Language and Mental Models in Trigger-Action Programming

Allowing individuals to customize and manage the functions of their smart devices is essential in various contexts, ranging from automating daily routines in smart home environments to customizing educational paths in school classes. End-User Development (EUD) through Trigger-Action Programming represents a promising approach to enabling users -- regardless of prior programming experience -- to achieve their goals defining automation. However, for non-programmers to effectively manage complex scenarios involving multiple variables and events, EUD solutions must support the assumption of appropriate reasoning strategies and mental models. Prior research suggests that specific linguistic cues can facilitate the representation of trigger-action dynamics, thereby enhancing users' ability to define the behavior of their devices. This thesis investigates the role of linguistic elements in guiding non-programmers in the definition of trigger-action rules for programming their smart devices, supporting users in assuming effective mental models and reasoning strategies. Through three studies on language, this work highlights design guidelines for developing systems that better support non-programmers in defining trigger-action rules. Moreover, two systems based on Trigger-Action Programming leveraging two different compositional approaches -- form-based and LLM-based -- are explored to examine users' interaction with the system and the impact of language and mental models in real-world applications