I had never made a rigorous distinction between role-playing and improvisation before, and they do share many traits. The authors of this article focused on theatrical improv, and game-style role playing, because these approaches are the most useful to design work.
Improv focuses on giving the actors a very barebones setting, and the actors make a story up in real-time. There is no reversing the action of the scene, and there are no corrections. Anything that goes awry must be acted upon and worked into the story. As the story unfolds, the actors develop a cognitive consensus about the scene, even though they had no idea what the scene would be before starting the performance.
Role-playing, while also unscripted, focuses on each actor playing a certain roughly-predetermined character in a scene. Game or entertainment-based role-playing is different from psychological role-playing in that it focuses on telling a story, rather than conditioning the role-player into certain behaviors.
Role-playing seems to me to be especially useful in figuring out what a certain type of user would do with a user interface or software package. On an informal level, working out how I’d accomplish a task if I didn’t have the knowledge to throw together a simple Python script is quite useful. On a more formal situation, role-playing with multiple participants could offer insights on how the thing being designed fits into the group and social structure of a situation.
The overarching question the authors asked was how useful and practical are improv and role-playing in the design process. Improv seems to be hard to constrain in a way that is useful for information gathering, but role-playing thrives on constraints and can answer more specific questions. Improv seems to really shine in situations where truly novel ideas is more important.
Role-playing reminded me a lot of the testing system engineers do toward the end of spacecraft-based projects at NASA. Once the system is built and approaching readiness, the operations team will all show up and attempt to use the system in a simulated launch or other critical operation. The system is fed simulated data from the spacecraft, and the team acts out their parts as if the data is real. At some point during the exercise, an operator will get ‘carded’ meaning that somebody walks up to them and lays a card on their keyboard that may say something like “the spacecraft has gone into safe mode” meaning its computer is mostly shut down and the satellite is tumbling in space. The team needs to react to this situation, and it’s considered a good test of how well the team is working together.