Collecting Insights into How Novice Programmers Naturally Express Programs for Robots

End-user programming of robots holds promise to help users lacking specialized software development skills customize robot behaviors. To enable such novice control, it is essential to design an end-user robot programming system that best supports how users naturally express what they want the robots to do. In this paper, we conducted a two-part online user study with fifteen participants to analyze how novice programmers intuitively solve robot-related tasks (part 1) vs how they code robot behaviors using a static version of Cozmo Codelab, a visual block-based robot programming system (part 2). Our part 1 results show that novice programmers do not intuitively consider programming tasks using an imperative procedural programming paradigm, which is typical of current visual block-based robot programming systems. Additionally, we find that users who initially adopted an imperative procedural programming paradigm (in part 1) for task completion in part 2 wrote less accurate robot programs. The mismatch between natural tendencies and enforced coding structure has implications for the design of easily learned end-user robot programming interfaces and systems. It suggests that block-based visual imperative programming systems, though at present one of the most accessible ways to program robots, could benefit from new interfaces and targeted design enhancements.