Tagging in Assisted Tracing

Abstract

Assisted tracing is the process where human analyst vets and makes decisions concerning the automated method's output. Current research reveals human fallibility in this process, and shows that analyst often makes incorrect decisions that lead to inaccurate final trace matrix. To help enhance analyst performance, we leverage tagging in assisted tracing. Specifically, we implement tagging as a front-end feature that allows analysts to freely mark what they feel worth externalizing during tracing. We then carry out an experiment to investigate the tagging practices of 28 student analysts in vetting requirements-to-source-code trace matrices. Our study shows that tagging is readily adopted by analysts, tags produced in tracing follow power laws, and tags greatly enhance the precision of analyst-submitted final trace matrices. Our work opens up new avenues for researching improved ways to foster analyst-tool integration.