Similarity detection

not plagiarism detection
I’m in the process of looking at the role that similarity detection tools (e.g. Turnitin and SafeAssign) can play in helping students improve their writing skills, and detect their own errors. My personal experience is that it’s a valuable tool – as long as you spend enough time explaining to students what should be similar, what shouldn’t and thus what to do about it. All the research I’ve found would seem to suggest the same (though from the work that Lynn Graham-Matheson & Simon Starr (2013) did the students were far more likely to think that the staff saw it primarily as Police force, than the staff thought they did). Morris (2015) suggests that could be the language used when staff introduced it.
What I’m looking for now, though, are any studies that counter this view? Has anyone got any work that suggests similarity tools are best used as a plagiarism police force?