Replication package for LLM Support for Discovering Cognitive Biases in Requirements Engineering

Replication package for the paper LLM Support for Discovering Cognitive Biases in Requirements EngineeringAbstract: Cognitive biases, as "errors" or "deviations" in logic that affect decision making, are a relevant concern in the area of Requirements Engineering (RE). Due to their intangible nature, these biases are difficult to detect and observe, as is the case with other cognitive phenomena. Despite their importance, this difficulty explains the limited number of studies that examine cognitive biases in RE. Therefore, a broader overview of biases that are relevant to RE is needed. We started conducting a study to systematically identify and exemplify cognitive biases in RE, and found that the process is resource-intensive, requiring the involvement of multiple experts, and considerable time and cognitive effort. This challenge led us to explore whether a Large Language Model (LLM) can support the task of identifying and exemplifying cognitive bias in RE and alleviate the need for these resources. With two different prompts, we conducted an initial exploration with GPT-5 mini in assessing the relevance of and exemplifying biases in RE activities. The two main contributions of this work, thus, comprise: (1) the introduction of a systematic framework for cognitive bias exploration assisted by LLMs, and (2) the introduction of a framework to assess LLM adequacy in finding cognitive biases in a specific context.