Toward responsible scientometrics: a lifecycle framework of normative data practices for research evaluation | Scientometrics | Springer Nature Link

Abstract:  Responsible scientometrics has become a central concern in the context of increasing reliance on bibliometric indicators for research evaluation and policy decision-making. While existing frameworks such as DORA, the Leiden Manifesto, and CoARA emphasize principles of responsible metrics, less attention has been paid to the underlying data practices that condition the validity, transparency, and reproducibility of scientometric studies. This paper argues that responsible scientometrics cannot be achieved solely through improved indicators or evaluation principles; it also depends critically on the normative structuring of data practices across the entire research lifecycle. We propose a lifecycle-based framework that conceptualizes scientometric research as a sequence of interdependent data practices, including data source selection, data retrieval and acquisition, data preprocessing and standardization, data analysis and visualization, data interpretation, data storage and documentation, and data sharing and citation. For each stage, we identify common methodological risks and propose normative guidelines aimed at improving transparency, reproducibility, and contextual validity. By integrating methodological considerations with principles of responsible research assessment, the framework highlights how data practices are not merely technical procedures but constitutive elements of evaluation outcomes. The proposed framework contributes to scientometrics in three ways: (1) it systematizes scattered methodological recommendations into a coherent lifecycle model; (2) it extends responsible metrics discussions by foregrounding data-level normativity; and (3) it provides a practical reference for researchers, reviewers, and policy stakeholders engaged in research evaluation. Ultimately, this work aims to strengthen the methodological foundations of scientometrics and support more transparent, accountable, and context-sensitive evaluation systems.