On July 2, DigiShape organized a working session within the Water and Climate working group of the Dutch AI Coalition. The goal: to explore how AI can contribute to climate challenges. One of the concepts discussed was ‘AI for flood disaster management’, a spontaneous idea by Christine Bel that needs further elaboration. We spoke to her about it.
Christine, what is your position?
“As Innovation Manager AI at TU Delft, I connect science with practice so that our research contributes concretely to helping solve societal problems. The AI for Water and Climate session was an opportunity to explore how AI can provide solutions to challenges around water and climate.”
Can you talk more about your idea of using AI for flooding?
“There is plenty of research being done on predicting floods with AI, including by TU Delft. But what do you do with that information? What measures do you take in case of a predicted or ongoing flood? Besides technical solutions, such as dikes or sandbags, managerial aspects play a role. For example, what do you do if everyone tries to get out of a city by car? Will your assessment protocols still work then? All sorts of procedures, communication, politics and unpredictable human behavior come into play, and AI may be able to help integrate these complex factors into decision-making.”
Do you have an example of how AI can help in crisis situations?
Laughs: ‘To be honest, I’m mainly basing it on what I heard at the meeting and on Netflix, because I’m not a water expert. In the series ‘Wielka Woda’, based on a true story, those involved have to decide whether to breach a dyke to save a town, at the expense of a village. No one dared to take that responsibility. I won’t reveal what happened in the end, but I think AI can quickly calculate scenarios in crisis situations and provide options, so you don’t end up in endless political discussions.’
What do you think are challenges in applying AI in this context?
‘The first challenge is the availability of data. We need reliable data on floods, infrastructure and human responses. In addition, ethics are important. AI can calculate scenarios, but that does not mean those scenarios are also administratively feasible or desirable. Think of the surcharge affair: there, insufficient thought was given to the undesirable consequences of deploying algorithms. With floods, for instance, you have to weigh up: intervene actively now with certain consequences or risk worse consequences? This has to be embedded in a system that also includes these kinds of ethical and managerial considerations. And the final decision always lies with humans.’
What could be a next step for developing this idea further?
‘Finding scientists and researchers interested in this topic and exploring the possibilities for funding. In doing so, it is obviously important to work with partners such as the Directorate-General for Public Works and Water Management, water boards and municipalities.’
How do you look back on your role in the working session?
‘It was a valuable experience to talk to experts and share a spontaneous idea. My role is mainly to make connections and see how we can move forward with such ideas. If any scientists are interested, I can put them in touch with potentially interested TU researchers.
