I suspect this test gives central values to many test-takers because the internal consistency of the questions has not been verified. Anyone can come up with t

I suspect this test gives central values to many test-takers because the internal consistency of the questions has not been verified. Anyone can come up with ten questions and claim "these measure a programmer's fondness for abstraction". Unfortunately, we cannot verify this claim. What we can do is check statistically if the answers to the ten questions correlate strongly enough that we think they measure the same thing, whatever that is. This involves a trial run of the quiz followed by statistics such as Cronbach's alpha and principal componoment analysis. Often one finds that the questions do not measure the same thing at all! It is difficult to come up with a set of internally consistent questions. When we add together inconsistent questions (as is done here to get the final score), we don't get stronger signal: the noise from the uncorrelated questions cancels out and almost everyone gets a result near zero. I know OP is not the author, but if the author reads this: if you're lucky, the data from this audience mey reveal a subset of two or three questions per dimension that are internally consistent. Use only those in the quiz going forward! Then you can trial more questions one at a time and keep only those that are consistent with the first set. (Is this not p-hacking? Yes but in this context who cares.)