Creating ElectionGPT and FactCheckGPT

For Piyush Aggarwal, an editor from India Today the JournalismAI Fellowship provided the opportunity to develop tangible solutions from conceptual ideas. 

When Piyush Aggarwal, an editor from India Today, applied for the JournalismAI Fellowship in 2023, he had one idea. By the end of the Fellowship, he had contributed to creating not just one, but two chatbots. 

ElectionGPT allows users to engage in conversational queries about election data, while FactCheckGPT aggregates and presents fact-checked claims from various sources. 

Aggarwal said the chatbots have been developed as a generic framework, meaning they can be replicated and tailored to fit the context of any country. 

He applied for the Fellowship with Ankit Kumar, Open Source Investigations Lead and Editor at India Today. “He (Kumar) comes from an editorial background like a proper hardcore journalist, but he understands the technical stuff and it's important to have someone like him who can complement your skills, who understands technology and how it can be applied in a newsroom scenario,” said Aggarwal.  

ElectionGPT dashboard

Measuring the Impact

“When ChatGPT was launched, it changed the whole AI landscape, especially for newsrooms and how they function,” said Aggarwal. Adding that “this was one of the Fellowships where you will get a chance to work on AI on actual projects, which can be implemented in a newsroom, not just as a concept level, but as something tangible, which can be used in a newsroom,” he said. 

Aggarwal said that the Fellowship gave him confidence. “It gave me confidence because we could build something from scratch, which was an idea when we started, and by the end, we had not one, but two products,” he explained. 

During the Fellowship, Aggarwal said he learned nuances of prompt engineering, how to write an AI algorithm, and how to build AI products using AI.

Aggarwal said mentorship was important in navigating AI development hurdles. "We encountered challenges in training the model to comprehend numerical data. We had discussions with mentors and they were able to help us find a solution,” he said. 

"During the Fellowship I got the chance to speak to like-minded people from different newsrooms across the globe and to talk to experts like Charlie Beckett or other mentors from Knight Lab, such as Scott Bradley, Jeremy Gilbert, and Joe Germuska, who have been working at the forefront of AI,” he noted. Adding that “they [experts] helped us understand the nuances of Generative AI”. 

In 2023, he was invited as a panellist at the Global Investigative Journalism Conference in Sweden, where he presented a basic version of the ElectionGPT chatbot. “That was a big moment for me in terms of my career, because every journalist aspires to go to that conference, and being invited as a speaker was a big thing for me,” said Aggarwal. 

The Fellowship helped him understand where AI can be used in a newsroom. Additionally, “It helped me improve my technical and practical understanding of AI,” he said.  

Tailoring for Impact

Following the conclusion of India's 2024 general elections, Aggarwal and his team refined ElectionGPT to better serve the nation's political landscape. Collaborating with various newsrooms, including the India Today Group, they customised the chatbot to provide real-time election data and analysis.

Aggarwal and Kumar have started their second journey with the Fellowship this year. Together with his team, they aim to create an open framework for chatbots, making AI-driven and data-driven solutions more accessible to everyone. 

Watch Team Daisy's presentation at the 2023 JournalismAI Festival

Previous
Previous

Meet the 2024 JournalismAI Academy for Small Newsrooms MENA cohort

Next
Next

JournalismAI 2023/24 Impact Report