Big data psychologyTools Lavelle-Hill, Rosa (2020) Big data psychology. PhD thesis, University of Nottingham.
AbstractThis thesis investigates what big data can add to the psychological study of human behaviour; and how Psychological theory can inform developments in machine learning models predicting human behaviour. It works through the difficulties that arise when the fields of machine learning and psychology meet. While machine learning models deal well with big datasets, they are designed for prediction, neglecting psychologists' desire to, not just predict, but understand behaviour. Psychology does well at using theory to specify models and explain the variance within a sample, yet can fail to consider how transferable the findings are to new samples. This research harnesses over a million loyalty card transaction records from a high-street health and beauty retailer linked to 12,968 questionnaire responses measuring demographics, shopping motivations, and individual differences. Equipped with real world behavioural records, and information on potential psychological and demographic drivers of behaviour, this thesis explores the ways in which psychological research can be undergone using big data to better understand three main areas: well-being, environmental behaviours, and anxiety symptoms. This thesis has the goal of marrying the strengths of traditional psychological methodology (utilising theoretical knowledge, quantifying uncertainty, and building interpretable models) with the exciting possibilities afforded by big data, all whilst ensuring that the models are generalisable and do not overfit. The following chapters discuss and evaluate novel research in this space, as well as the difficulties encountered, and compromises made, in undertaking `Big Data Psychology’.
Actions (Archive Staff Only)
|