Skip to content

Features

Source Alert : Teacher's perspective

Why is it cool?

In the AI-era it became very easy for students to miss out on the learning goals of academic writing, by simply letting AI write a significant amount of text for you. Our source alerts on the one hand help the student to avoid unintentional misconduct. On the other hand, the data from the writing pattern analysis can be reviewed thoroughly, if there is doubt, regarding the integrity of the student’s work, and provide an objective data basis to come to a decision in that case.

Source alerts are notifications you will receive when our writing pattern recognition algorithm recognizes a pattern suggesting copied or copy-typed text in your document. You can review them at different places within the app (e.g. activity stream), to classify them accordingly as one of the following: Citation, My Text, AI-generated, or AI-revised.

  • Usability
    • Reviewing the source alerts in detail is a ‘last resort’, when the suspicion is already significant. First, have a look at what the sources were classified as. The most relevant classification regarding misconduct is ‘My Text’. If a student never uses this one but rather always gives a source for what was being copied into their document, then it is very unlikely to have cheated.
    • For each source alert, explain how the teacher can “work“ with it.
    • Source alerts categorized as citation: if the citation was automatically detected, the details section will contain the corresponding reference from the document, e.g. the footnote contents. This should be verified for consistency and sufficiency. In addition the student might have added a comment to explain the source.
      if the citation reference has not been automatically detected we recommend to verify if there is a valid reference in the document – to do this you can inspect the in-text report.
    • Source alerts categorized as AI generated: the details section should give you more information on the prompt(s) used
    • Source alerts categorized as AI revised: if the details reference the prompt history you can see what text was there already before the revision in question. if the student did not add the prompt history you should check the in-text report to see what fraction of the text section was already there – see in-text example below 
    • Source alerts categorized as “own work”: the student marked this as their own work, but the creation history could not be recorded by Mentafy. Therefore, consider whether this categorization is credible, taking into account the importance, frequency, and overall length of such sections, as well as the student’s explanation in the comments section.
    • Unresolved source alerts indicate that an unusual writing pattern was found, but it could not be assigned to a category automatically and also the author did not classify it. In such a case you should review the corresponding text to decide its relevance for evaluation.
  • Technical
    • Our writing pattern algorithm analyzes the text, that is entered in the main document. Multiple dimensions of this writing pattern are considered to decide, whether a text was copied or originated directly from the author’s mind, like the speed of the input, the number of recursions, how many corrections have taken place, or the pattern of writing breaks. Based on the interplay of those variables, we will raise a source alert, i.e. mark the text passage that does not look like it was written by the author directly but rather copied from another source.

© 2024 Mentafy. All Rights Reserved.