In academic Human-computer interaction, it’s often difficult to publish a systems paper without a quantitative evaluation. What’s so silly about this is that the implementations of these systems are often so unpolished that, as Bret Victor points out, the expected differences in simply doing a well-designed version of the same idea would exceed the measurement variance.
This field is trying too hard to be a science. Perhaps it can become a design science (per Herb Simon), but these evaluations rarely make sense as a natural science.