StuDocu

Helping students exchange study notes with each other and differentiating between student and AI-generated documents.

The case in a nutshell

Studocu approached Miyagami to help solve the following design and engineering challenges:

To create outperformed industry-available options through custom software, doubling the accuracy of detecting AI-generated content.
To accurately assess the origin of the documents, differentiating between user-generated content from AI-produced documents.
To maintain the authenticity of their content library and continue to provide reliable resources for their community.

The client

Studocu is a platform where students exchange their own study notes with each other. However, they were facing a major issue: differentiating between documents generated by users and those produced by artificial intelligence. This challenge was impacting their business model, as it became increasingly difficult to ensure the authenticity of the content.

The challenge

Studocu was struggling to accurately determine whether the documents being uploaded to their platform were user-generated or AI-generated. Without a reliable method to differentiate between these two, the integrity of the platform's content could be compromised, potentially diminishing user trust and engagement.

Whether you need a one-off product or long-term support with a wider digital transformation, you need the best digital partner. Find out why we are the right choice.

Custom AI-generated content detection model

We aimed to maintain the authenticity of their content library and continue to provide reliable resources for their community.

We created a robust data structure capable of generating AI content in bulk

This large dataset was used to fine-tune an advanced pre-trained language model, ‘Roberta’, specifically tailored to detect AI-generated content. Our approach allowed for a comprehensive comparison between user-generated content and AI-generated material before the public availability of ChatGPT.

The custom model we created doubled the accuracy of detecting AI-generated content

Thanks to this solution, Studocu is now able to accurately assess the origin of the documents, differentiating between user-generated content from AI-produced documents. More important, they can maintain the authenticity of their content library and continue to provide reliable resources for their community.