Simulating Participant Behavior in Experience Sampling Method Research

The Experience Sampling Method (ESM) is applied widely for collecting self-reports from participants in free-living environments. Preserving high compliance in ESM remains challenging, especially when a study lasts more than a few weeks. Markedly, participants get increasingly bothered by prompts delivered at inconvenient moments. To alleviate that, personalization techniques have shown their potential. Particularly, ESM protocols that delivered prompts at more convenient times have significantly fewer drop-outs. Such personalization may lead to sampling bias, while ESM should be ecologically valid. Therefore, it is critical to equip experimenters with tools that enable trade-off analyses between the minimization of dropout versus the maximization of ecological validity. This paper lays the foundations for such analyses: we propose a novel ESM-specific participant behavior simulator, demonstrate its resemblance to real-life data and expected behaviors indicated by psychological theories. Such simulators enable trade-off analyses and they can help avoid the cold start of reinforcement learning agents.