FWIW - I studied at the first 102 lines of Shog's title filter data before getting bored. In those, I count (in my obviously subjective judgement):
20 questions posted with improved titles
16 questions posted with worse titles
16 questions posted with titles no better nor worse than before
8 crap-sounding questions abandoned
0 good-sounding questions abandoned
I guess you can count that as the title filter being a net force for good or evil depending upon how important you consider each of those outcomes to be. I remain anti.