|Institution:||University of Saskatchewan|
|Keywords:||Human Behaviour; Data Collection; Architecture; Smartphone; Android; Pipeline; Programming Language; Wireless Sensor Network; Contextual Survey; English-like Language; Ecological Momentary Assessment; Participant Tracking; Spatio-temporal Behaviour|
|Full text PDF:||http://hdl.handle.net/10388/ETD-2014-12-1851|
Understanding human behaviour is key to understanding the spread of epidemics, habit dispersion, and the efficacy of health interventions. Investigation into the patterns of and drivers for human behaviour has often been facilitated by paper tools such as surveys, journals, and diaries. These tools have drawbacks in that they can be forgotten, go unfilled, and depend on often unreliable human memories. Researcher-driven data collection mechanisms, such as interviews and direct observation, alleviate some of these problems while introducing others, such as bias and observer effects. In response to this, technological means such as special-purpose data collection hardware, wireless sensor networks, and apps for smart devices have been built to collect behavioural data. These technologies further reduce the problems experienced by more traditional behavioural research tools, but often experience problems of reliability, generality, extensibility, and ease of configuration. This document details the construction of a smartphone-based app designed to collect data on human behaviour such that the difficulties of traditional tools are alleviated while still addressing the problems faced by modern supplemental technology. I describe the app's main data collection engine and its construction, architecture, reliability, generality, and extensibility, as well as the programming language developed to configure it and its feature set. To demonstrate the utility of the tool and its configuration language, I describe how they have been used to collect data in the field. Specifically, eleven case studies are presented in which the tool's architecture, flexibility, generality, extensibility, modularity, and ease of configuration have been exploited to facilitate a variety of behavioural monitoring endeavours. I further explain how the engine performs data collection, the major abstractions it employs, how its design and the development techniques used ensure ongoing reliability, and how the engine and its configuration language could be extended in the future to facilitate a greater range of experiments that require behavioural data to be collected. Finally, features and modules of the engine's encompassing system, iEpi, are presented that have not otherwise been documented to give the reader an understanding of where the work fits into the larger data collection and processing endeavour that spawned it.