ChatGPT possesses the ability to identify potential causes of bugs in a program, which can be used for fault diagnosis. Although ChatGPT cannot always provide accurate responses, it can provide the confidence level that is trained to be correlated with the accuracy of the response, and the confidence level helps verify the accuracy of responses. Through preliminary trials, we found that the explanatory power of potential causes and the confidence level became low even for accurate responses when runtime information is needed to identify the causes of bugs. In this study, we propose a method to construct prompts based on the target program and its execution traces to improve the accuracy of fault diagnosis by ChatGPT. Through case studies using five bugs from Defects4J, we obtained the following two results: (1) For four bugs, the explanatory power of the responses to potential bug causes improved using the information contained in the execution traces. (2) For three bugs, the confidence level was higher for the accurate responses when execution traces were available than when they were not. These results suggest that by using program execution traces in prompts, the accuracy of fault diagnosis by ChatGPT can be improved.

Biosignals reflect the mental states of software developers and could improve support technologies for software development activities.Although several technologies for software development support using biosignals (BioSDS) have been proposed, BioSDS has not yet been deployed in actual software development workplaces.As a prerequisite for industrial deployment, BioSDS must be well understood and accepted by software developers.However, the current level of their acceptance has not been comprehensively assessed.In this study, we conducted a survey to clarify the current level of acceptance of BioSDS and potential attributes that influence the level of acceptance.We defined eleven use-cases based on six previous primary studies related to BioSDS, and then asked developers at Hitachi, a Japanese IT company in the FORTUNE 500, about the level of acceptance of each use-case.Our analysis of eighty-six responses revealed that four out of eleven use-cases had some level of acceptance by software developers.In addition, we found four attributes that affect the level of acceptance: subject to be measured, objectives, interventions, and timing.These findings help to identify barriers to the adoption of BioSDS in the workplace.

Minimizing the risk of harm to stakeholders is required when incorporating AI tasks into business processes (BPs). Oneway to reduce the risk is to redundantly incorporate both of AI and human tasks. Multiple BPs incorporating them can be considered and it is not obvious which BP can minimize the risk. In this study, we propose a risk assessment methodto design BPs incorporating AI tasks following the ISO guideline that recommends managing the severity and likelihood of each risk source. A case study showed that the proposed method might be useful for designing BPs incorporating AI tasks.

Descriptions of program identifiers improve the maintainability of programs. Modern software projects maintain proper descriptions by following coding conventions. However, software projects maintained for a long time have two problems: (i) descriptions at incorrect locations and (ii) no descriptions. We propose the method of generating a identifier dictionary for managing identifiers and their descriptions, which enables developers to refer to identifier descriptions from anywhere within programs. The method involves two steps: (i) extracting identifiers and descriptions from design documents and programs and (ii) generating descriptions using information-retrieval and machine-learning methods. We applied the proposed method to COBOL programs and design documents of a legacy system that has been maintained for over 20 years as a case study. The proposed method obtained the descriptions of 83% of identifiers and reduced the cost of locating files to be modified by enhancing search keywords using the identifier dictionary. This means that the proposed method can improve the maintainability of systems maintained over many years.

Symbolic execution can automatically generate test input values that cover the execution paths of programs. It enables us to test functions of even huge COBOL legacy programs, but the execution time substantially increases when the number of instructions in the input program is large. In this research, we propose a method that removes instructions that do not affect execution paths using program slicing. Applying the method to three functions in COBOL programs reduced the execution time by up to 70% by removing instructions that manipulate variables not related to the execution paths. This result suggests that our method enables symbolic execution to generate input values even from huge COBOL programs.