Machine learning based analysis of gender differences in visual inspection decision making

While machine learning is most often concerned with learning from humans, the fact that human behavior systematically differs for (groups of) people with different gender, age, education or cultural background is widely ignored. Obviously, such differences are reflected in the training humans provide to machine learning algorithms that in turn affects the induced models. A coherent set of experiment design and analysis methods is presented which was applied for studying gender differences in visual inspection decision making. Detailed results from a study with 50 female and 50 male subjects are reported. Although immediate performance measures were almost equal, highly significant differences in the structure of induced decision trees have been found (p = 0.00005). This demonstrates the value of our contribution for researchers intending to investigate the otherwise hidden structure of cognitive gender differences rather than their merits.