If they were BINARY duplicates, I might trust something 'Automatic', because there is no 'is best' decision.
Then there are the 'version' duplicates. (I will use Baen for this example)
The Free Library has the Authors (dead tree?) Original
Baen, re-released a 'recent edited' version (fixed typos?). FD finds these
Now days , Baen is releasing 2nd editions (of the original e-book). Completely re-edited (made more current, PC?)
No way, do I want these kind gone
BTW
Checking (viewing) duplicates was the reason I requested the 'reuse viewer' option. I keep the viewer (placed) on my second monitor, where I can see what the results (tap V) for each book that FD found. Only then, do I delete 'true dups'