Four Ways To Make AI Work For Sustainability

There are high expectations that AI can help solve complex issues around sustainability and climate but we must beware the pitfalls.

Enlisting artificial intelligence to help solve complex issues surrounding sustainability and climate change is creating plenty of hype.

Governments and organisations are creating AI strategies, pushing for increased AI adoption, and setting high expectations for what AI can do. But recent research shows that most expectations to date around AI aren’t being met.

There is a need to better understand the realistic potential of AI in addressing sustainability challenges.

Our research suggests four ways to bridge these gaps between expectations and reality at the AI-policy interface.

While there is immense hope in the promise of AI for addressing complex sustainability challenges, this research shows there’s a significant gap between expectations and real-world applications.

AI is not yet at the technological tipping point of being able to solve our sustainable development policy challenges and the multitude of human factors at play need to be identified and addressed as much as the technological potential.

A key issue is the variation in understanding AI’s potential, compared to where the technology is at currently.

This has led to a wider discussion about the possible applications of AI. As a result there is a disconnect between what AI can do now and social and environmental needs.

Academic researchers in AI are highly optimistic about its promise in addressing sustainability policy challenges but are often not across the intricacies of policy development.

This means they may not be able to understand if and how AI technologies could support decision-making.

Many also recognise the myriad ways AI could exacerbate sustainability issues if not handled with due consideration.

Governments worldwide are creating national AI strategies and policy frameworks. The motivation for doing so comes in part because they see AI as beneficial in decision-making and for addressing complex public policy issues, such as sustainability.

But it also stems from a fear of being left behind in the next “space race”.

Despite high expectations for government use of AI, there is very limited evidence of actual implementation. Related use cases that have made it into the public sphere are often entrenched in scandal.

For instance the (mis)use of algorithms to detect and (falsely) accuse people of welfare fraud in Australia (Robodebt) and the Netherlands (SyRI). Or the widespread use of Facial Recognition Technologies (FRT) in policing and security contexts across the United States and the UK, leading to accusations of bias and racialisation, as well as other rights abuses.

Consultants and think tanks play a major role in shaping the narrative around the promise of AI, and push for accelerating adoption to avoid getting stuck in “pilot purgatory” with small scale use that doesn’t result in impact.

However, while “scaling up” could result in greater benefits and positive outcomes, real world policy problems are complicated and ensuring a tool “works” across multiple contexts takes time.

There are four practical recommendations for translating AI’s promise into practice:

Document and Evaluate: It’s crucial to document and rigorously evaluate AI applications in real-world settings to understand their true impact and effectiveness.
Focus on Mature Technologies: Prioritise AI tools that are proven and reliable rather than speculative, experimental technologies.
Problem-Centric Approach: Start by clearly defining the policy problem, then select the most suitable AI technology to address it, rather than assuming AI is always the answer.
Adaptability to Complexity: AI solutions must be flexible and continuously evaluated to handle the dynamic and multifaceted nature of sustainability issues.

The journey from AI’s promise to practical implementation is still unfolding. Bridging the gap between AI’s potential and practical use in policymaking is crucial for harnessing its benefits while mitigating risks.

By focusing on real-world evidence and thoughtful application, we can better harness that potential without falling for the hype.

(Published under Creative Commons from 360info)