Abstract
For the purpose of reducing human error, warnings should be designed to match warning users' performance level (Lehto, 1991). However, the performance level on which an individual works depends on the task being performed and the individual's knowledge level. We hypothesized that the use of warnings will result in greater relative reduction of errors for individuals with high knowledge content (HKC) than for individuals with low knowledge content (LKC). An experiment with 40 participants was designed and conducted to test this hypothesis. Results of the experiment partially supported the proposed hypothesis only for high inference tasks but not for low inference tasks. The result validated Lehto's theoretical analysis of when and how warnings can help reduce human error.