A perfect test would, first, try to prevent correlation between questions. It would be a waste to ask 4 questions when 3 suffice to predict the last answer with good accuracy.

After that, some questions supposed to be perfectly correlated with previous are added. These should, however, use different words or concepts altogether, to avoid detection. These questions are meant to establish consistency, and would pick up on randomness

25 strikes me as too few to get into that kind of verification. I seem to remember serious personality tests to usually contain in the range of 100 to 200 questions.

You're on to something. I recently took several, and they contained a lot of questions (100+). I noticed that many of the questions were testing suspiciously similar things, only worded differently.

>It would be a waste to ask 4 questions when 3 suffice to predict the last answer with good accuracy. //

How you ask a question can perturb the answer. If an axiom is that you're determining static elements of personality, then you'd need to ask about the same facets in different ways to avoid skewing the analysis.

"I am quick to react?"

vs

"I take time before reacting to make a measured response"

Both a priori sound positive, so people will tend to "somewhat agree" (as the answer format of these things goes), but they are seemingly antithetical.

Of course they're not really antithetical they're just shallow questions that aren't really going to discern much of your personality. People can react quickly/slowly according to the needs of a situation. Maybe the value is in measuring perceptions of one-another, despite the ease with which we are misguided.

Also, I’m sure many people tend to stray away from extremes.