I keep a digital copy of Ranganathan’s Five Laws of Library Science on my desktop. Not for decoration. I go back to it the way a carpenter goes back to a plumb line when something does not look right.
Last year I was sitting with an engineer from a technology company who was explaining, with considerable enthusiasm, how their new AI system personalises information delivery, ensures every user finds what is relevant to them, saves search time dramatically, and continuously learns as new content enters the system. He was proud of it. It was genuinely impressive.
I asked him if he had ever heard of Ranganathan.
He had not.
Shiyali Ramamrita Ranganathan was a mathematician from Tamil Nadu who became a librarian almost by accident, spent a year studying at University College London, and came back to India to write, in 1931, what I now think of as the oldest AI design specification in existence. The Five Laws of Library Science. Books are for use. Every reader his book. Every book its reader. Save the time of the reader. The library is a growing organism.
Read those laws again, slowly, with an AI system in mind.
Every piece of information exists to be used, not stored. Law one. Every user should find information suited to their specific need, not a generic average of what most people look for. Law two. Every piece of information should find the right person, not just wait to be discovered. Law three. The system should reduce friction between need and answer. Law four. And the entire system must grow, learn, and adapt as new knowledge enters it. Law five.
The engineer I spoke to had spent three years building a system that Ranganathan had specified in a single paragraph in 1931. Without a neural network. Without training data. Working from first principles about what information systems are for.
There is a further layer to this. Ranganathan’s other great contribution was Colon Classification, a faceted system that breaks any subject into its component dimensions: Personality, Matter, Energy, Space, Time. It was, for its era, an extraordinarily counterintuitive idea. Instead of placing a subject in one fixed location, you describe it by its multiple facets and allow it to be found from any of them. Modern knowledge graph construction, the architecture that makes AI reasoning possible, does exactly this. Different vocabulary. Same intellectual operation.
I do not raise any of this to suggest that AI engineers should have read more library science. Most of them are solving hard problems well, and the tools they have built are genuinely useful. I raise it because there is a generation of LIS professionals in India who were taught these ideas and then told, implicitly or explicitly, that they belonged to a declining field.
They were taught the theory of intelligent information systems at a time when nobody was calling it that. That training has value now that was not visible then.