The scientists behind the study of scenic beauty

Suzy Moat and Tobias Preis are Professors of Behavioural Science and Directors of the Data Science Lab, Warwick Business School and Fellows of The Alan Turing Institute. Here they speak to us about the origins of the science of scenic beauty study.

Suzy Moat and Tobias Preis Suzy Moat and Tobias Preis

Are beautiful environments good for our health and happiness? In our Data Science Lab at Warwick Business School and at The Alan Turing Institute, we’ve been using AI and an online game to seek some scientific answers to this age-old question. And it turns out that canals and rivers have a very important role to play.

Millions of ratings of beautiful locations

In our Data Science Lab, we’ve spent years analysing large volumes of online data, telling us about what people are searching for on Google, or what photos people are posting online. We’ve used this data to get quicker insights into how diseases spread, or how well the economy is doing.

Back in 2014, with our colleague Chanuki Seresinhe, we started to wonder whether we’d spotted a way to use online photos to help us understand whether beautiful environments are good for our health and happiness too.

The key to answering this question came from an online game called Scenic-or-Not, which shows people photos and asks them to rate them from 1 to 10. From Scenic-or-Not, we had access to over 1.5 million ratings of over 200,000 locations all across Great Britain. Exactly the data we needed to crack this puzzle.

Cities and towns

By analysing Scenic-or-Not data along with data from the Census, we found that people who live in more scenic areas report their health to be better. Working with our colleague George MacKerron, who had developed a mobile app to gather data on the happiness of more than 15,000 people, we also found that people are happier in more scenic locations. Importantly, we found that these relationships hold in built-up areas such as cities and towns too.

We talked to planners about these results. To our delight, they were interested in what we’d found, but they wanted to know what makes a location more beautiful. From Scenic-or-Not, we had 200,000 photos that had all been rated – but we needed to find a way to work through all these photographs and figure out what had made people give certain photos higher ratings than others.

Artificial intelligence

We realised we could use a kind of AI called 'deep learning' to help us get data on what was in each of the 200,000 photographs.

Using this data, we found that beautiful locations are not only natural locations or green locations. Certain natural elements such as valleys, mountains and trees are of course associated with higher ratings. However, some man-made elements such as cottages, castles and bridges also help boost the score of a scene.

Crucially, in built-up urban locations, we found that the biggest boost to scenic scores was provided by the presence of a canal. Rivers also score highly.

We also found we could train some AI to rate photographs for itself, to help us produce even more data on how scenic different locations are.

What makes a canal particularly scenic?

From our previous work, we know that canals and rivers play a very important role in making built-up urban areas feel more beautiful. But not all canals and rivers look the same. What makes a canal or a river particularly scenic, and particularly beneficial to the people who walk past it?

Rate this Scene

We’ve worked with the Canal & River Trust to create a new game called Rate this Scene. Rate this Scene shows you pictures of canals and rivers all across England and Wales, and like Scenic-or-Not, asks you to rate them from 1 to 10. The game has tens of thousands of photographs that you can rate.

Rate this Scene will create an exciting new dataset on the beauty of the canals and rivers of England and Wales. The photos that you rate in this game have both locations and times associated with them, so that we can see how beauty varies across different locations and across time too.

This new dataset will help in three ways:

  • It will give the Canal & River Trust excellent data on the beauty of the canals and rivers they care for.
  • It will help us understand what makes certain canals and rivers more attractive, which in turn can help inform positive action.
  • It will also help us make our scenic AI even better at finding particularly beautiful canals and rivers in new photographs that haven’t been rated before.

Many decisions are made every year about building new structures, demolishing old structures, and maintaining the neighbourhoods we live in. Our findings provide evidence that the beauty of these environments may have consequences for our health and wellbeing.

This makes it crucial to ensure that we have excellent data on how beautiful different locations are, so that these decisions can be as well informed as possible.