Technical benchmarking standardizes evaluation for model performance comparison, often overlooking interpretive processes of meaning-making in digital humanities exploratory research. Digital humanists engage in computational strategies to explore corpus structure and uncover thematic patterns through an iterative research design of testing different methodologies towards the goal of meaning-making. We conceptualize this form of research as exploratory computation and argue that this research design process can be clarified and communicated through transparent evaluation. This discussion article proposes a framework that qualitatively evaluates the effectiveness of a computational methodology for exploratory tasks by structuring self-reflection on positionality, task evaluation, and retrospection. Through a case study application of the framework on exploratory data visualization, we demonstrate its conceptual and practical utility for early-stage collaborative evaluation of research design. By making legible processes of humanistic interpretation, this article advocates for centering context and pedagogical transparency in computational research design and benchmarking.
Nguyen et al. (Thu,) studied this question.