The project

At the Turing, in collaboration with her colleagues Seresinhe, Alanyali and Preis, Dr. Moat is investigating to what extent we can exploit images from sources such as Flickr and Instagram to develop new indicators of human behaviour and our experience of the world we live in.

In a recent paper, Moat and colleagues used over 1.5 million votes from the online game Scenic-or-Not to train a deep learning model to rate the beauty of outdoor locations (Seresinhe, Preis and Moat, 2017, “Using deep learning to quantify the beauty of outdoor places”, published in Royal Society Open Science). When trialled on pictures of London, the model identified photos of Big Ben and the Tower of London as being amongst the top 5% scenic views.

These new results add to their earlier findings that people who live in areas rated as more scenic report their health to be better, even when data on greenspace is taken into account (Seresinhe, Preis and Moat, 2015, “Quantifying the impact of scenic environments on health”, published in Scientific Reports).

In particular, the study helps answer questions around what makes a location scenic, for example suggesting that flat, featureless green spaces are not considered attractive, whereas man-made elements such as historic architecture and bridges can boost the aesthetics of a scene.

In parallel, Dr. Moat is developing Turing projects with a number of policy stakeholders. Moat recently led the organisation of an ‘Urban Analytics Data Dive’ at the Turing, hosted in collaboration with the Office for National Statistics Data Science Campus. The event saw organisations such as the Cabinet Office’s Policy Lab and the Department for Communities and Local Government pose urban policy challenges for 50 data scientists, who set out to solve them over two days, using data from sources such as Strava, Zoopla, and satellites.

Research interest

Our everyday usage of the Internet leaves huge volumes of text and images in its wake. Dr Moat’s research investigates whether we can use online data from sources such as Google, Wikipedia and Twitter to measure and predict human behaviour in the real world.

Her previous work has touched on problems as diverse as linking online behaviour to stock market moves (with Preis, Curme, Stanley, et al.), estimating crowd sizes (with Botta and Preis), and evaluating whether the beauty of the environment we live in might affect our health (with Seresinhe and Preis).