Increasingly cost-effective environmental sensors combined with AI-supported evaluation tools promise faster and more insightful environmental planning.
The need for higher decisions about how we use ecosystems and natural resources is much more urgent today because the Draft law on accelerated approval require faster assessments.
As a part of our research at Digital dooran open-access and collaborative project to gather data across land and water, we found that there’s a real appetite amongst Iwi and Hapū tribal groups to have interaction with AI.
Overburdened kaitiaki environmental watchdogs recognized the chance that AI could help integrate fragmented environmental data sets while improving analytical capability quickly and cost-effectively.
Based on this need, the Kuaha Matahiko project developed a working AI trained using environmental data from Aotearoa New Zealand, demonstrating that a tipping point is approaching where bespoke AI is rapidly becoming a sensible option for kaitiaki groups, even small ones.
However, caution is suggested. Previous experience show that algorithm-based systems often limit us to practices that reproduce existing inequalities in data collection and exclude imagination in regards to the results.
These problems often arise as a consequence of two interrelated issues: the legacy of ad hoc data collection and the customarily false belief that larger data volumes mean greater accuracy.
The “precision trap”
First, useful AI systems require large amounts of knowledge at high speed and in high volume. The Parliamentary Commissioner for the Environment has warned successive governments that New Zealand’s environmental data system ad hoc, opportunistic and underfunded.
Existing environmental databases largely reflect the priorities of government-led agricultural science and up to date efforts to watch its environmental impacts. Our environmental data also suffer from a scientific ignorance of Māori education.
Long-term environmental data sets are very useful, but they supply very incomplete coverage of places and problems, and we will't return in time to repeat data collection. Identifying the gaps and biases introduced by a history of uneven, exclusionary data generation is critical, because this data (and the assumptions it accommodates) shall be used to coach future AI.
An agricultural robot working on a sensible farm and spraying corn fields with fertilizer
Secondly, AI guarantees security and precision. But a Study examines precision agriculture describes the risks that arise after we confuse the high volume and granularity of massive data with high accuracy. An exaggerated belief within the precision of massive data can result in an erosion of checks and balances.
This is becoming a much bigger problem as algorithms change into an increasing number of opaque. Most algorithms at the moment are incomprehensible. This is as a consequence of technical complexity, lack of user understanding and deliberate strategies by developers. It blinds us to the risks of inaccuracy.
If we ignore the opacity of algorithms, we risk falling right into a “precision trap.” This occurs when belief within the precision of AI results in an unconditional acceptance of the accuracy of AI results. This danger exists due to political, social and legal value we attribute to numbers as a trustworthy expression of objective “hard facts.”
These risks increase rapidly when AI systems are used to predict (and control) future events based on precise but inaccurate models that usually are not based on observations. But what happens when AI results form the premise for evaluation and decision-making? Do we then have the choice of not believing them in any respect?
Avoiding an “iron cage”
A possible future lies in what the German sociologist Max Weber called “iron cage of rationality“Here, communities change into trapped in rational, precise and efficient systems which can be concurrently inhumane and unjust.
To avoid this future, we must proactively create inclusive, comprehensible and diverse AI partnerships. This isn’t about rejecting rationality, but about mitigating its irrational consequences.
Our evolving framework for data and AI governance relies on the principles of discoverability, accessibility, interoperability and reusability (JUST). These are very useful. They are also blind to the social history of knowledge collection.
The Failure of the 2018 census is a stark example of what happens when historical inequalities are ignored. We can't recreate the environmental data now we have. But latest AI systems need to pay attention to the impact of past data gaps and construct them into their design. That may additionally mean going beyond awareness and actively enriching data to fill gaps.
Expanding the worldview of AI
Data and AI must serve human ends. Indigenous data sovereignty movements demand the suitable of Indigenous peoples to own and manage data about their communities, resources and lands. They have inspired frameworks generally known as CAREwhich stands for common good, authority, responsibility and ethics.
These provide a model for strengthening data relationships that puts thriving human relationships first. In Aotearoa New Zealand The Māori Data Collection was established in 2019 as an independent body to enable Māori to access, collect and use their very own data. Their data management model is a practical example of those CARE principles.
An even greater step forward can be to expand the worldview of AI. Serving human goals means exposing the assumptions and priorities inherent in several AIs. That, in turn, means opening up the event of AI beyond the “WEIRD” – western, educated, industrial, wealthy, developed – viewpoint that currently dominates the sector.
Training an AI for Māori organisations using environmental data from Aotearoa New Zealand is one thing. Creating an AI that embodies mātauranga Māori and the responsibility for all times embedded within the Māori worldview is a more radical thing.
We need this radical vision of AI, which consciously builds on the inspiration of various worldviews, to avoid locking ourselves within the cage and shutting ourselves off to the longer term.
