30 minutes with...the human geographer using AI to find out how we feel about rivers

CIWEM’s digital-content editor Miriam Habtesellasie speaks to Dr Helge Peters, researcher at Oxford University’s School of Geography and the Environment.

Can you tell us a little about your professional background?

I'm a human geographer, which means that I study how people relate to the natural environment. I started doctoral studies at the University of Oxford in 2014 and completed my PhD there in 2019. I have continued to work as a researcher at the School of Geography and the Environment at Oxford since then.

My current work is focused on CAMELLIA, which stands for Community Water Management for a Liveable London. It’s a large project involving Oxford and Imperial College London, UCL, the British Geological Survey and many partners including the rivers charity Thames21.

Can you briefly talk about background behind the river sentiment app?

My work with CAMELLIA is centred around developing participatory methods, such as community modelling, where we try to give community members a say in decisions about blue-green infrastructure, decisions that would otherwise be left to experts alone.

We're facilitating this by bringing water professionals and scientists together with residents, using tools such as participatory computer modelling to make discussions on the future of water infrastructure more inclusive.

As part of this I interviewed residents across four London boroughs to find out what issues they were facing in relation to the water environment, from flooding through to water quality.

One key finding while completing those interviews is that though there is a lot of public support for water quality measures such as SuDS, it’s difficult to secure funding for them. I was told that this is partly because water quality is not valued very highly in economic appraisal models, despite the obvious benefits to the local community. The two were seemingly at odds.

At the same time during the 2020 lockdown people started paying more attention to their local water environment and consequently we saw a surge in the number of people talking about water issues, such as sewage pollution on Twitter.

So, I decided to bring these two observations together. With the support of a senior colleague, and having won a small grant, I teamed up with Nathanael Sheehan, a brilliant young computer scientist, and we began building the river sentiment app.

Can you explain how the river sentiment app works?

The river sentiment app is basically an online map that lets you look up a river in the Thames basin and find out how healthy it is and how it makes people feel.

The app maps a sentiment polarity score on to more than 400 rivers within the Thames river basin. This score is based on a sentiment analysis of more than 3 million tweets, mined by accessing Twitter’s API, that mention any water body in the Thames river basin by name.

So, if you tweet and you mentioned, say, the river Cherwell in Oxford, then it'll likely be picked up by our algorithm. And then that tweet becomes part of our data set of these millions of tweets.

We then perform a sentiment analysis on this data set. Sentiment analysis is part of a suite of technologies under the umbrella term emotional AI, which is about algorithms that sense how people feel.

In our case we are using an algorithm that calculates the sentiment polarity of your tweet by labelling the language it contains as expressing negative emotions (red), neutral emotions (yellow) or positive emotions (green). This traffic light system is then layered onto a thematic map of the rivers in the Thames basin, and we also include the rivers’ ecological status as defined by the Environment Agency (EA).

So, users can look up any river in the area that they're interested in, either by zooming in on the map or using the search function and find out about the river’s ecological status and how it makes people feel.

Who do you hope the end users of the app will be? And what do you hope they will do with the insights?

I hope the app travels far and wide, and that lots of people start using it! Thames21 have been brilliant at helping us understand how local river-action groups might use this technology.

So, for instance, if you're leading a local river-action group, you might use this to demonstrate to your stakeholders how sentiment about a river has changed over time, to become more positive or negative. The latter is a feature we’re currently building.

Another possible use is to utilise the app to identify those rivers that currently have very little engagement on social media and investigate what could be done to better connect residents with their local environment.

Did your research findings make your feel more/less hopeful about our ability to tackle river-quality issues?

I do believe in the power of collective action to drive change towards a more sustainable approach to the environment.

It’s quite heartening to see through our research how communities are getting together to ask the EA, councillors, members of parliament, water companies, really everyone in the water sector, which probably includes water professionals more broadly, to do better.

So, I'm actually quite hopeful that eventually rivers will get better as more people care for them and use tools such as social media, among a whole suite of tools, to hold polluters to account and keep regulators on their toes.

How do you predict AI will support policymaking and mobilise action on water quality and rivers?

That's a big question. There are many different applications of AI, and it's a whole suite of technologies under one umbrella term. What I can say is that I'm pretty convinced that we can, in principle, harness AI tools for positive change.

That’s to say, if, and that's a big if, we bring the concerns of affected communities to bear on the development of these technologies. That is what I would like to see more of in the field. Rather than promising salvation through technology, AI researchers should bring affected communities into the development of these technologies through co-design and co-creation methods.

What’s next for the app?

We plan to officially launch the app in March. For now, you can view and test the prototype here http://thames21ox.web.app. This is a beta version and we encourage everyone to get in touch and report any bugs and issues that they find.

The app is static at the moment, which means users can’t put in additional data, but we hope to make it more dynamic, and hopefully have a function that will continually feed in Twitter data. If we can find the resources to support its continued development, we also plan to have a wider geographical range and include additional features.

We would definitely love to hear from water and environmental professionals on how useful they found our app and what we can do to improve it!