may depend on whether different treatments are delivered by similar kinds of treatment providers. When treatment providers volunteer to deliver innovative practices in an experiment, it is unclear whether outcomes depend on the content of the treatment, enthusiasm of the providers for the new practice, or both. In such situations, the potential for what we describe as differential predisposition of volunteers for a new treatment raises a question of external validity. Methods We describe the process by which 14 out of 29 mediators across seven Danish police districts came to deliver a new, restorative conferencing method of conducting face-to-face meetings between offenders and their victims, in comparison to longstanding mediation methods. Results We negotiated with all seven District mediation leaders and all 29 of their mediators to use partial random assignment of 14 of the mediators to deliver the new, restorative model. The 14 trained providers of the new method were substantially similar in several measureable characteristics to the 15 other mediators who continued to use the preexisting model, but we cannot measure directly the extent or balance of their predispositions for delivering each model. Conclusions While small work teams pose obstacles to simple random assignment of treatment providers to deliver experimental practices, the random assignment of victims and offenders to two different models of service might be made more externally valid by use of partial random assignment of service providers.