Table 3 Data justice characteristics for more equitable climate action AI.
From: Harnessing human and machine intelligence for planetary-level climate action
Social justice characteristic | Application to data justice for a less biased human-in-the-loop AI | Implications for a climate action AI design | |
|---|---|---|---|
| Â | Injustice issues | Justice measures | Â |
Instrumental (fair ends) | Inequitable distribution of the benefits or risks of data sharing or AI programming. | Guaranteeing the rewards of AI reduces lessen rather than worsen economic and social vulnerability. | Inclusive and diverse pre-training datasets create more accurate and efficient climate models that can help in better vulnerability estimation, leading to improved climate mitigation and adaptation planning at urban, regional and national scales. |
Procedural (fair means) | Exclusion of key stakeholders from data discussions and obfuscation of biases within AI programming. | Ensuring that individuals and groups have a meaningful voice in the design and operation of AI systems, including the right to consent (or withdraw consent) for the use of their personal data. | Effective consensus generation for climate action by leveraging diverse and context-specific collective intelligence. Improved renewable energy forecasting and energy management in a changing weather context that promotes energy justice in vulnerable areas. |
Distributive (fair distribution) | Consolidation of intellectual property and big data among corporate or technology elites | Maintaining open and accessible data, and community based ownership platforms (e.g., cooperatives, community interest companies) and prioritizing minority or vulnerable groups in digitalization processes and outcomes. | Better identification of climate vulnerability groups and their social structures, resulting in better mitigation and adaptation measures across diverse population groups. Enables robust decision-making and accurate estimation of the distributive impact of disruptive climate technologies associated with carbon capture and storage. |
Recognition (fair voice) | Discrimination against particular demographic or indigenous groups and failure to respect privacy | Including groups in design and governance of data systems. Providing stronger accountability measures. | Enables collective intelligence at the pre-training stage, which can embed valuable grounded knowledge and perceptions of climate change impacts. It further strengthens an epistemic web of planetary health challenges. Climate models become more grounded and context-driven, leading to better forecasting and disaster preparedness. |
Spatial justice (fair geography) | Uneven siting and location of digital infrastructure, as well as uneven access to data. | Providing access to digital infrastructure and resources as a common pool and open resource rather than a restricted product, improving accessibility for all stakeholders. | Climate models become spatially effective across varying granularities. It enables better mitigation and adaptation planning in vulnerable areas, as well as improved disaster resilience. Collective intelligence gets strengthened and streamlined due to the shrinkage of the global digital divide. Cities are planned with greater emission reduction potential. Social structures are made more resilient through diverse and inclusive people-centric datasets, leading to the efficient use of collective intelligence. Better energy management and renewable energy forecasting across geographical scales will lead to improved accountability for emission reduction actions. |
