EPSRC Network+ proposal: Human-Like Computing

A computer that can read a person’s intentions, not just their typed commands, is the goal of this research network. The core problem is that today’s machines are brilliant at processing data but terrible at understanding what a human actually wants. When you ask a computer to “find a good restaurant,” it might return a list of highly-rated places, but it cannot grasp that you mean “somewhere quiet for a business meeting” or “a place my picky child will eat.” This gap between raw computation and human-like reasoning limits how naturally people can collaborate with machines. If this research succeeds, it could transform how we interact with every system that relies on human input—from medical diagnostic tools that interpret a doctor’s unspoken concerns, to manufacturing robots that adjust their actions based on a worker’s subtle gestures. The network will bring together computer scientists, psychologists, and engineers to build machines that learn from accumulated background knowledge, not just labelled data. This is primarily fundamental science. The network is not building a specific product. It is creating the conceptual and technical foundations for a future where computers are genuine collaborators, not just tools that require precise, literal instructions.