Detect AI-Generated Text: Can Turnitin Spot ChatGPT’s Work?

The emergence of AI-powered text generation tools, such as ChatGPT, has opened new horizons for content creation. These tools have garnered significant attention from students and educators alike, raising questions about their detection in academic settings. In this article, we delve into Turnitin’s AI writing detection capabilities and explore how they work to maintain academic integrity. Know how to to Detect AI-generated text with Turnitin.

Can Turnitin Detect AI-Generated Text?

Turnitin, a widely-used to Detect AI-generated text, has adapted to the evolving landscape of content generation. It now offers AI writing detection capabilities, aiming to address the concerns surrounding the use of AI in academic assignments. This feature is a valuable addition to educators’ toolkits, helping them ensure fairness and uphold academic integrity.

Understanding Turnitin’s AI Writing Indicator

Turnitin’s AI writing detection is integrated into the Similarity Report. This indicator provides an overall percentage representing the portion of a document that AI writing tools, like ChatGPT, may have contributed to. It also offers a detailed report highlighting specific text segments predicted to be AI-generated. Importantly, this information is visible only to instructors and administrators.

However, it’s crucial to emphasize that Turnitin’s AI writing indicator does not make a final determination of academic misconduct. Instead, it offers data that educators can use to make informed decisions in alignment with their academic and institutional policies. Thus, the AI writing indicator should not serve as the sole basis for actions or definitive grading measures.

How Turnitin’s AI Works to Detect AI-generated text

Turnitin’s AI writing detection process involves breaking down a submitted paper into segments, each comprising a few hundred words or approximately five to ten sentences. These segments are then overlapped to consider the context of each sentence.

Each sentence is scored on a scale of 0 to 1 to determine whether it was authored by a human or generated by AI. If a sentence is identified as AI-generated, it receives a score of 1; if it’s deemed human-authored, it gets a score of 0.

The model calculates an overall prediction by averaging the scores of all segments within the document, indicating the proportion of AI-generated text.

Supported Language Models

Turnitin’s AI writing detection model is currently trained to identify content generated by GPT-3 and GPT-3.5 language models, which include ChatGPT. Additionally, it can detect content from GPT-4 (ChatGPT Plus) since the writing characteristics align with earlier model versions. Efforts are ongoing to expand the model’s capabilities to detect content from other AI language models.

Parameters Considered by Turnitin’s Model

The AI detection model takes into account the probability sequences of words generated by AI models like GPT-3 and ChatGPT. These models tend to produce observations in a consistent and highly probable manner, contrasting with the inconsistent and idiosyncratic nature of human writing. Turnitin’s classifiers are trained to identify these differences in word probability sequences, enabling them to detect AI-generated content.

Training of Turnitin’s Model

Turnitin’s model is trained on a diverse dataset encompassing both AI-generated and authentic academic writing across various subjects and regions. The dataset includes content from second-language learners, non-English speakers, diverse educational institutions, and less common subject areas. This comprehensive training minimizes bias and ensures accuracy to detect AI-generated text

Availability and Usage

Previously submitted assignments can be checked for AI writing detection by resubmitting them to Turnitin, provided AI writing detection is enabled for your account. Currently, the feature is available for documents submitted in long-form English.

Customizing AI Writing Detection

Administrators have the option to enable or disable AI writing detection in their account settings. Disabling the feature removes the AI writing indicator and report from the Similarity Report, making it invisible to instructors and administrators until re-enabled.

Impact on Workflow

Turnitin’s AI detection capabilities seamlessly integrate into the Similarity Report, ensuring a smooth experience for users. There is no major change in existing workflows or the use of the Similarity Report.

Integration with Learning Management Systems

Users can access the AI writing indicator and report via their Learning Management Systems (LMS), such as Moodle, Blackboard, or Canvas. While there is no embedded AI writing indicator in the LMS user interface, users can access the AI score by entering the report.

Authorship Detection vs. AI Writing Detection

It’s essential to differentiate between Turnitin’s AI writing detection and authorship detection within the Originality feature. AI writing detection assesses the likelihood of AI-generated content in a submission, while authorship detection focuses on identifying assignments not written by the student, without specifically indicating AI involvement.

In conclusion, Turnitin’s AI writing detection is a valuable tool in the academic landscape, promoting academic integrity while acknowledging the evolving role of AI in content creation. It empowers educators to make informed decisions, maintaining fairness and honesty in educational institutions.

Related Articles

Top 25 AI Tools Like ChatGPT: Your Ultimate Guide to ChatGPT Alternatives