A quasi-experimental comparison of no-code visual analytics and Python programming for medical AI literacy

A significant “computational divide” hinders the widespread adoption of AI competencies within the medical workforce, as traditional training approaches, heavily reliant on Python programming, present technical barriers for clinicians without an engineering background. This quasi-experimental study compared a no-code visual analytics platform (KNIME, n = 52) with a historical Python coding cohort ( n = 60), using the Dual Knowledge Mapping Framework (DKMF). KNIME-course outcomes were described using task-completion rates and the LEI; the primary between-cohort comparison used overall LEI. Both cohorts had individual-level overall LEI values calculated using the same AHP weighting scheme; for the historical Python cohort, original individual-level educational records were retrospectively harmonized to the four LEI dimensions using prespecified common scoring rules and conceptually corresponding, tool-specific questionnaire items scored on a 5-point scale. Full measurement equivalence across cohorts was not established. The KNIME group achieved a higher post-course LEI (M = 4.49, SD = 0.72) than the Python group (M = 4.05, SD = 0.25; Welch t = 4.19, p < .001; Cohen’s d = 0.84). Mean task completion was 96.15% in KNIME under the same task-level definition. Because the cohorts were sequential and nonrandomized, the between-group finding is an association and does not establish a causal advantage of KNIME or DKMF.