AI & ML
Theory building in software engineering: Operationalization
arXiv cs.SE0 views
arXiv:2412.02384v2 Announce Type: replace Abstract: This work is part of a research project whose ultimate goal is to systematize a procedure for constructing theories in software engineering. The proposed methodology involves four phases: conceptualization, operationalization, testing, and application. In previous work, we described the conceptualization process. This paper presents a set of procedures for systematizing the operationalization phase in theory building. Specifically, it translates the concepts and propositions obtained from the previous conceptualization into constructs and empirically testable hypotheses. The operationalization phase is structured across three distinct dimensions: the practical procedural steps, their strict formal mathematical specification, and their evaluation through an illustrative empirical case study. Grounded in critical realism, we operationalize the qualitatively derived concepts and propositions into explicit constructs and logical hypotheses. It is important to emphasize that a novel strategy has been designed that reduces the number of hypotheses resulting from this process. In the use case described, more than 30% of the initial hypotheses have been eliminated because they are redundant. This paper is a pioneering contribution in offering comprehensive guidelines for theory operationalization using logical implication and describes a strategy based on graph theory that tackles the problem of hypothesis explosion and maintains the principle of parsimony.
Source:https://arxiv.org/abs/2412.02384PreviousParanoiaEval: Benchmarking Unnecessary Defensive Work in Agentic CodingNextReLoop: Structured Modeling and Behavioral Verification for Reliable LLM-Based Optimization
Related posts
7.9% to 48.1%: LLM rewrote Qwen weights, no training
0dev.to
Spousal Open Work Permits in Canada: Who Qualifies in 2026 and What to Prepare
0dev.to
Reading ingredient lists from phone photos: a crop bug, a model bench, and why the AI never gives the verdict
0dev.to
Edital exige guardrail
0dev.to
Convert an SVG to a PNG in the Browser, Without Uploading the File
0dev.to
Generative Simulation Benchmarking for planetary geology survey missions for extreme data sparsity scenarios
0dev.to