Some good thoughts and analysis here, but I think your experiment design might need a little more tweaking.

One example, the image choice on the people vs objects experiment may have introduced bias unnecessarily because some of the “object” photos have body parts, plus the diversity of people is very limited in such a small sample. Perhaps, try a bigger sample size, more randomization of photo subjects, and/or photos of identical scenes with and without people?

Much of what a viewer finds enticing in a photograph is subjective, and it’s difficult, if not impossible, to control for overall interestingness without a massive sample size of diverse photos.

There are other issues with the experiment, but trying not to troll here. Good writing.