ElectionGPT: How AI-powered tools supported India’s elections––and what may be next for AI in the public sector

In 2024, elections took place in some 70 countries, home to half the world’s population. In India alone, 642 million people cast ballots in the general election, making it the largest democratic exercise in history. Ensuring the smooth conduct of Indian elections for hundreds of millions of people was no small feat and required the dedication of millions of support staff. 

Our team worked with the government of Uttar Pradesh where millions of registered voters participated in the elections. To support this monumental task, IDinsight developed ElectionGPT–– an AI-powered tool designed to help election officers in India.  The tool served as an indispensable resource, answering thousands of questions with speed and accuracy, while saving officials countless hours.

As experts note, AI has the potential to improve electoral efficiency and engagement, yet also brings risks and concerns of misuse. While many fears did not come to pass in 2024, they remain, and strong governance of this technology is important to ensure it is not misused. In this case, we developed a safe, beneficial tool to increase efficiency – providing officers with information that was not easy to access or find, despite being publicly available. This article explores how ElectionGPT supported India’s election, and explores other promising and safe future applications of AI in the public sector.

What was ElectionGPT?

One of the toughest problems election officials faced was quickly and accurately locating relevant regulations and laws from thousands of pages of material. To alleviate this burden, IDinsight wanted to create an AI-powered tool that could:

1. Take a natural language question from an official
2. Find the most relevant official documents for the initial query in just a few seconds
3. Draft a summary answer from these documents
4. Return the answer alongside a clear list of sources to the user

After initial testing, ElectionGPT was launched in early February 2024 and quickly gained traction among election officials through word of mouth, most notably in Uttar Pradesh. The system could handle queries via a desktop website and through mobile browsers. It amassed over 16,000 user questions to date.

Impact on Election Efficiency

With over 8,000 pages of documents––some including grainy scans of old papers alongside handwriting––searchable in seconds, ElectionGPT significantly reduced the time required for officials to find answers in a vast corpus of regulations.

Based on our estimates, the system saved between 1,000 to 3,500 hours through its 16,000 queries, demonstrating substantial efficiency gains to officials across several states in India. This time saving not only improved the efficiency of election operations but also allowed officials to make quicker, more informed decisions, enhancing the overall electoral process.

Transparent AI’s Role in Government

The principles behind ElectionGPT can be applied to various government sectors where large volumes of regulations and data must be navigated quickly. In healthcare, AI tools may help medical professionals quickly access protocols and guidelines. In public services, streamlining access to regulations and procedures for civil servants may improve service delivery. In education, AI may assist educators by providing quick access to policy documents and instructional materials.

ElectionGPT’s success illustrates the potential of AI to enhance service delivery in government by making information retrieval faster and more accurate. The system’s built-in page-specific citations allowed officials to verify information quickly. This feature is crucial in maintaining the integrity and trustworthiness of AI-driven tools in public sector applications.

The accuracy and reliability of AI systems like ElectionGPT depend heavily on the quality of data they are trained on. By providing domain-specific, up-to-date information, we can ensure that such systems deliver precise and reliable services. This principle can be applied to various government sectors, where large volumes of regulations and data must be navigated efficiently.

Looking at the system diagram above, it is easy to see that the highlighted “Step 2,”  where the system searches through its source documents, can easily be altered to use different source documents. Therefore, the system could be as knowledgeable in other areas of government regulation as it was in election regulation. Some great use cases are:

1. Public Services

Streamlining access to regulations and procedures for civil servants can enhance service delivery, reduce administrative burdens, and improve response times for citizen inquiries.

2. Healthcare

AI tools can help medical professionals quickly access protocols, guidelines, and patient records, streamlining the diagnostic process and improving patient care.

3. Education

AI systems can assist educators by providing quick access to policy documents, instructional materials, and best practices, allowing for more effective teaching strategies and curriculum planning.

4. Environmental Protection and Regulation

AI can help environmental agencies access regulations, compliance documents, and scientific data quickly, aiding in the monitoring and enforcement of environmental laws.

5. Transportation and Infrastructure

AI systems can assist in managing regulations and guidelines related to transportation, ensuring compliance with safety standards and streamlining infrastructure development projects.

ElectionGPT not only supported the smooth conduct of the 2024 Indian elections but also showcased the transformative potential of AI in government operations. By continuing to refine and expand such systems, we can significantly enhance the efficiency and accuracy of government services, ultimately benefiting citizens. The success of ElectionGPT demonstrates that AI has the potential to revolutionize public sector operations, making them more efficient, transparent, and reliable.