Detecting Plagiarism

Using Technology to Detect Plagiarism

Three major approaches to detecting plagiarism have been recommended: One is the use of plagiarism detection programs such as Turnitin.  This site compares the submitted student writing sample to a database of existing content, including web pages, articles, books, and other uploaded files. It then generates a similarity report that identify matching text and provide links to the original sources. Instructors are warned, however, that “the similarity index alone cannot determine plagiarism. Turnitin cannot detect if a student has included the appropriate references for the text that is identified as similar to some other document. A paper with a high similarity rating needs to be examined to determine if plagiarism has occurred.” Guidelines for Using Turnitin

The University of Kansas Center for Teaching Excellence urges caution when using Turnitin and lists the following issues to consider:

Use results, which indicate similarity between a student’s paper and papers in the database, as information, not proof if plagiarism. Flagged material means that similarities were found, not necessarily that the student has plagiarized. Flagged material may include quotes, citations, and common descriptions that have been appropriately used. Therefore, to make the best use of the results, instructors are urged to do the following:

  • Talk with students. If a paper comes back with a high number of flagged items, talk with students about the results and your expectations for academic integrity and writing. Give them an opportunity to learn from their mistakes and make necessary changes.
  • Avoid definitive cutoffs. Setting a rigid cutoff (e.g., 10% or 15%) fails to consider the weaknesses in plagiarism detectors. A better approach would be to set ranges of percentages that lead to various actions by students, such as
    • Meeting with the instructor and discussing the assignment and creating a plan for revising it
    • Reviewing the plagiarism detector report, making revisions, and attaching a brief summary of what they did
    • Discussing the reports in groups in class, looking for similarly flagged work and providing a summary of possible revisions to the instructor

Dig Deeper Into Similarity Reports. Learn about the details of Turnitin and how to adjust many of the settings. Possibilities include the following:

  • Exclude bibliographies and references. If students use proper citation style, these will always be flagged.
  • Adjust the number of consecutive words checked. Using a smaller number (e.g., five) will lead to more material being flagged.
  • Adjust for names. A student’s name will be flagged if a student has turned in a paper before, as will any name in previously submitted papers.
  • Take students’ previous submissions into account. Once a student paper is submitted to a database, that paper becomes part of the database; therefor, the paper that is resubmitted is very likely to have a large percentage of material flagged by a plagiarism checker.

Help Students Understand Disciplinary Conventions and Plagiarism Reports. Talk with students about what plagiarism detectors do and how they work. (Turnitin has good explanatory material on interpreting its reports.) Talk with students about your expectations and about how you will interpret the reports from Turnitin.

  • Establish common understanding. Instructors should help students understand what plagiarism is and why they are trying to prevent it.
  • Work with students on citation. Explain what common knowledge and what a unique idea is.
  • Work with students on citation style. Make it clear to students what they need to include in citations and the particular citation style you want them to use.
    What to consider when using a plagiarism detector

Additionally, Turnitin reports can be used teach students about the proper use of quotation marks, how to cite sources, and how to paraphrase. Another way to use Turnitin as a teaching tool is to allow students to submit their own work and view their similarity reports, which will allow them to check their work and correct and resubmit it, if necessary.

Sources

Carnegie Mellon University Eberly Center for Teaching Excellence and Educational Innovation
Guidelines for Using Turnitin

Canvas at Yale
Turnitin Overview

University of Waterloo Centre for Teaching Excellence
Using Turnitin in Your Class

Cornell University Center for Teaching Innovation
Learn to use Turnitin

University of Kansas Center for Teaching Excellence
What to consider when using a plagiarism detector

Stanford University Teaching Commons
Guidance on technology tools for academic integrity

A second approach to detecting plagiarism is to do a Google search on a specific group of words or a phrase common in the document being checked to determine whether the search identifies sources in which these words have been used. Specifically, an instructor can find a short phrase that appears in the paper or paragraph and perform an exact phrase search by typing the phrase into the Google search window and surrounding it with quotation marks. If the paper is available online, it will usually come up in the search.
Why Students Plagiarize

Identifying Signs of Plagiarism

A third approach is to learn to identify the signs of plagiarism in student writing. These include the following:

  • Lack of references or citations
  • Citations of obscure sources
  • Citations of sources in another language
  • Reference to outdated or very old material
  • Reference to unpublished material that is unavailable
  • Reference to websites that are no longer available
  • Mixed citation styles (e.g., MLA, APA, and Chicago)
  • Inconsistent margins and fonts
  • Unusual formatting (skewed tables, broken lines, mixed subheading styles)
  • Uneven writing style (some very sophisticated sentences and some very simple ones)
  • Abrupt changes in tone, tense, style, grammar, or sentence structure
  • Vocabulary or word usage that is not usual for the student writer
  • Papers that are slightly off topic or do not follow instructions
  • Last-minute requests to write on a different assignment or topic

If they suspect plagiarism, instructors can try talking to the student about the assignment in order to learn about their process and determine their familiarity with its content. Plagiarism is more likely if the student is unfamiliar with the sources or information in the paper, cannot summarize the main points in the paper, or cannot answer specific questions about the content of the paper or their research and writing process.

Sources

Massachusetts Institute of Technology Writing
Resources for Teachers: How to Detect Plagiarism

University of Illinois Center for Innovation in Teaching and Learning
Discouraging and Detecting Plagiarism

Westmont College Psychology
Reducing and Detecting Plagiarism

University of California/Berkeley Graduate Student Instructor Teaching & Resource Center
Detecting and Addressing Plagiarism

Generative AI Detection Tools

With the widespread use of Generative AI tools, instructors have become concerned that students will use them to respond to assignments and then submit the output as their own. This concern has been accompanied by the development of tools to detect whether a written assignment was generated by AI. These tools (e.g., Turnitin) scan the submitted text and identify what percentage of that text was generated by AI. There is some controversy, however, about the effectiveness and utility of these tools and whether they should be used at all. Thus, some institutions of higher education do not support the use of Gen AI detection tools. For example,

The Cornell University Center for Teaching Innovation indicated that “We currently do not recommend using current automatic detection algorithms for academic integrity violations using generative AI, given their unreliability and current inability to provide definitive evidence of violations. We believe that establishing trusting relationships with students and designing authentic assessments will likely be far more effective than policing students.” Detecting AI Generated Content

The Duke Learning Innovation & Lifetime Education offered the following statement about the use of AI detection software: “We don’t recommend AI detection software as a part of your AI policy for three main reasons.

  1. The products are unreliable. Research on AI detection softwarefrom MIT highlights the false positive and negative rates. Open AI (the company behind ChatGPT) withdrew its own detection software due to the software’s unreliability.
  2. Detection software is biased against certain segments of learners, for example non-native speakers, as research from Stanfordshows.
  3. As AI changes, detection software cannot keep up.”
    Artificial Intelligence Policies

Academic Technology Solutions at the University of Chicago compared four detection tools against both AI-generated and human-composed text. Their findings suggested that no tool is infallible. “Each had one or more vulnerabilities that could pose problems when using it to detect possible student misconduct–either by failing to detect AI-generated text with certainty, or by falsely flagging human-composed text as AI-generated.” They concluded that instructors should not rely solely on results generated by AI detection software to determine whether a student has committed academic misconduct; if detection software is used, it should be in conjunction with other evidence. AI and Plagiarism Detection Software

According to the University of Kansas Center for Teaching Excellence the best approach to use the using the results generated by Turnitin is to collect additional information by (a) comparing the student’s work that was flagged to previous work, looking at style, tone, flow, complexity, and use of citations; (b) trying another detection tool and comparing results; (c) talking to the student about the problematic findings; and (c) giving the student an opportunity to redo and resubmit the assignment. Why you should use caution with AI detectors

Identifying Signs of Plagiarism Due to Gen AI Use

As indicated, some sources have advised that the best way to use reports from AI detection software is to combine the results with instructor evaluations of students’ work. Thus, instructors can use their own judgement to detect material that has been generated by AI and either combine it with evidence from AI detection tools or use it independently. Possible clues that Gen AI has been used to generate text include the following:

  • Lack of fluency, word repetition
  • Poor or inappropriate punctuation
  • Short, less complex sentences
  • Grammatical, syntax, and spelling errors
  • Writing that is too perfect, with no typos or errors
  • Hallucinations, logical inconsistencies, irrelevant or factually incorrect information
  • Tone is flat and writing sounds formulaic
  • Lack of creativity or originality
  • Approach to assignment that is significantly different than that presented in course
  • Repetitive use of “the,” “it” or “is”
  • Strange metaphors
  • Sources that are obscure or unavailable to students
  • Bibliographic entries that don’t make sense, sometimes with problematic links
  • Style is very different from previous writing for course
  • Fake or dead-end links

Sources

Cornell University Center to Teaching Innovation
Detecting AI Generated Content

Duke University Learning Innovation & Lifetime Education
Artificial Intelligence Policies

University of Chicago Academic Integrity Solutions
AI and Plagiarism Detection Software

University of British Columbia UBC Guidance
Guidelines for All Uses of GenAI in Teaching & Learning

Carnegie Mellon University Eberly Center for Teaching Excellence & Educational Innovation
Generative AI Tools FAQ

University of Missouri Academic Technology
Detecting AI Plagiarism

Northern Illinois University Center for Innovative Teaching and Learning
AI Detectors: An Ethical Minefield

University of Kansas Center for Teaching Excellence
Why you should use caution with AI detectors