Assessing the science–society relation: The case of the US National Science Foundation's second merit review criterion