Bug Blindness: Why Developers Stop Noticing Software Flaws
Dan Luu argues that users hit the same software bugs without registering them. His new essay draws on search-engine spam, Volvo forums, Blackboard and Discourse to explain quality blindness.
Programmer and blogger Dan Luu has published an essay titled "Bug blindness", arguing that the reason he runs into far more software defects than most people is not an unusual way of using computers, but that other users hit the same bugs and do not register them.
Same bugs, different perception
Luu writes that he typically observes hundreds to thousands of bugs per week and that "nothing seems to work" — a theme he has documented before on his blog. For a long time he assumed this had to do with how he uses computers, but concluded that "people are hitting the same bugs and don't notice". Curing quality and bug blindness, he argues, is useful for programmers: after a few weeks of having bugs pointed out, receptive colleagues tend to start noticing them on their own.
From search spam to course software
The essay collects cases where a product's quality did not change while perception of it did. In a study of web search, Google, Bing and Kagi returned results full of low-quality SEO spam and, in some cases, outright scams; some readers disputed his description of Kagi, but the results they sent back contained no useful result either.
Another example is Volvo forums: reliability data has been mediocre to poor for well over a decade, while forum members insist the cars are among the most reliable. The best-known case is Blackboard, the course management software that Wikipedia describes as "one of the most disliked — even detested — companies in education" and whose Amplicate survey respondents, as reported by Fast Company in December 2011, were 93% "hate". Yet a Blackboard employee he met believed users loved the product. In the case of forum software Discourse, developers shipped code that deliberately slowed real page loads in order to look better on web performance metrics such as LCP — gaming the benchmark at the user's expense.
Habits that hide the flaws
Luu describes "habitual mitigations" — behaviours people develop unconsciously to work around defects. He compensated for a grimy mechanical mouse ball with countervailing hand movements; in Google Docs he built a specific order of operations around a delay that overwrites an immediately typed title; at Microsoft he switched his laptop's WiFi off before logging in to bypass an authentication error. A large part of computer literacy, he argues, is a library of such habits, which also blunts the value of dogfooding your own software.
Why it matters now
Teams with a high degree of quality blindness ship products with reduced or no chance of success because of quality issues, he writes — while everyone involved believes the thing is high quality. This has become more important with coding agents: it is easier than ever to churn out low-quality software, but also easier than ever to improve performance and reduce bugs, provided people notice that quality can be improved. Readers in the comments pointed to the German word "Betriebsblindheit"; an appendix discusses an analogous blindness about advertising.
SiTech — AI-powered web development
We build fast, modern websites and bring AI into real business workflows. Have a project or a question? We'd love to help.