Academic Integrity & Learning Philosophy
LLM Grader is designed to transform the homework experience from a high-stakes “one-off submission” into a low-stakes “iterative conversation.” To use this tool effectively, it is important to understand its role in the broader course ecosystem.
1. A Formative, Not Summative, Tool
LLM Grader is not intended to provide an authoritative evaluation of a student’s success on a particular problem.
Large Language Models can occasionally hallucinate, miss subtle context, or provide inconsistent feedback. Therefore, the results generated by this tool should not be viewed as a final certification of mastery. Instead, think of it as a Personal Teaching Assistant—a tool that provides immediate, iterative feedback to help students identify and fix their own gaps in understanding.
2. The Iterative Learning Model
We advocate for a model of “Infinite Attempts.”
- Mastery through Iteration: Students are encouraged to engage with the AI feedback, refine their logic, and re-submit as many times as necessary to arrive at the correct answer.
- Open Collaboration: Students may discuss problems with peers or consult other AI tools to interpret the hints provided by LLM Grader.
- Expected Outcome: Under this model, it is expected that most students will be able to arrive at a “correct” answer before their final submission.
3. Recommendations for Instructors
Because students are encouraged to iterate toward the correct answer with AI assistance, we recommend that instructors value these results similarly to class participation.
- Homework as Engagement: Use LLM Grader to ensure students are “doing the work” and grappling with the material throughout the week.
- Verification of Mastery: Use high-stakes, in-person, or “unplugged” assessments (such as pen-and-paper quizzes and exams) to verify individual mastery of the core concepts.
4. Digital Signatures & Data Integrity
To support institutional requirements and prevent casual tampering, LLM Grader includes an optional Digital Signature feature using Public-Private Key infrastructure.
What it Secures
The digital signature ensures that the file submitted to an external LMS (like Gradescope) is the original, untampered output from the llmgrader portal. It prevents students from manually editing their scores or feedback text within the data file before uploading it.
What it Does NOT Secure
- Identity: Since use can be anonymous or guest-based, the signature does not confirm the identity of the person at the keyboard.
- Originality: The signature does not confirm that the student did not use other AI tools to generate the answers they provided to the grader.
The digital signature is a tool for data integrity and technical accountability; it is not a replacement for traditional academic proctoring.