Recommendations

Machine learning technologies are expected to play a significant role in addressing climate action - they can improve our understanding of complex systems, optimise the working of existing systems, and advance scientific research. However, the use of AI is also fraught with multiple challenges and risks.

This brief provides recommendations for global philanthropy and development organisations to support the responsible development and use of AI for climate action. It is important to remember that AI may not always be the best solution to address a problem. Its use should be weighed against other technological and non-technological solutions, particularly as there is a risk that it may distract or divert resources from less “flashy” tools or approaches.

1. Localised and Participatory Data Collection

   Localised data sets that reflect the priorities of vulnerable communities and geographies are needed for inclusive and effective AI interventions. Private players are not incentivised to collect such data unless it is commercially viable. Investments by global philanthropy and development agencies can fill this gap through support for consultative and problem-driven data collection and curation; data communication and visualisation initiatives that support local stakeholders in evaluating and monitoring data insights; and citizen science initiatives.

2. Interdisciplinary Collaboration

   Solutions that deploy AI will only be as good as the definition of the problem, but many AI models are based on available data and insights generated from that data, rather than a specific theory of change about how the problem at hand can be addressed. Building such a theory of change requires domain knowledge and localised expertise. Development organisations can support platforms and initiatives for sustained interdisciplinary collaboration between data scientists, climate scientists, domain experts and local stakeholders. Particular attention should be paid to ensuring that indigenous knowledge and practices are incorporated. The ability to access and foster such partnerships is a unique leverage that development organisations have over private technology companies.

3. Ongoing Evaluation

   Investments in AI for climate action are growing in the region. However, there is almost no evidence of the impact and outcomes of these interventions. Development organisations are well-positioned to develop an impact assessment framework and related measurement tools for their investments, which can also be used by the broader ecosystem. These frameworks and measurement tools should be developed with civil society and community organisations, and linked to a broader theory of change for AI use in climate action. This can also help build a repository of evidence of successful examples of AI use, including best practices and how values such as equity and privacy are safeguarded.

4. Equitable Data Sharing

   New arrangements for equitable and safe data sharing are needed, that are rooted in principles of collective responsibility, reciprocity and accountability. Development organisations can invest in experiments to develop new legal and technological tools for data sharing. For example, new forms of data licensing and contracting for climate data are required in the region - having such licenses or contracting arrangements in place can give actors confidence in sharing their data. These contracting tools should enable parties to decide how to allocate data access and usage rights efficiently, as well as the types of liabilities associated with the data.

5. Demonstrative Projects and Pilots

   Global philanthropy can play a catalytic role in providing civil society and community organisations with the financial and technical resources they need to develop and test pilots. Even beyond the pilot stage, financial support is required to sustain the adoption and use of AI interventions. They can support and help unlock the value of pilots, which governments and multi-lateral agencies can then scale in the region. Seeing the success of these pilots can help catalyse government action and can also be used to advocate for further investments in digital infrastructure, data collection, and community-centred models for technology design.

6. Foundational Research

   An opportunity area common across all domains is the potential of ML for advancing scientific research and knowledge for climate action. Market dynamics alone may not facilitate such research, nor a thorough phase of testing and evaluation. While the market is flushed with applied AI applications, global philanthropy has a role in driving such foundational research.

7. Greener Computing

   The energy and water consumption of AI and other harmful environmental impacts are expected to increase. Using AI to address climate action can thus be counterproductive unless advancements are made in green computing. Global philanthropy and development organisations can support foundational research in more energy-efficient and less polluting forms of computing.

Digital Public Goods (DPG) and Digital Public Infrastructure (DPI) are expected to play a pivotal role in supporting climate action. This action brief outlines recommendations for global philanthropy and development organisations to advance equitable, safe and trustworthy DPGs and DPIs for climate action.

As these organisations advance the DPG and DPI agenda, it is important to recognise that open technical architectures are a means, not an end - openness will not in itself enable positive climate action or support localised climate interventions. On the contrary, open data systems can compromise the data sovereignty of communities and regions. Moreover, the focus on technological innovation through DPGs and DPIs should not become a substitute for political action nor reduce the responsibility of powerful players to change policies and practices.

1. Centre a coordinated climate vision

The development of many current DPGs for climate action is reliant on individualised efforts by small groups across civil society, academia and development organisations. They tend not to be informed by a coordinated strategy for climate action at a local, national, or regional level. A more impactful way forward would be to develop a coordinated action plan for DPGs aligned with government strategies and plans for climate action. Global development agencies and philanthropy can play an important coordination role, working with national and local governments and regional bodies to identify and support the type of DPGs required for effective climate action.

2. Support Community-driven DPGs

Many DPGs are developed top-down and led by technical experts. Instead, we need to develop DPGs that respond to the specific climate vulnerabilities of communities - that help them adapt to climate change and advocate for action from government and relevant authorities. Global development organisations and philanthropy can play a critical role in enabling bottom-up and purpose-driven data collection and tool building that local communities use and manage. One of the main advantages of a DPG approach is that it can enable a wider community of actors to build tools relevant to their local contexts. Local civil society organisations will also need financial and technical support to do this.

3. Finance DPG maintenance and sustainability

A recent survey found that the top issue concerning DPG providers is the need for a sustainable business model and financing. Considerable costs are entailed in not just building DPGs, but also regularly updating and maintaining them. It is not surprising that some of the more notable DPGs have been built and maintained by large international organisations (UNDP), academic institutions in partnership with international donor agencies (DHIP2) or volunteer networks with alternate financial means (iSPIRT). Global philanthropy and development agencies can support DPGs for climate action by financing open-source teams to build and maintain DPGs. In addition to financial support, open-source developer teams will also benefit from access to domain experts and support on scaling, governance, and risk mitigation strategies.

4. Develop Assessment Frameworks and Capacities

Along with investments in building and maintaining DPGs, evaluating their outcomes and impact on climate action is important. This can also help identify what type of tools are most effective and help guidelines and best practices for future DPG investments. Impartial and publicly available evaluations can build trust in DPGs and drive adoption among key stakeholders. Furthermore, these evaluations must not be top-down but participatory - allowing target communities to provide feedback on their experience and establish further areas for improvement. This will also require providing communities with the resources, tools, and capacities to evaluate these tools. Global philanthropy and development agencies can help develop frameworks, tools, and mechanisms that assess the impacts of existing and new DPGs on climate action on an ongoing basis.

5. Considered approach to DPIs

One of the key use cases for DPI is the provision of benefits and relief to climate-vulnerable populations. Before embracing this approach, it is important to learn lessons from the use of digital IDs in humanitarian contexts. Where global development agencies support DPIs for climate action, they must first assess the potential risks and the availability of risk mitigation measures, including an assessment of the political and regulatory environment especially in areas of conflict. Technology should not be the only means to deliver benefits and alternate routes should be considered so that we can reach the most affected populations.