The Dilemma of AI in Environmental Approvals: Lessons from Robodebt
Amid an ongoing conversation about environmental protection, a proposal from the Minerals Council of Australia to implement artificial intelligence (AI) for speeding up national environmental approvals has sparked heated debate. Conservationists warn that this strategy could replicate the failures seen in the Robodebt scandal, risking threatened species even further.
The Proposal: AI in Environmental Assessments
The Minerals Council of Australia has called on the government to invest $13 million to trial AI technologies, aiming to assist companies in preparing environmental applications and aiding federal decision-making. This initiative seeks to streamline processes that have historically been bogged down by lengthy assessments.
A Warning from Experts
Experts at the Biodiversity Council, comprising independent specialists from 11 universities, have expressed skepticism. They argue that while AI may benefit straightforward tasks, automating complex environmental assessments could mirror the Robodebt crisis—where flawed, automated decisions led to unjust outcomes for recipients of welfare. The biodiversity experts emphasize that this lack of transparency could exacerbate risks to endangered species.
Understanding “Robodebt”
The Robodebt program, operational between 2015 and 2019, erroneously accused over 400,000 Australians of welfare overpayments due to automated systems that lacked oversight. The fear is that a similar approach could be taken with AI in environmental assessments, endangering biodiversity without sufficient checks and balances.
Complex Legislation and AI Limitations
Lis Ashby, a policy and innovation lead at the Biodiversity Council, points to inherent flaws in the country’s primary environmental law—the Environment Protection and Biodiversity Conservation Act (EPBC Act). She argues that its vague language complicates assessment processes, which could be further exacerbated by AI’s inability to navigate such ambiguities effectively.
More Read
“Current regulations impede rules-based decision-making,” Ashby notes, illustrating a fundamental issue where AI may not efficiently solve the existing problems but instead complicate them.
Technological Optimism or Caution?
Brendan Sydes, national biodiversity policy adviser at the Australian Conservation Foundation, acknowledges the potential role of technology but remains cautious. He believes that while AI could optimize the enforcement of nature protection laws, it should not dominate the decision-making process. Instead of rushing towards AI solutions, he advocates for more comprehensive data collection on threatened species and habitats.
The Data Dilemma
Prof. David Lindenmayer, a forest ecologist at Australian National University, emphasizes a critical barrier to AI-driven assessments: the lack of reliable data on Australia’s threatened species. He argues that AI decisions are only as credible as the data they utilize. Unfortunately, many species lack the necessary location and habitat data, which could lead to misguided and detrimental decisions.
“AI automation risks using outdated or inaccurate information, ultimately compromising biodiversity,” Lindenmayer warns.
Rethinking Approvals
Following a review in 2020 that indicated the EPBC Act was failing to protect species adequately, the Albanese government reformed environmental laws last year. Yet, experts like Prof. Hugh Possingham from the University of Queensland advocate for a human-centered approach. He suggests that employing more assessors rather than leaning on AI would yield better results in environmental protection.
Possingham highlights that training AI systems would require high-quality data, which is currently absent. The historical dataset corresponding to the EPBC Act approvals is not reliable, indicating that a focus on human expertise remains crucial.
Innovative Yet Controversial
Tania Constable, CEO of the Minerals Council, has defended the proposal, arguing that using AI is an innovative approach that could enhance environmental protections while improving efficiency. She believes it could assist both regulators and project proponents in navigating the complexities of environmental assessments.
A government spokesperson has reiterated that despite exploring the use of AI, project approvals will ultimately rest in human hands. The focus remains on making applications easier, with AI serving as a support tool rather than a decision-maker.
The Future Landscape of Environmental Protection
As discussions around the utilization of AI in environmental assessments continue, the importance of transparency, reliable data, and expert judgment cannot be overstated. While there may be potential for technology to play a role, the lessons learned from past failures must guide future implementations to ensure that both biodiversity and human welfare are protected effectively.
Inspired by: Source
