How Do Humans Manage Conflicts Across Code and Models? A Controlled Scenario-based Study

This is the replication package for a controlled scenario-based case study investigating how humans detect and resolve syntactic conflicts across two software-artifact representations, Java source code and UML diagrams, and whether that detection and resolution behaviour transfers between them. Fifty-four student participants each judged up to 44 matched scenarios (20 Java syntactic, 20 UML syntactic, 4 semantic contrast), spanning conflict, no-conflict, and pseudoconflict situations, presented through 35 randomized form versions. For each scenario, participants classified it as a conflict or not, chose a resolution from a fixed option set, and reported their confidence and perceived difficulty.The package includes the full study protocol, the scenario instrument and informed-consent materials, anonymized raw and processed response data, the complete statistical analysis pipeline (Python), all result tables and figures, plain-language explanations of every statistical method used, and per-research-question written answers with full supporting detail. Five research questions are addressed, covering detection and resolution consistency across ground truth, cross-language agreement, sensitivity to structural cardinality and conflict kind, and the effect of participant experience and sequence position (learning/fatigue). All shared data is anonymized in accordance with the study's informed consent terms and GDPR; no raw or identifiable participant data is included. The analysis pipeline is fully reproducible from the included raw form exports via three entry-point scripts (build_dataset.py, build_extraction.py, build_analysis.py).