Semi-Automated, Large-Scale Evaluation of Public Displays

This article presents a scalable, semi-automated process for studying the usage of public displays. The process consists of gathering anonymous interaction and skeletal data of passersby during public display deployment and programmatically analyzing the data. This article demonstrates the use of the process with the analysis of the Information Wall, a gesture-controlled public information display. Information Wall was deployed in a university campus for one year and collected an extensive data set of more than 100 000 passersby. The main benefits of the process include (1) gathering of large data sets without considerable use of resources, (2) fast, semi-automated data analysis, and (3) applicability to studying the effects of long-term public display deployments. In analyzing the usage and passersby data of the Information Wall in our validation study, the main findings uncovered using the method were (i) most users were first-time users exploring the system, and not many returned to use the system again, and (ii) many users were accompanied by passive users who observed interaction from further away, which could suggest a case of multi-user interaction blindness. In the past, logged data has mainly been used as a supporting method for in situ observations and interviews, and its use has required a considerable amount of manual work. In this article, we argue that logged data analysis can be automated to complement other methods, particularly in the evaluation of long-term deployments.