Oxford researchers are tackling misinformation in wildlife reporting with a crowd-trained AI tool that’s already processed millions of media headlines—transforming how conservationists harness digital data to protect nature.
A research team at the University of Oxford's Environmental Change Institute is leading an innovative project to refine how wildlife is identified in digital news, helping conservationists direct their resources more effectively.
In today's digital age, online media shapes public perceptions of wildlife. Conservationists and researchers rely on media coverage to understand how people engage with and discuss wildlife-related issues. However, current keyword-based searches often misclassify content, leading to misleading insights. For instance, an article titled 'Toronto Blue Jays Score Season High' might be mistakenly flagged as wildlife-related when it actually refers to a baseball team.
To address this challenge, a team led by Dr Diogo Veríssimo is developing a machine learning (ML)-based filtering tool that can more accurately distinguish between genuine wildlife references and unrelated content. By comparing human-classified data with AI model outputs, the researchers aim to create a system that quickly and precisely identifies wildlife-related news, reducing misclassification, misinformation, and misinterpretation of wildlife data.
A Brief Introduction to the Zooniverse
Why Does This Matter for Conservation?
Dr Veríssimo explains: 'Understanding public sentiment towards wildlife is critical for conservation efforts. When people care about animals, they are more likely to support protection measures. However, tracking public attitudes isn't always straightforward. Online media provides valuable insights, but only if we can accurately extract meaningful data. This is where AI can make a real difference. '
The project aims to develop an advanced filtering tool to help researchers and conservationists access clearer, more reliable insights from digital media. The team plans to publish their findings as an open-access research paper to ensure transparency and broad usability, demonstrating how the tool can enhance media-based conservation research.
How You Can Get Involved
A key part of the project's success relies on volunteers helping to create a 'gold standard' dataset by reviewing and classifying article titles to confirm whether they genuinely mention wildlife. This ground-truth data will be used to train and refine the AI model, making it even more accurate.
'We're inviting people from all backgrounds to take part in this effort, ' says Dr Veríssimo. Volunteers can play a direct role in improving conservation research by dedicating just a little time to sorting article titles. Together, we can build a tool that benefits scientists, conservationists, and, most importantly, the wildlife we all want to protect.'
Since the project launched just a few weeks ago, nearly 3,000 volunteers have already contributed, generating close to 2 million classifications- a testament to the power of people-driven research.
You can join the project, informally known as the 'Nature SPAM Filter,' on the Zooniverse citizen science platform.
Find out more and contribute to the Nature SPAM Filter project.