How can we find patterns in large amounts of data? How can we create optimal prediction models? How can we generate insights from political documents and speeches and social media? How do organisations need to change to institutionalise the use of big data in decision making? How can the data revolution help us address some of the grand challenges facing society? These questions can be answered by bringing together substantive expertise with knowledge of statistics and computer science. Developing long-term capacity in data science our department also hosts one of 15 national Q-Step Centres that provide training in advanced analytics for undergraduates in the social sciences.
Q-Step is a £19.5 million programme designed to promote a step-change in quantitative social science training in the UK. Funded by the Nuffield Foundation, ESRC and HEFCE, Q-Step was developed as a strategic response to the shortage of quantitatively-skilled social science graduates. Four academic departments contribute to UCL’s Q-Step Centre offering state of the art training in quantitative research methods, data analysis and visualization.
2. Natural Language Processing
Language used in social and professional communication can be processed and analysed in systematic ways ranging from classification to entity extraction and sentiment analysis.
Making textual data the subject of analysis allows us to learn about social and political processes and provides us with a distinct input that differs from more standard data such as attitudinal or behavioural measures. Current projects at UCL include e.g. sentiment analysis of political media reporting, mapping intra-cabinet politics from budget debates, using debates in the United Nations for foreign policy analysis and human rights monitoring, tracing policy diffusion over time in the network of MPs based on debates in the UK House of Commons, extracting information from events data in the media for conflict prediction.
Experiments (aka A/B testing) provide researchers with the ability to identify causal effects of treatments.
We are interested in the causal effect of interventions whether these are policies, programs, or institutions. The ultimate goal is to identify the causal impact of a specific treatment on a defined outcome. Current and past research projects have studied civic engagement encouragement and framing effects in austerity debates.
4. Predictive Modelling
Many areas of social life rely on accurate prediction of future events. We specialise in predicting political events.
Based on parametric, semi-parametric, or non-parametric models one can try to leverage data on past events to make probabilistic statements about future events. In our department we have several active research projects, e.g. predicting voter behaviour based on structural and individual information, predictive models of energy policy compliance based on patterns of consumer behaviour, and predictive models of violent conflicts and conflict dynamics.
- Project: Aid Attitudes Tracker (2013-2018)
Funder: Bill and Melinda Gates Foundation
Collaborators: Jennifer vanHeerde-Hudson (UCL), David Hudson (UCL), Joe Twyman (YouGov), Marianne Stewart (University of Texas, Dallas), Harold Clarke (University of Texas, Dallas), Will Tucker (Will Tucker Consulting)
- Predicting The Escalation Of Conflict
The key objective of this research project is to forecast conflict escalation of intra-state conflicts. In the past few years an increasing number of researchers are becoming interested in forecasting