NEW DELHI: Deputy Chairman of the Rajya Sabha, Harivansh Narayan Singh, underscored the importance of human institutional memory in shaping application of artificial intelligence (AI) in legislatures.
He emphasised that if AI is to be introduced into legislative institutions, it must be accountable, contextual and trustworthy. He made these remarks at a workshop on the adoption of AI in Parliament, held during the 28th Conference of Speakers and Presiding Officers of the Commonwealth in New Delhi.
During his address, he also outlined the range of digital and AI tools being developed to improve efficiency within the Indian Parliament.
Highlighting the need for a hybrid approach to AI development, he said, "When a human enters a new organisation, they bring with them two essential attributes: skill and knowledge. Skills can be acquired, transferred, or outsourced. Knowledge, however, is contextual and deeply embedded within the institution. Parliamentary knowledge is unique. It is built over decades through debates, rulings, conventions, and constitutional practices. The same principle applies equally to Artificial Intelligence."
He further noted that the institutional knowledge of humans is central to the development of accountable AI for parliaments.
Stressing the importance of oversight, he said, "Human oversight and the ability to intervene must be integral to the system. Innovation without restraint carries risk, while restraint without innovation can lead to stagnation." He added that Parliament must therefore maintain a careful and deliberate balance between innovation and restraint.
Elaborating on the practical use cases of AI already being implemented in Parliament, Harivansh cited the use of AI models for translating official documents, analysing parliamentary debates and framing questions in 22 languages.
"We have developed a Parliamentary Language Dictionary consisting of approximately 48 thousand terms, which has been integrated into a custom AI model designed exclusively for parliamentary use. This has significantly increased acceptability among internal users and improved its accuracy. Human translators remain firmly in control, with AI serving as an assistive tool," he said.
However, he cautioned that in a legislative context, "artificial Intelligence in Parliament must be rooted in truth, bound by ethics, guided by human judgment, and accountable to democratic values". He also called for enhanced parliamentary cooperation among Commonwealth nations in the application of AI.
The two-day conference was hosted by India for the fourth time, following the previous editions held in 1971, 1986 and 2010.