UTU Instructions for Turnitin

Turnitin systems are used through the Moodle platform as additional features. Ordering a Moodle workspace with Turnitin Feedback Studio assignments can be done in three ways. A fourth option is adding Turnitin assignments to a workspace already in use. This option is suitable for both Feedback Studio and Integrity.

1. The primary recommendation is to make an order through the PEPPI system. When making an order for a teaching implementation in PEPPI, the workspace automatically receives the name and abbreviation in the teaching information.

• From the Learning Environments tab of the PEPPI system, you can create a Moodle workspace for yourself. The prerequisite is that you have teacher or planner rights in Peppi.
• In the desktop view, select "Add Learning Platform" and "UTU Moodle". Click "Save" to create the workspace. With this command, the Moodle workspace is generated and Peppi immediately adds teachers and already approved students to the course, except for participants who have not previously logged in to UTU-Moodle and are therefore missing from the Moodle user database.

• Make sure that the correct unit information is selected on the "Responsibilities and Guidance Areas" tab of the Peppi implementation - without it, the course will not be transferred to Moodle.

• Do not change the generated short name for the course! It is a unique link between Peppi and Moodle. If it is changed for some reason, restore it to match the abbreviation in Peppi so that the data transfer integration works.

• As students are approved for the Moodle course after its creation, student and teacher information will be updated on the Moodle course participants list the next day. Students will see the course on their "My courses" list.

2. Another way is to order a Moodle-Turnitin workspace through Request for a new course -link. In the order details, mark the course name, abbreviation, category, and additional information as the rationale for ordering a course area with Turnitin.

• As the course category, choose the faculty/department responsible for the course or where you work.
• For the course abbreviation, use the Peppi identifier or other suitable abbreviation for the course. The same abbreviation can only be used on one Moodle course. The year suffix related to the course name is a functional separator for otherwise similarly named courses.
• Your reason for ordering a workspace may be supervising a thesis seminar or checking theses, or another form of work that requires Turnitin checks.
• The order form warns if any mandatory information about the course order is missing. After sending the order, you will receive an email notification when your workspace order has been approved.
• If you want your students to find the course on their My courses list, add them to the Participants list of the workspace. Alternatively, you can send the course address to students by email and ask them to log in to the workspace.

3. The third way is related to workspaces that are ordered for special needs. For example, interdisciplinary courses required for reviewing the research plans of those applying for postgraduate studies are ordered directly from the Turnitin system's administrator or via helpdesk@utu.fi.

4. The fourth way to use Turnitin is to open a Feedback Studio assignment on an existing Moodle workspace site using the "Add an activity" function. Alternatively, you can open the Integrity Assistant for Moodle Assignment, Forum, or Workshop activities.

Detailed instructions with screenshots for options 1 and 2 can be found in the Moodle Teacher's Guide. Option 3 requires contacting Turnitin support/helpdesk. Instructions for activating and editing Turnitin assignments on your own for option 4 can be found in chapter 4 of this guide.

Creating a learning and assessment environment begins with selecting the content plan and layout solution for the workspace.

It is recommended to use the ready-made templates containing Turnitin's Feedback Studio assignments for all theses checks. You can also modify course templates from them for other teaching.

Even in basic use of the templates, it is useful to include content that guides scientific writing in accordance with RCR in addition to intermediate and final checking assignments.

The assignment calendars of the Turnitin assignments in the templates must have the same period of validity as the Moodle course area in order to avoid functional disruptions. The types of Turnitin templates are:

If the life cycle of the guiding materials is several years, the content can be transferred from the old course to the new one using the copy function. 

• How to copy contents from the old course to a new course.

It is useful to use Turnitin in all stages of the writing process!

It can identify problematic elements not only in your own writing but also in the used sources, and allows for timely response to these challenges. Turnitin can also help you find new relevant sources.

If the Turnitin report suggests that a source you have used may be plagiarized, either remove it from the text or transparently cite the source with the appropriate critical approach to address your problematic information source. Relying on plagiarized information as such constitutes as indirect plagiarism.

When using Turnitin templates, they already contain pre-configured Turnitin Feedback Studio assignments. If you build your own workspace, provide at least two separate assignments to support source checking: 1) an intermediate assignment to support independent writing processes, and 2) a final assignment for submission of the finished work.

In between these tasks, you can also include a commenting assignment that focuses on text structure, language usage, and methodological issues.

The instructions you provide using the tools on the Review tab of the report are stored in the review report. Students can refer to them when preparing the next version. The settings for the commenting task are the same as for the intermediate assignment.

The intermediate checks of the work document should be done in the same workspace as the final check assignment in order to avoid the mirroring of parallel versions of the work (e.g. 95% similarity) and the need for reference filtering in final comparison reports.

• You can also use the settings for the final check assignment in the submission assignments for home exams.
• The more flexible settings of the intermediate check assignment are suitable for the assessment of essays, learning diaries as well as translation and calculation exercises.
• If you suspect content recycling within a study group, you can set Report Generation Speed to "Generate reports on due date" so that all works submitted at different times are compared against each other when the assignment's validity period ends. This way, their similarities become visible. You can also use this setting when comparing corpus materials in your research.
• The group assessment function is even more flexible with the Turnitin Integrity check associated with Moodle's own Assignment activity: the student gets the Turnitin report immediately and again at the end of the course. If the first result is dubious, the student still has time to make a corrected version and learn the principles of transparent use of source information through experience.
• In courses based on the principles of flipped learning, you can support source-informed thinking by activating Turnitin Integrity in Moodle's Forum and Workshop activities. (See instructions in Chapter 4.)

Turnitin Feedback Studio is suitable for both originality checks and scoring or qualitative annotation of texts. With Turnitin, you can also comment on image materials, but Turnitin does not compare the originality of images.

You should plan the use of assessment support tools when setting up your workspace. You can control the functions on a red, blue, and black display level.

• The red similarity layer displays a link list of correspondences in descending order of percentage. Identified source and peer publications can be viewed in a screen window.
  Turnitin highlights in color the sections of the comparison text that match the reviewed text. This two-way comparison helps to profile the author's use of sources or to detect that both the author and the reference author have used a third common source.
  If you open the Turnitin report in Text-only format, you can see the timestamps of when the references were published on the Internet.
• The blue instructor feedback layer includes three sets of functions: 1) quick feedback with drag-and-drop text boxes, 2) general feedback in text or audio message, and 3) rubric and grading forms. You can edit and create your own quick feedback phrases and rubrics.
  The new, separate blue layer function is AI writing detector, which identifies the text sections produced with ChatGPT. Turnitin’s AI writing detection indicator is available to non-student users. The texts must be in English and in length between 150 and 15,000 words.
• In the black display level, you choose whether the correspondences and feedback are displayed simultaneously or separately.
• The top bar displays grades (e.g. 80/100) and you can navigate between submission reports using the <-> navigation button. This helps ensure consistent use of criteria.
• From the white menu buttons, you can download the text to your desktop and view the metadata of the recordings.

Search for more instructions using keywords: How do you use Feedback Studio Moodle V2 ?

Adding a Feedback Studio assignment to a course area is an easy process which starts from the Admin Tools menu:

^{Managing tools  >>  Turn editing on  >>  +Add an activity or resource  >>  Add Turnitin Assignment 2}

When the task is activated, most of its settings are already correct. Some of the settings depend on the course's pedagogical goals and assessment format.

Implications and recommendations for mission settings:

Connecting the new Integrity helper to Moodle's Assignment, Forum or Workshop activity are alternatives to the Feedback Studio feedback task.

In the same Moodle course, you can use both the Feedback Studio and Integrity's inspection functions. Integrity is suitable for the evaluation of partial performances during the course and Feedback Studio for wider practice tasks and theses.

Implementation of Integrity starts from the management tools menu of the course view. In the following, it is attached to the Assignment:

^{Managing tools  >>  Turn editing on  >>  +Add an activity or resource  >>  Add Assignmen}

When the Moodle Assignment is activated, its settings are edited according to the Moodle guidelines. The settings for the Turnitin Integrity assignment can be found under its own heading. Recommended options are marked in red.

See more instructions (Integrity assistant is based on Turnitin Similarity):

• Using Turnitin with Moodle

Tell the students at the beginning of the course or supervision relationship that it is beneficial to start using Turnitin already in the drafting phase of their texts!

Depending on whether you open a Turnitin assignment for continuous use to support long writing processes or for a small time window as a weekly assignment submission channel for a course, its settings, usage, and monitored indicators are slightly different.

Below is a list of characteristics for those Turnitin Feedback Studio assignments that do not have restricted submission times, and for those assignments (either FS or Integrity) that are used on a scheduled basis.

Managing graded assignments in Turnitin

If you use scoring or another grading scale in your assignments, you can download score reports from Turnitin in Excel, PDF, or CSV format for later use. In Integrity-supported Moodle activities, you can directly use Moodle's own grading management tools.

Search for more instructions: How do you use Feedback Studio Moodle V2 ?

Turnitin is a tool for evaluating the transparency of source use. It identifies intentional and unintentional text matches. While it does not replace the expert judgment of a reviewer in assessing the quality and ethical use of sources, it can expedite the process.

• Most violations of the responsible conduct of research (RCR) are related to an individual's unethical behavior as part of a community. Turnitin does not recognise this.
• Plagiarism may be an indirect result of these problems of the social and qualitative dimension, but it is a small manifestation of all forms of violations of the responsible conduct of research, which Turnitin can recognise.
  Turnitin marks potential places of plagiarism in the texts, i.e., identical wording in different sources. These word matches may be problematic, resulting from copying, reliance on the same source, or the standards and standardised practices of the vocabulary used in reporting - and are acceptable in such cases.

Plagiarism, i.e., unacknowledged borrowing, may be associated with the data collection process as one of its stages.

• Collecting data onto one's own desktop can be straightforward copying from different sources. As such, it meets the characteristics of plagiarism.
• Unacknowledged borrowing does not lead to problems as long as the borrowed information remains for the borrower's private use, and it is not shown or used outside of one's own desktop.
• Copying information becomes plagiarism - which is one form of academic dishonesty - when it is published in one's own name or as part of one's own text, with insufficient or misleading notation.
• Academic misconduct is involved when plagiarism is used in exchange for something, such as a credit-producing exercise or a thesis related to a degree.

The problem disappears when the collected information is linked with careful references, necessary usage permits are ensured, and the referenced content or quotation is written as an argumentative part of one's interpretation, thus forming the quotation right as defined in the Copyright Act.

• See: FairUTU >> RCR violations
  Marjut Salokannel, Avoimuus vs suojaaminen - avoin tutkimusdata oikeudellisesta näkökulmasta (TY 2015)
   

Turnitin's search robots mine millions of pages of the Internet and licensed e-publications into its reference database every day. When generating comparison reports, Turnitin uses these pre-indexed contents, so the review does not reach the very latest publications or contents published in closed environments or copy-locked formats. Therefore, Turnitin's comparison report is only a partial view of the information universe. It does not replace the expertise of teachers and supervisors.

Here's how to review a comparison report:
See first basic interface functions:  Turnitin Feedback Studio Walkthrough !

• The Turnitin report opens when you click on the submission's % -value displayed on the assignment's home page.
• On the right side of the report is a red equality menu and a blue review menu.
• Open the similarity view by clicking on the red number, which is the percentage of matches in the text. On the right side of the view, there is a link list of reference materials. Matches of less than one percent are usually bibliographic information. Most of the correspondence coloring can be found in the publication's bibliography and footnotes.
  - The footnotes on the sample page contain reference information and citations from the texts of four different source publications.

• Below is the same example view when reference 1 (dokumen.tips) has been selected for close inspection. By browsing through the "All Sources" mode you can see how the cited quote fits into the source publication and how widely and efficiently the source has been used. Arrow 1 points to the beginning of the correspondence between the text and the source text.
• Gaps in uniform copy coloring (gap = different character string) can have noteworthy reasons. Arrow 2 shows that in the reference, "l'Amazzone" has been written together with the previous stem, whereas in the checked article, that error has been corrected.
• The gap may also be due to the syllabification of the comparison text if the so-called "hard syllabification" has been used.
• Sometimes, gaps are a sign of wider copying of the reporting format: only the unique result values of own research layout do not color, but the rest of the narrative structure is colored because it is plagiarised.

If there have been intermediate checks on the text in multiple course areas during the writing process, the report view can be completely colored when the similarity to the parallel recording is 100%.
By filtering out reference links step by step, you can refine the picture of the roots of the text.

^{See: Excluding sources via match overview}

The main focus should be on the consistency and clarity in marking sources. According to studies, the most serious errors due to negligence or fraud relate to the indication of a source reference, either by covering the used source data in a narrower text area or by including one's own text to the text interpreted as source information within the reference.

Identifying text manipulations

The Turnitin system detects manipulation attempts related to the code of the text. It points out with a red flag icon above the match % if the document contains e.g. texts embedded in images or other code deviations affecting the content.

• The traditional method of modification is to replace spaces with dots or other punctuation marks in the color of the publication base (usually white), in which case the code string cannot be directly compared to references.
• Hard hyphenation (i.e. not automatic) for a narrow column can lower the match value of the content.
• More advanced ways are invisible code additions to the file.
• A more significant problem is the modification of the text using language tools. At its simplest, it means using paraphrase generators that plug into your Internet browser. The copied text can be transformed into another form, while keeping its essential contents fairly original. A more advanced method is the "rewrite", "condense", "translate to language X + translate from language X to home language" procedures performed with language model tools, which can be supplemented with language maintenance tools from another operator to erase the "fingerprints" of the language model.

During March 2023, Turnitin released language model-based ghostwriting detection tools for all its products, meaning Feedback Studio and Integrity at the University of Turku. We will supplement these instructions accordingly as soon as possible.

• How do you use Feedback Studio Moodle V2 ?
  Turnitin blogi, artikkeleita AI:n vaikutuksista kirjallisiin suorituksiin 

Identifying AI-generated texts

The AI writing indicator will show an overall percentage of the document that may have been AI-generated. To open the new AI writing report, select the AI writing indicator. Please note, only instructors and account administrators will be able to see the indicator and the AI writing report at this time. 

The AI writing report contains the overall percentage of prose sentences contained in a long-form writing format within the submitted document that Turnitin’s model determines was generated by AI. These sentences are highlighted in blue on the submission text in the AI writing report. 

The percentage, generated by Turnitin’s AI writing detection model, is different and independent from the similarity score, and the AI writing highlights are not visible in the Similarity Report.

Turnitin’s AI writing detection model only highlights text that is highly likely to be AI-generated. This is to help ensure that students are treated fairly whilst safeguarding the institution’s academic integrity standards. The percentage is interpretive and should not be used as a definitive measure of misconduct or punitive tool. Instructors should use this indicative percentage to help them decide how to best handle work that may have been produced or partially produced by AI writing tools.

In order for a submission to generate an AI writing report and percentage, the submission needs to meet the following requirements:

• File size must be less than 100 MB
• File must have at least 150 words of prose text in a long-form writing format
• File must not exceed 15,000 words
• File must be written in English
• Accepted file types: .docx, .pdf, .txt, .rtf

See: AI Writing Detection

When you, as the instructor, comment on the use of sources, you can use bubble comments, inline comments, recording voice comments or text summary comments. In the example below, the instructor asks for the reason behind the changes in words compared to the original text. - The explanation for the differences may also be that Turnitin has found a reference for another historical edition.

<Video view of scrolling comparison texts, coming soon>

You can attach a rubric or grading form from the Grademark options ("Launch Rubric Manager") to your Turnitin assignment. The templates of the forms are based on criteria for skill levels according to different standards.

With Turnitin’s graphing tools, you can effectively track student performance.

At the end of active use of the course environment

• Make sure that accepted performances have been recorded or transferred to the performance register. Do not use Moodle as an archive for student performances. Collect student feedback on the course implementation if necessary.
• Review the phenomena that have raised suspicions of cheating in academic performances. Serious suspicions of cheating have been dealt with the support of an Investigator and reported, but you can send observations about text manipulation and fraudulent use of information to vilppiraportit@utu.fi - the University's quality system is developed with the help of this feedback.
• Instruct students who have not completed the course on how to apply for the next academic year's course implementation.

Finally

• You will make a plan for further use or updating of the guiding materials in the workspace for the next academic year's course.
• You will assess your need to use the Turnitin feedback tools more extensively in the next course implementation.
• You will hide the workspace from view (Administration tools: "Hide course"), after which the course name will display a logo  

The evaluated documents that remain in the workspace are documents containing personal data, the storage time of which is 6 months in accordance with the University's guidelines.

The storage time for theses is longer and regulated in such a way that storage does not take place on the Moodle platform.

Bachelor's theses are stored for 10 years (archiving responsibility lies with the faculties), and theses related to Master's and doctoral degrees are subject to permanent storage, which takes place in the UTUPub publication archive of the University Library.

The Moodle workspace is not suited for archiving, since the lifespan of the workspace is short. It is not necessary to retain intermediate check assignments for theses. Even the final check assignment can be deleted once the thesis has been assessed and recorded in the Peppi register. At that point, the thesis has already been transferred to the UTUPub archive.

Process images for checking Bachelor and Master's theses and doctoral dissertations. Functions as a generic check list.

AI Writing Detection helps educators determine if a text has been produced, edited, or translated using a large language model (LLM). Like large language models, its detection system is also based on a language model that applies the latest generation of transformer deep learning architecture.

AI Writing Detection is available through Turnitin Feedback Studio and Moodle’s Turnitin Integrity assignments.

The system is currently fine-tuned to work with the English language. Optimizing for the scientific community's most widely used language precedes expanding functionality to other languages supported by Turnitin.
 

HOW DOES THE AI DETECTOR WORK?

A document uploaded to a Turnitin assignment is divided into text snippets in two ways: sequentially and in sequences that overlap each other. These individual snippets are analyzed by the detection tool, which determines the likelihood that they are generated by generative AI. The system then calculates average scores for all analyzed text snippets and creates an overall assessment of whether and to what extent the text is written by AI. Turnitin’s language model-based analysis is capable of recognizing subtle statistical features in AI-generated writing, primarily trained to detect AI assistants based on OpenAI language models.
 

WHAT ARE THE DETECTABLE FEATURES?

Language models based on transformer deep learning architecture have the ability to link individual text tokens (words or subwords) to each other in many nonlinear ways, facilitating associative relationships between tokens. The computational infrastructures of large language models from OpenAI, Antropic, Google, and others, which support hundreds of billions of parameters, combined with training datasets as extensive as the entirety of the indexed internet, endow language models with the ability to perform complex reasoning tasks and produce natural language.

While advanced AI assistants write in a very 'human-like' manner, there are statistical structural features in their texts that AI systems trained and optimized for detection can identify. Each language model has its own handwriting, as it were, which reflects in its structural features. These characteristics arise from how language models sequentially produce tokens according to the probability distribution defined for them. The token sequences created by language models are more schematic in structure than descriptions written by a human on the same topic or concept.

Simple concepts that describe the comparison of machine- and human-generated text are "perplexity" and "burstiness." Perplexity refers to the statistical "fluidity" of a flow of words, while burstiness represents deviations from typical sentence structures, expressions, and their extents. Additionally, the overall picture is defined by a large set of statistical dependencies in a broader continuum that distinguishes human writings from texts produced by language models. To verify this broader framework, Turnitin's AI Writing Detection only processes texts that are at least 300 words in length.
 

ARE FALSE DETECTIONS A SIGNIFICANT PROBLEM?

The level of false positives must be below 1%. Minimizing this value (FRP = False Positive Rate) is a condition for the acceptance of the detector's use. Before the system was implemented on April 4, 2023, Turnitin conducted reliability tests using, among other things, 800,000 texts stored in the system before 2019 as a reference. These texts, uploaded to Turnitin before the release of GPT-3, were assumed to be written by humans. In addition, the system was tested on a data set consisting of thousands of texts written purely by humans, purely by AI, and by humans and AI together.

When tested with this dataset, the False Positive Rate of the Turnitin AI Writing Detection system was 0.007 (0.7%) at the document level and 0.002 (0.2%) at the sentence level.

Reliability Criteria:

• Precision. The ratio of true positives to true positives + false positives. High precision means few false positive results, i.e., situations where text written by a human is incorrectly identified as being written by AI.
• Recall. The ratio of true positives to true positives + false negatives. High recall means that the system can identify a large portion of texts written by AI.
• F1 Score. The numeric value 2 x (Precision x Recall) / (Precision + Recall). F1 scores are the harmonic mean of precision and recall, which describes the balance of the metrics. F1 is useful for comparing systems that may be optimized for either precision or recall.
• False Positive Rate. The ratio of false positives to false positives + true negatives. A low FPR number is desirable to avoid classifying texts written by humans as generated by AI.
• Accuracy. The ratio of true positives + true negatives to all cases. Overall accuracy describes how large a portion of all identifications (positive and negative) was correct.
   

ARE THE TURNITIN ORIGINALITY REPORT AND AI DETECTION REPORT LINKED TO EACH OTHER?

They are not. Turnitin's plagiarism detection and AI writing detection are two separate processes, but they open in the report view.

When you open the Similarity Report, the AI writing detector is found in the sidebar. The AI Detection Report opens in its own tab in the window from which the Similarity Report was launched. Any pop-up blocker must be set to allow pop-ups from the Turnitin system.

Sentences identified as created by AI are highlighted in the report with a blue shading color. The detector displays one of three possible states:

• BLUE, with a percentage range of 0–100: The submission has been processed successfully. The percentage indicates how much of the estimated text volume the AI detector determines to be created by AI. If the text includes lists or tables depicted with dashes or bullet symbols that do not form complete sentence structures, they are not processed by the detector.
  According to the tests, the possibility of false positives is higher when the overall level is low, between 1–20%. The detector then displays an asterisk (*).
• GRAY, without a displayed percentage (--): The AI detector is unable to process this submission for one of the following reasons:
  The submission was made before April 4, 2023, when the AI detector was implemented; reports of historical submissions require re-uploading of texts.
  The submission does not meet one of the following format requirements:
  • The file must be under 100 MB
  • The file must contain at least 300 words of prose text
  • The file must not exceed 30,000 words in length
  • The file must be written in English
  • The file type must be .docx, .pdf, .txt, or .rtf.
     
• ERROR ( ! ): Turnitin has failed to process the submission. Try again later. If the file meets all the above file requirements and this error state still appears, please contact the helpdesk.
   

See

• Frequently Asked Questions
• Artificial Intelligence in Teaching and Studying
• Artificial Intelligence in Research

• Turnitin End-User License Agreement
• Turnitin system Privacy statement
• Accessibility statement