Essay Finalist … Varsha Vijay
We are pleased to share with you an essay from one of our essay finalists: Varsha Vijay from Notre Dame High School in San Jose, California. His essay illustrates his attentiveness to relevant issues that face all of us as it relates to climate change as well as issues relevant to him as a student. Varsha explores the issues with biases in data and the associated impacts.
His essay is incredibly thought provoking. We invite you to read it below.
Essay from Varsha Vijay:
When I first watched Planet Earth, narrated by David Attenborough, there was one moment I could never forget: a polar bear, stranded on a shrinking ice floe, adrift in a vast, melting sea. Its fur, soaked and matted, clung to its body as it struggled to find footing on the vanishing ice. The image was haunting, not just because of the bear's silent desperation, but because it felt like I was watching something irreversible unfold. It wasn’t just the bear’s struggle that unsettled me; it was the realization that this wasn’t a distant tragedy. This was the consequence of human action, unfolding in real-time. The bear wasn’t just floating. It was sinking with the weight of our choices. I wanted to do something. I wanted to harness the power of AI to predict and prevent climate disasters, to turn data into a tool for protection rather than a record of destruction.
But as I dove deeper into machine learning and climate data, I realized something troubling: data isn’t always as neutral, objective, or useful as we assume. Biases lurk in datasets, shaping AI predictions in ways we don’t always see. Incomplete records can lead to misleading models. The very tools we rely on to fight climate change can become flawed, reinforcing disparities instead of solving them. If the foundation is unstable, how can we build technology that truly makes an impact?
The biggest challenge in our use of data and technology isn’t the technology itself. It’s the integrity of the data we feed into it. I learned this firsthand when I worked on a research project predicting wildfire sizes in California. I had access to over 72,000 wildfire records, and at first, the sheer volume of data seemed reassuring. But as I trained machine learning models, inconsistencies began to emerge. Some years had missing data. Certain regions had vastly different reporting standards. When I ran my model, I saw the effects firsthand: predictions skewed, favoring well-documented regions while overlooking fire-prone areas with incomplete data. I felt a sinking feeling in my stomach. It was frustrating, knowing that the technology I was working with could be so much more, but the data was failing to give vulnerable communities the attention they needed. This was personal. I was committed to protecting those at risk, but the data wasn't allowing me to do that effectively.
This wasn’t just an isolated issue. I saw it in environmental research, where pollution studies often overlooked or misrepresented the struggles of marginalized communities disproportionately impacted by climate change. These communities, already vulnerable, were further ignored in data that failed to capture the full scope of their experiences. I noticed similar challenges in my own academic work, where data gaps hindered the accuracy of models, especially when dealing with regions that weren’t adequately represented. Even peers working on AI ethics projects shared their struggles. They encountered data sets that didn’t reflect the diversity of the populations they aimed to serve. In all these instances, the technology itself wasn’t the problem. It was the incomplete, biased data that shaped it. When the foundation of data is flawed, technology merely magnifies those existing disparities, reinforcing them rather than solving them.
Data, in many ways, is a reflection of who we choose to see and what we choose to record. If the polar bear had never been filmed, its struggle might have gone unnoticed. If climate models exclude certain regions, their struggles remain invisible. The challenge, then, is not just about building better AI but about ensuring the data we use is complete, fair, and representative of reality.
So, what can be done? This question lingered in my mind until I realized that the solution begins with action. For me, the first step was ensuring the integrity of my wildfire dataset. I worked to clean, standardize, and supplement missing data, advocating for open-access environmental datasets so researchers wouldn’t have to start from scratch. Through my experience, I saw how crucial it was for researchers to collaborate and share data to ensure no one was left behind. But this is only a small part of the solution. We need a larger movement, one that involves students, educators, and organizations, toward ethical data practices.
I realized that solving this issue wasn't just about analyzing datasets. It was about ensuring that every researcher, from students to professionals, understands the importance of data integrity. I’m committed to ensuring that data, whether in my future work or through advocacy, is both ethical and inclusive. For me, the first step is sharing this message and leading the charge for better data practices. Schools could integrate data ethics into computer science courses, making students aware of biases before they even write their first algorithm. Research institutions could push for more standardized, transparent datasets, ensuring that gaps in data don’t become blind spots in solutions. Organizations fighting climate change could invest in crowdsourced data collection, empowering local communities to contribute to datasets that represent their realities. But most importantly, we need to shift the way we think about technology. AI is not an all-powerful solution, nor is data an unquestionable truth. They are tools, only as good as the hands that shape them. If we want a future where AI truly combats climate change, it won’t be enough to build smarter models. We must first ensure that the foundation, the data itself, is just as strong.
I still think about that polar bear sometimes, floating on its shrinking ice floe. But I no longer see it as a passive symbol of loss. It’s a call to action, a reminder that technology should not just document the world’s unraveling. It should help us stitch it back together. I’m committed to ensuring that the future of technology, driven by data, is one that reflects the full spectrum of humanity and safeguards our planet.
