Concerns Arise Over AI Proposal for Environmental Approvals
Conservationists and scientists have expressed apprehensions regarding the mining industry’s initiative to implement artificial intelligence in expediting national environmental approvals. They fear this could lead to “robodebt-style” failures, thereby jeopardizing already threatened species.
The Minerals Council of Australia is advocating for a $13 million investment from the government to pilot the use of AI technology aimed at streamlining application processes and assisting federal decision-making.
However, the Biodiversity Council, which comprises independent experts from 11 universities, cautions that while AI may facilitate straightforward tasks, automating environmental assessments could result in catastrophic failures akin to the robodebt scheme—where flawed algorithms issued erroneous debt claims against welfare recipients, ultimately pushing some species closer to extinction.
Robodebt refers to an automated debt recovery initiative that, between 2015 and 2019, inaccurately accused a significant number of welfare recipients of receiving excess payments.
Liz Ashby, who leads policy and innovation at the Biodiversity Council, noted that Australia’s principal environmental legislation—the Environment Protection and Biodiversity Conservation Act—contains ambiguous language and offers broad ministerial discretion.
She articulated that “The prevalent vagueness exacerbates the current duration of assessment processes, obstructing rules-based decision-making by human evaluators. This lack of precision would pose even greater challenges for an AI-driven tool.”
“Establishing explicit regulations within the National Environmental Standards, particularly detailing unacceptable practices, would expedite assessment times even without AI assistance and is crucial for any prospective AI integration,” Ashby added.
Brendan Sydes, the national biodiversity policy adviser with the Australian Conservation Foundation, expressed skepticism toward the Minerals Council’s initiative.
He stated, “While technology holds promise in ensuring nature protection laws yield effective conservation outcomes, AI should serve as a tool, not a decision-maker.”
He recommended that the federal government concentrate on addressing existing data deficiencies related to threatened species and their habitats.
Professor David Lindenmayer, a forest ecologist at the Australian National University and a member of the Biodiversity Council, pointed out that research indicates approximately one-third of Australia’s threatened species lack sufficient monitoring, with others reflecting only sporadic data.
Assessors historically remedied these gaps by consulting field experts. “AI decisions rely on the quality of data available. Unfortunately, comprehensive data remains elusive for many threatened species across Australia—often lacking even basic geolocation information,” he remarked.
“Relying on AI could potentially result in decisions predicated on flawed or antiquated data, thereby failing to safeguard biodiversity.”
The Albanese government introduced reforms to environmental regulations last year after a 2020 review indicated these laws were inadequate in preserving species and habitats.
Professor Hugh Possingham, a prominent conservation biologist at the University of Queensland, added that AI tools necessitate a substantial dataset for training purposes.
The past two decades of EPBC Act approvals do not constitute appropriate training material, as evidenced by their failure to protect the environment.
To enhance assessment efficiency, he advocated for increasing human resources dedicated to this task.
Tania Constable, the chief executive of the Minerals Council, expressed disappointment over comparisons to Robodebt, asserting that the proposal is both innovative and capable of reinforcing environmental protections while optimizing efficiency.
Constable remarked, “This proposed method aims to bolster human decision-making through AI tools for both regulators and project proponents, assisting in navigating the complexities inherent in assessments and approvals under the EPBC Act.”
A spokesperson for the federal government indicated that budgetary considerations would be addressed in due course, noting that the environment department is evaluating how AI could facilitate application processes.
“Ultimately, decisions regarding project approvals will remain solely in the hands of assessment officers—not AI,” the statement confirmed.
The spokesperson added that AI technologies possess the potential to save time, mitigate uncertainties, and simplify technical jargon.
Source link: Theguardian.com.
