I'm involved in TPC (technical program committee) task for some international conference. In one paper I'm reviewing, I noticed that the similarity index is just 2% using iThenticate which seems impractical (for me). Though the type/context of the paper is an essay paper, discussing current status and challenges of some phenomenon/technology/topic, there like no graphs and discussion of numerical/simulation results. I'm suspecting whether the authors are trying to cheat somehow? Any tips how to proceed since this is my first TPC task?
I don't think I've ever seen an iThenticate paper with 2% similarity. Less than 10% is rare enough.
However: so what? Low similarity isn't a problem, only high similarity is. The latter implies plagiarism, but the former ... I can't think of any way the authors can cheat by having low similarity. If it's a fake paper generated by SciGen then perhaps it would have abnormally low similarity, but you've already ruled that out since you understood the paper. It's also presumably possible to force a paper to have low similarity as well, simply by continuously rewriting the parts that iThenticate flags.
I'd assess the paper on its content and not worry about the similarity.
A high portion of the similarities usually come from the reference list, because most of the references in a new paper will already have appeared in the reference lists of earlier papers.
For the paper you are looking at, it may be that either the reference list is extremely short, or they only cite papers that have not been cited anywhere else yet. Both would be rather suspicious for the type of essay paper you are describing.
When reviewing this paper, have an extra close look at the selection of references. Do the authors really cover the state of the art? Do they give enough and appropriate references for the things they are discussing?