Building scientific knowledge implies making ontological assumptions, experiments are theory-laden and data are a partial construct informed by researchers’ preferences. In short, there are various valid perspectives on reality. Of course, this does not mean that in science, ‘anything goes’. Data exclude some explanations, more experiments allow to narrow down possible theories. Nevertheless, science always remains a balancing act between constructivism and empiricism. In this CLEA seminar, Karin Verelst will shed a light on this complex balance against the background of the ever-growing realm of interdisciplinary research. She will use Poincaré’s approach to scientific knowledge - the perspectival principle of ‘conventionalism’ - to do justice to the dual nature of science as both invention and discovery without falling into the relativist trap.