AI for climate won’t work unless it’s built with communities

Is a caterpillar a pest? It’s not a question I expected to be stumped by at all. In a world where you feel unshakably close to any information if you have a smartphone with internet in your hand, the classification of a caterpillar is not supposed to be particularly baffling. Yet Aranya Sahay’s beautiful film ‘Humans in the Loop’ gently tells you to question the veracity of the assumptions we often make about nature, reminding us how invaluable the quiet wisdom of communities whose lives are built around nature, is. 

In a now viral clip from the film ‘Nehma’, a woman from a local tribal community enlists as an operator at an Artificial Intelligence (AI) data center, labelling objects to train AI models used in various industries. She refrains from labelling a caterpillar as a ‘pest’, arguing from her experience that the organism in the photo she was required to label only ate rotten bits of leaves, helping plants fight infestations. However, the foreign tech company she works for is adamant and insists that she sticks to the labels they have sent her. The conflict, and the unequal power dynamic underlying it, lies at the heart of developing Artificial Intelligence that aspires to enrich natural ecosystems that for years have been protected and cherished by the people living amid them. 

In the New York Climate Week I attended this year, as conversations seeking ‘AI-led climate solutions’ gathered steam, a recurrent concern that surfaced was: how do you centre communities in an AI discourse, amidst the disparate access to technology resources across the world. 

The answer lies in the wisdom of what community leaders I’ve met have always said—solutions that aspire to make lasting, defining impact have to be designed with the people, not excluding them. And the work of community-based organisations in India, who have laboured through economic downturns and pandemics to carry resources, and sometimes just reassurance, for the country’s diverse, far-flung communities offers strong models of how that can be achieved. 

Recent surveys have revealed that 95% of Indian villages are covered by 3G/4G networks. It is widely known that in India and across South Asian societies, availability of technologies doesn’t guarantee equitable access to it. In fact, some families have just one mobile phone with just basic functionalities available to them. However, across the country, data-based early warning systems, monitoring systems, and predictive systems for disaster risk reduction have been deployed to disseminate critical information that is targeted at communities, many of whom don’t exist in an ecosystem that enables them to process or act on the information with adequate resources. 

The success of these systems has therefore traditionally depended on community networks, an informal chain of people held together by mutual trust and the shared experience of battling hyper-local challenges and oftentimes overcoming them with collective knowledge. And this puts them in a unique position to help build climate solutions rooted in the pulse of diverse eco-geographic landscapes and their socio-political realities. 

Take, for example, a community-based organisation we work closely with. Located in Rajasthan, where groundwater levels have declined by over 62% between 2008 and 2018, this organisation helped revive traditional water harvesting and recharging methods used by the locals for centuries with value additions based on scientific research and data-modelling. When the organisation began work in a cluster of villages years ago, among other issues, their traditional water conservation system was in disarray and clean drinking water was scarce. Their team, populated with people who have grown up around these villages and are intimately familiar with the cultural shifts that led to the gradual decline of water reserves and conservation structures, prioritised local knowledge and sustainability while designing solutions. Naadis or shallow ponds dug up by the locals to store rainwater were slowly becoming unusable due to silting. So, the organisation worked with the communities to build loose stone structures placed at optimal distances from the banks – and this stopped the gush of debris and silt from flowing into the naadis. They have also built a bio filter with varieties of abundantly available gravel and sand arranged in layers and then topped with an inexpensive chemical solution. This is fitted into water harvesting structures to ensure that the rainwater harvested is clean and safe to drink. The organisation used contextual data – most recent rainfall patterns, local economic profiles, gender roles, among others – generated from close interactions with communities over time and used it to design solutions for growing water scarcity. Having this information ensured that their solutions were cost-effective and had long-term use – an approach AI innovators could embrace to build solutions that can be scaled and adapted for India’s diverse communities. 

The availability of climate data in India often depends on the accessibility of the location – so cities and highly connected rural areas with strong tech infrastructures will often have complete and updated climate or weather data that could be used by AI models. However, in large swathes of the country facing extreme climate events official data is often found to be incomplete or completely unavailable, due to the absence of effective monitoring mechanisms. The information on changing climate patterns and their impact on people’s lives then rests with communities, whose lived experiences will therefore be important in reducing biases in AI climate models.

The scale of impact of an AI-based climate tech in a country—where people are mired in complex social relationships—will also depend on understanding the cultural systems that dictate how people embrace ‘solutions’. 

An Odisha-based group we have worked with built their climate interventions bottom up – informed by the urgent needs of the communities. Their community mobiliser, equipped with deep knowledge of the people they live among, brought a local small-holder farmer’s burgeoning struggles to the organisation – his crop yields had consistently dwindled, forcing him to take up wage labour. The organisation, comprising people with lived experiences of the landscape, examined his situation and suggested a method of rice cultivation that did not require intensive investments or resources the farmer couldn’t afford. With training on better soil aeration techniques, weeding, spacing of saplings, use of organic fertilisers and early transplantation, the farmer managed to increase his yield manifold. 

The insight that empathy brings will be critical in fortifying communities in the face of climate emergencies – as this organisation has illustrated. 

For years, scores of organisations working with rural farming communities have been analysing weather data and climate warnings to simplify the information in a fashion that is easy for locals to understand. Their involvement often extends to training them on farming practices that help them tide over these challenges. Community-based organisations in India are the largest reservoirs of socio-cultural data and empowering them to document this knowledge and make it accessible is a massive yet critical task for stakeholders in the social impact sector. And only long-lasting collaborations between AI entrepreneurs, community-based organisations and climate finance institutions will ensure that.

(The author is the CEO of EdelGive Foundation. Views are personal.)