The Need to Leverage Responsible AI to Transform Water Management
Paul Flemming & Blair Troutman — Published in Smart Water Magazine, 13/12/2023
The impacts of climate change, once viewed as a problem of the future, are increasingly challenges now. These challenges frequently emerge through water. Extreme events, be it floods or droughts, have devastating impacts that put immense stress on water systems and the communities and ecosystems that de- pend upon them. The need to embed resilience into water management, to make flexibility, redundancy and robustness inte- gral to the operations and capital planning of water systems, has never been greater. We need to responsibly leverage the power of artificial intelligence (AI) to address the complexity and uncertainty of water management under climate change.
When ChatGPT was released in 2022, the previously esoteric concept of AI was mainstreamed, unleashing a range of emotions from curiosity to existential angst. Addressing the role of AI in society requires an on-going debate and discussion in the realm of ethics, philosophy, and public policy. We must, however, continue to leverage and evolve the power of AI-guided decision-making regarding climate change.
One of the promises of AI is its ability to recognize patterns amongst massive amounts of data and general actionable insights to inform decisions. The staggering amounts of data in existence – which grows exponentially on a daily basis – make this functionality a necessity for any organization aspiring to make data-informed decisions. For water management purposes, however, the vast majority of this data is largely irrelevant. In this context salience is essential. This requires moving from a horizontal approach to AI tool development intended to address generic needs to a vertical approach, where AI is optimized for the water utility sector, trained on water utility-specific datasets and informed by water utility-specific characteristics and attributes, in order to generate salient intelligence.
This verticalization of AI for water is underway. Google’s Flood Hub and Upstream Tech’s Hydroforecast are just two of many examples of the deployment of AI to enhance water managers’ foresight capabilities. Earth observation (EO) data features prominently in many of the services currently in the market, but what role could the vast amount of utility gener- ated data play in shaping the salience of these services? What role could the utility sector play in shaping this water AI verti- cal to address utility-specific needs?
One, the sector should take an objective mapping approach to identify target utility business functions, the decisions made within those functions, and the data essential for each function. Second, utilities should develop, clean, and structure their data to train AI applications for utility specific applications. Third, these activities should be informed by a clear understanding of the analytical needs and objectives of decision makers so that AI tools are developed within and constrained by a clear decision-making context. What do I need to know, when and where?
An Industry 5.0 view of AI is essential, with AI augmenting and complementing human in- telligence and decision making, allowing humans to decide which decisions to delegate and which to retain. Other attributes that should undergird this work in- clude data accessibility, allowing data to be made widely avail- able not just within utilities but across the sector; interopera- bility, where data standards are developed to enable datasets that had been assembled and developed in isolation to be de- ployed as seamlessly as possible to generate derivative benefits; and transparency, where the understanding of AI results and use and are open and explainable.
Climate change is an “all hands on deck” moment for the water community. Leveraging the continually evolving power of AI is critical for building resilience in our water systems. Successfully managing the development and deployment of AI tools necessarily requires an active dialogue in our community to explore and address ethical issues associated with AI.