Supporting the Design and Analysis of HCI Experiments

Alexander Eiselmayer

DOI: https://doi.org/10.1145/3334480.3375038

2020-04-23

Abstract:Researchers in HCI commonly use controlled experiments to evaluate artifacts and interaction techniques. However, experiment design and statistical analysis are complex tasks that are prone to errors, especially for novice researchers. In part, this is because researchers need to make numerous decisions about the design and analysis while it is hard to immediately anticipate their effect. In this dissertation, I aim to study how interactive systems that provide real-time feedback, enable direct manipulation, and facilitate exploration positively influence decision making and reproducibility. My previous work, Touchstone2, shows how researchers can benefit from comparing trade-offs among experimental designs. I contribute an empirical understanding of how researchers design and analyze experiments, as well as a set of tools that support researchers during that process.