Gold plated shovels with rope handles: cognitive issues in display design and a case for performance testing

System performance is often described in terms MIPS, MFLOPS, IO bus speed, etc. However, these dimensions are not very useful in characterizing overall system performance when part of the system is not silicon based (e.g., when there is a human user). Thus in wishing to address overall system performance in man-machine systems we must toss the user into the fray. How should this be done? Often, we just ask users how they like the particular configuration. We have, after all, been inculcated with the dictum, "honor thy user", and it is not far from that pronouncement to, "the user knows best". Thus if the user is happy we can expect optimal system performance. Or can we? This paper focuses on the importance of acknowledging general quirky human behavior in the design and evaluation of computer displays.