We explore why optimizing for easy metrics like interface polish often masks foundational flaws in research rigor and ethical alignment. Drawing on Raluca Budiu’s warning that bad data is expensive and Zeeshan Khalid’s distinction between usability and accessibility, we argue that high scores can’t fix exclusion or invalid participant selection. The conversation extends to the community’s shift from AI hype to practical utility, highlighting how documenting pushback against dark patterns matters more than iterative tweaks when business goals contradict user trust.
