The benefits and risks of using AI in Nature Restoration

Imagine this scenario. You’re working on a project supporting agroforestry and green economic growth. Part of the project involves guiding local farmers on techniques and advice for growing new fruit trees. But you and your team are based in the capital, several hours away from the project, making travelling to provide this guidance time-consuming and expensive. However, a new AI tool is available that could provide the same advice remotely, via an app. Would you use it?

Using the app could lead to a revolution in how information is successfully shared between farmers. But what if the AI provides bad advice, inventing fake solutions that are inappropriate or unhelpful? And can less tech-savvy farmers be persuaded to use it in the first place? If you find these problems emerging, how can you fix them?

These kinds of benefits and risks arising from the use of AI tools are among many others that a new report commissioned by the Reversing Environmental Degradation in Africa and Asia (REDAA) programme attempts to answer. The report reviewed 68 artificial intelligence (AI) tools to assess the potential applications, prospects, risks and barriers of using AI for locally led nature restoration and conservation initiatives.

The REDAA programme supports research and action by offering grants and facilitating mutual learning to reverse environmental degradation. It is funded by UK International Development from the Foreign, Commonwealth and Development Office (FCDO) and managed by the International Institute for Environment and Development (IIED).

The report identified the following key prospects where AI could have the most beneficial impacts.

1. Biodiversity Monitoring, land use and modelling – AI tools can be used to help identifying species and tracking ecosystem health, as well as conduct real-time analysis of environmental changes like deforestation. AI tools could also help with predicting environmental disasters, enabling mitigation.
2. Local Research and participatory design – AI tools could provide benefits at local levels, if well integrated with local research systems and governance structures.
3. Local-level business models – These can benefit from AI tools like virtual assistants and chatbots that could provide tailored training or support to key groups, such as smallholder farmers.
4. Financing and funding – AI tools also exist that can help with proposal writing and grant management, which can improve funding access.
5. Inclusive Governance systems - AI tools have potential to improve Governance systems by increasing transparency and accountability, for example by simplifying complex information and providing real-time alerts.

However, the report also identified the following risks which were seen as having the potential to cause major problems.

1. Hallucinations – AI models can sometimes produce incorrect information due to flaws or limited context, leading to potentially coherent but incorrect responses. For example, in biodiversity monitoring, species could be mislabelled, or farmers could be told to plant crops in the wrong season.
2. Bias – AI tools could be bias towards certain species, ecosystems or regions due to data limitations or developer preferences. Such biases could lead to worse or even negative outcomes.
3. Transparency and Integrity – The ‘Black Box’ nature of most AI tools limits understanding and replicability of results, which could undermine decision making and make it harder for users to trust its outputs.
4. Privacy and Security – Weak regulations on AI and poor security measures could lead to data breaches or exploitation. There is also the danger of AI tools being used to monitor communities unlawfully.
5. Exploitation and sovereignty – Indigenous Peoples & Local Communities (Ips & Lcs) may lose power over their traditional knowledge and practices if these practices are inputted into AI tools, since they wouldn’t be able to control who uses the data or profits from it. This might reinforce existing power imbalances and undermine sovereignty and community rights.
6. Dependency – If communities are reliant on foreign AI tools, they may be forced into compliance with particular rules or conditions that could undermine traditional practices and agency, leaving them dependent on these tools, risking a catastrophe if the tools are withdrawn.
7. Unequal access – There is a potential for AI tools to worsen existing inequalities. For example, AI tools for farmers may only be affordable for richer farmers, leaving poorer ones further behind.
8. Misrepresentation of culture and traditional knowledge – AI can often interpret cultures or local research practices/methods through a lense of Western Scientific approaches, which can lead to solutions that might be inappropriate for Ips & Lcs.

The report also makes recommendations for key stakeholders, as well as ways for research programmes to engage productively with AI tools. These include ensuring the closing of the technology gap that exists among different groups, developing a clear and transparent AI policy, and ensuring good quality data for use in AI tools.

Ultimately, AI tools are here to stay and, as the report makes clear, they have huge potential to bring benefits for nature conservation and green growth. While the risks cannot be ignored, they can be mitigated. AI tools should be implemented with caution and with consideration for all relevant stakeholders. Doing so will help improve outcomes and make the tools more effective.

If you want your Agroforestry and green economic growth project to succeed, the best outcome could be to test the AI tool with the farmers collaboratively. This could allow flaws or bad advice to be found and fixed. It could also help those who might be less tech savvy learn how to better engage with the tool and improve its usefulness. If the flaws and risks can be mitigated, then the benefits of AI can be more keenly felt and can help improve outcomes across all kinds of projects.

- Alasdair Brown, IIED