As the mother of a toddler, working in academia at the intersection of AI and climate justice, I can’t help but see uncanny parallels between the behavior of my child and AI systems. Both are in the early stages of encountering a complex world and trying to orient themselves within it, shaped by the stories and information we feed them and by whose experiences we choose to center or consign to the margins.
I began noticing these parallels in the everyday work of giving instructions. A prompt, I realized, tells the system what to notice, what to ignore, where to begin and how to do things. My prompts to an AI system had to be specific, provide context, break tasks into parts and set explicit expectations—much like the way I teach something new to my toddler. If I just say “Clean the house,” my child, quite logically, interprets the instruction through their own limited framework of what “clean” means. But if I give a more specific instruction, “Go to the playroom, pick up all the toys and put them in the red box,” the task becomes clear and the outcome improves.
While this comparison may be amusing, it also carries a heavier realization: the parallel does not stop at prompting but extends to responsibility. In both parenting and AI models, we are not simply trying to produce the right response in the moment. We are passively shaping patterns of attention, judgment and action that may persist beyond the immediate interaction. What is repeated becomes familiar. What is centered becomes important. What is ignored can become invisible. That is where the stakes of responsible AI begin.
Responsible AI for climate governance
Raising a responsible child and building responsible AI both require a deep awareness of what we are training them to notice and value. Just as a child can grow into someone who is not only capable but also compassionate, AI systems can be designed not only to optimize outcomes but also to support fairness, accountability and long-term sustainability. In climate governance, this is particularly urgent because the goal is to build smarter systems that also help us become more responsible stewards of a rapidly changing planet. When designed responsibly, AI can help us see what has long been overlooked in climate governance. It can identify vulnerable communities that have historically been excluded from decision-making, strengthen early warning systems, enable faster and more targeted responses to extreme weather events, optimize energy systems, accelerate the transition to renewables, and help cities anticipate heat stress, flooding and infrastructure strain. It can help shift climate governance from reactive crisis management to anticipatory, justice-oriented planning.
It takes a village
It takes a village to raise a child, and perhaps the same is true of AI. A responsible AI system cannot be raised inside a narrow room of technical expertise, institutional settings or market priorities. It needs a wider perspective from those whose knowledge has too often been treated as peripheral: women, frontline communities, Indigenous communities, diverse ethnic and racial groups, low-income communities, disabled people, and others living at the intersections of marginalization. Instead of treating these individuals as “stakeholders,” we should treat them as knowledge-holders whose lived experience can reveal risks, harms and possibilities that technical systems often fail to see. Their experiences reveal what dominant datasets often miss: unpaid care work, gendered vulnerability, environmental exposure, displacement, informal economies, cultural knowledge and the everyday strategies people use to survive systems not built for them. Without these perspectives, AI may still appear intelligent, but it will inherit a dangerously partial view of reality just like a child raised without a wider circle of care and accountability. Such a child may still learn language, routines and patterns of behavior, but their understanding of the world might remain constrained by the limits of the environment in which they were raised.
The need for ethical guardrails
Just like a child can be guided to correct mistakes, AI systems require continuous oversight and accountability. These systems inherit and amplify the assumptions built into them. Left unchecked, they can reproduce the very inequalities they are meant to address. Like any learning system, AI requires a continuous practice of evaluation, monitoring and repair. As it becomes embedded in decisions about infrastructure, risk, adaptation and resource distribution, the stakes are no longer merely technical. They are social, political and environmental. Responsible AI, therefore, requires deliberate ethical guardrails that shape how systems learn, decide and act in ways that account for power, justice and ecological consequences. If AI is to play a meaningful role in climate governance, it must be shaped by justice, care and responsibility from the very beginning.
A responsible AI system in climate governance must be constructed using datasets that are diverse and representative, thereby minimizing the reproduction of existing exclusions and biases. This principle extends to model design, where fairness, equity, transparency, and robustness must be operationalized as core design criteria rather than treated as secondary considerations. AI systems must also have clear accountability structures, making them answerable not only to the institutions that develop and deploy them, but also to the communities whose lives and futures are affected by their outputs.
With these issues in mind, the development and governance of AI systems should be guided by key evaluative questions: What forms of climate harm must be anticipated and mitigated? Who holds the authority to intervene when failures occur? A responsible model, then, cannot be limited to recognizing patterns. It must be capable of questioning—creating space for uncertainty, alternative ways of knowing, and what is not immediately legible in the data.
Raising the future in the present
AI and children share a profound truth: they will outlive us, carrying forward the worlds we helped teach them to see and value. The systems we design and the values we embed do not disappear; they persist, shaping decisions, norms and possibilities long after their origins fade from view. In climate governance, where the consequences of today’s decisions already unfold across generations, this continuity is immediate, consequential and impossible to ignore. What we are shaping, whether in people or in systems, is not just capacity, but orientation. And the question that remains is not whether these AI systems will influence the future, but whether they will reproduce the limits of the present or help us move beyond them.
Pavithra Priyadarshini Selvakumar is a postdoctoral research scientist at the Columbia Climate School. She is interested in exploring how AI and climate justice can be integrated to support equitable resilience planning in frontline communities.
Views and opinions expressed here are those of the authors, and do not necessarily reflect the official position of the Columbia Climate School, Earth Institute or Columbia University.
