A proposed UK R&D value for money framework tackles benefits that cannot be priced. How would it work, and what safeguards could make judgements credible?