Simon Thorne wrote this interesting and thought-provoking piece about catastrophic consequences of human errors associated with spreadsheets. Some of my earliest research and publications were in this area of ergonomics, and I am still interested in software quality. Here are some thoughts:
First of all, it is possible we can apply the ideas of large language models and generative AI to assuring quality in a few areas. We can train, fine-tune, and task the models to search for the most-common issues that violate implicit and explicit requirements and expectations of the document, spreadsheet data, or software output. Writing software tests is difficult and time consuming. Generative AI models already write manu of my unit tests. We can expand their use for negative tests in software, spreadsheet checkers, and document scanners.
Secondly, the engineering quality assurance (building quality in) and quality control (measuring quality) that has moved from older engineering disciplines such as civil engineering, construction, manufacturing, and software engineering can be applied to clerical knowledge work. The magical "second pair of eyes" that checks work from accounting through code reviews should be applied to spreadsheets, documents, and knowledge work artifacts.
It is possible my cognitive bias to find patterns where none exist is suggesting an idea that is infeasible. Or perhaps these concepts can be valuable. What are your thoughts?
No comments:
Post a Comment