Construct clarity has a considerable influence on research practice. The Altmetrics Manifesto promised to broaden the notion of impact. However, reliance on citation theory has limited this promise, confining altmetrics to bibliometric principles by treating interactions with scholarly outputs that occur outside the norms and values that rule science through the same lens as citations. This study argues that this constitutes a form of ontological misalignment (fallacy of reification), whereby altmetrics are assigned entitative properties when they in fact capture processual aspects of engagement and use that may precede or contribute to societal impact. This misalignment is manifested as use of counts of social media and online interactions with scholarly outputs as direct indicators of societal impact. Aligning ontology and epistemology requires the redefinition of altmetrics as process-based indicators. We describe recent empirical developments applying interaction and networked approaches to translate this perspective into practice. We argue also that calls for production of socially relevant knowledge and increased interest in science communication represent a timely opportunity for use of altmetrics as process-oriented monitoring tools. To study societal impact, we suggest their use in a mixed-methods research design to identify engagement patterns that can be further explored using qualitative methods.