CogNetEval: An interactive cognition-informed readability evaluation tool for network layouts

Network diagrams are widely used to explore relational data, yet assessing whether a layout supports user cognition and refining problematic regions without disrupting the surrounding spatial context remain challenging. We present CogNetEval, an interactive tool that integrates network rendering, cognition-informed readability evaluation, diagnostic heatmaps, and constrained clique transformation. CogNetEval maps candidate cliques of four to six nodes toward regular polygons and selectively commits transformations according to explicit criteria for regularity, displacement, crossings, and node-edge occlusion. Across 20 paired Fruchterman–Reingold layouts on six networks, the method accepted 18.8%–45.4% of candidates, increased regularity on accepted cliques, and kept mean displacement below one-thousandth of the layout diagonal. The hard node-edge-occlusion constraint was satisfied in all runs, while crossings decreased empirically across all six networks. A perceived-utility study with 15 participants found the diagnostic heatmaps useful. A targeted behavioral experiment ( N = 12 ) found no statistically significant overall differences; an exploratory analysis of the more complex Dolphin stimuli showed higher path-tracing accuracy after transformation ( d z = 0.74 ). CogNetEval provides an evaluate-then-refine workflow for selectively improving local clique structure while controlling global layout disturbance.