[ARCHIVED] Turnitin Recalculation of Similarity Score

[shime]

Hopefully, this is the right place to ask this question. We have an issue where a student turns in an assignment and receives a similarity score from the Turnitin LTI in Canvas of 0.0%. If another student then turns in the same (or very similar) assignment at a later date, their similarity score (for example 98%) is calculated based on the similarity to the previous student's submission. However, shouldn't the Turnitin LTI recalculate the previous student's similarity score (say from 0.0% to 98%)? With a small class it might not be hard to find the duplicate submission. But, with a large class, you would have to go through each assignment (assuming the submission was even in the same class) to find the duplicate.

[nwilson7]

@shime When you click on the details for the score, doesn't it tell you where the similarity came from?  It should have a list of where that content matched and if it is content from the web, it provides links and if it is other students it should tell you that.  

I see what you are saying but technically the first student may not have plagiarized (although they may have been in cahoots with the student that did) as when they submitted, there were no other papers like theirs.  If somehow a student copied my work (like maybe they peer reviewed mine so they knew what I wrote) and I was unaware, I would hate to be blamed for plagiarizing.  I believe that is how Turnitin functions.  There would also need to have a time frame for your idea to work correctly such as within the same assignment or something.  Otherwise, does it go back to a submission for 5 years ago and mark that now as 98% similar since something new matched it? (playing devils advocate, not disagreeing with your idea).  

This would be something to reach out to Turnitin about since it is their software doing all the steps.  I could see this being helpful if it matched submissions in the same assignment.

-Nick

[shime]

When I click on the similarity score, it shows the similarity score % but does not show where it came from. I also didn't have the radio button next to "Show source text" in our Turnitin admin settings. That may resolve this issue. And, you mention exactly the scenario I was thinking of in that you don't necessarily know which student plagiarized. Turnitin just assumes the first submission is the original, when it may not be. We have had instances where a student gets another student's password, logs into said student's Google Drive and copies that student's work and submits it as their own. The student's work that was copied is not at fault, but it's easy to determine where the plagiarized work came from if it's in Google Drive. If it's not in Google Drive, then you have to determine who wrote what first. Not asking Turnitin to do that, but knowing which papers match (regardless of time of submission) would be helpful.

Shawn

[chriscas]

Hi @shime,

With my general knowledge of plagiarism detection (my institution has been a turnitin/vericite customer for almost 20 years now), I can pretty confidently say that I haven't seen any systems that work as you describe.  In pretty much all systems I know of, the first time something appears in the the database (by web crawling, journal ingest, student submissions, etc), it's considered the original source.  Subsequent submissions that match text are marked as similar to that original.  You do bring up an interesting point about student submissions from the same course though...  I could see where it may be handy in that scenario to mark all similar submissions.  I think there would be some many technical challenges going that route though...  One, the checking and reporting back to Canvas is generally done at the time of submission. Changing that would mean turnitin would have to store a lot more information about each submission for a much longer period of time so it could attempt to change similarity scores later.  Then there would be a question about how narrowly or broadly to apply the new logic.  Just for submissions from the same exact assignment in one course, for all submissions in a course, for submissions across different courses at the same institution, etc...  In my mind, anything but the most narrow cases there would seem to open up even more issues.

All that being said, if the main issue is figuring out where some particular matched text on a submission came form, your Turnitin administrator should have some control over the info released to instructors.  They can choose to release the text and submission details for entries in your institutional database of submissions, which would allow you to identify where the match came from (screenshot below, we have the most basic Turnitin version, so I believe every Turnitin customer should have these options or more). 

Hope this helps a bit!

-Chris 

[shime]

I see your point. However, I will say, ultimately, Turnitin is assuming the first submission is the original. If one student copies another students work, then submits it before the first student, Turnitin would assume the copied work were the original. I also had not selected the radio button next to "Show source text" in the Source Text Visibility category in our Turnitin admin page. Perhaps that may resolve our issue. Thank you!

Shawn

[nwilson7]

@shime You make very valid points!  I currently work in higher-ed but spent 12 years in K-12 and we had issues with people sharing password or forgetting to log out on a shared computer.  Not sure we had any plagiarizing, our students just liked to be malicious and either delete the other students work or upload inappropriate content to the other students' google drive.  I was the Google Domain Administrator so most of that aways came to me in hopes I could track down who did what, so I feel your pain!

-Nick