Middlebury Community Updates
AI and Academic Integrity
On November 6th, The Dean of Students notified all Middlebury students of changes to the Honor Code. One key change concerns the use of AI.
“AI Use Now Explicitly Included in Plagiarism Definition
The Honor Code now clearly states that representing AI-generated content as your own without citation is considered plagiarism. This includes both verbatim and paraphrased content. This change was approved by faculty and did not require a student vote.”
You can review the updated plagiarism statement here. Changes like this provide a good opportunity to revisit the AI statement in your syllabus and, given how quickly tools and practices are changing, continue discussing your expectations around AI use with your students.
What’s DLINQ been up to with faculty?
- Two-person audio in target languages – Here’s a quick example showing how to generate target-language audio from a script. This example is in Korean and uses Google’s AI Studio voice generator.
- OpenStax textbooks in NotebookLM – We mentioned in October’s update that Google released a number of OpenStax textbooks in NotebookLM. This PDF-to-NotebookLM workflow lets us create a very similar experience using any openly-licensed textbook. This example uses a physics text from OpenStax. (Automatically splitting the PDF into chapters using Python is a bit technical, but we’re happy to handle that part. 😊) You can experiment as viewers in this Notebook.
- Efficient prompt sharing – This is a URL generator that uses special URLs to auto-fill Gemini prompts and can trigger specific Gemini tools or custom Gems (Click here for an example.). This should make sharing prompts with students much easier. We also created a WordPress plugin that will expand our ability to collect, categorize, and share prompts. You can explore an example using prompts from the Career Development Center.
AI in the World
ChatGPT
ChatGPT-5.1, released on November 12 and, rolling out gradually, is described as “warmer, more intelligent, and better at following your instructions.” There seems to be real focus on the customization of tone and style (which may or may not be related to OpenAI’s decision to allow “mature content” in December). Another of its highlighted features is that the base model can now automatically route questions and tasks to other specialized models as needed without requiring the user to select a specific model themselves. If it works, this will provide advantages in terms of accuracy, energy usage, cost, and simplicity.
AI Browsers
Continuing last month’s topic of AI-first browsers, ChatGPT launched their Atlas browser on Oct. 21st (and Microsoft re-launched Edge two days later). BrowserOS is another new option. It’s the first Open Source browser which allows you to pick the AI model of your choice. If you’re having trouble differentiating, AIMultiple has a solid benchmarking article that covers 8 of the more popular AI browsers. The tests also paint a picture of what these tools can actually do. It is important to stress that many of the benefits these tools provide rely on having deep access to your data and the automation opens up new security risks that we have not dealt with previously and almost certainly do not fully understand.
AI in Canvas
Canvas, Middlebury’s learning management system (LMS), continues to develop AI-enabled features. When possible, we leave the decision to use these tools up to each instructor. For example, this summer we enabled two optional features: Smart Search and Discussion Summary.
The newest round of Canvas AI tools is branded “IgniteAI.” DLINQ is reviewing these features as they become available. We welcome instructors to test them with us. We’re now reviewing the following options:
- IgniteAI Rubric Generator – create assessment rubric criteria automatically based on an assignment description
- IgniteAI Translations for Inboxes and IgniteAI Translations for Discussions & Announcements – generates translated versions of inbox messages, course announcements, and discussion posts in a limited menu of languages
- IgniteAI Question authoring for Quizzes – upload source material (PDF, doc, text) to generate quiz items aligned with specific learning objectives and standards
If you would like to help us review these features, schedule a DLINQ consultation.
