Content-Anchored
Quiz Generation

Content-Anchored
Quiz Generation

Content-Anchored
Quiz Generation

Verify student authorship by auto-generating a quiz from the student’s own submission to test comprehension of what they wrote.

Problem

Detection tools flag AI-written work but can’t prove authorship. A student who wrote it can defend it.

Role

Lead Designer, POC owner

Stack

Figma, Claude API, internal quiz engine

Outcome

3–5 questions per submission; 82% of pilot flags matched follow-up review outcomes.

The hypothesis

If a student wrote a paper, they should be able to answer questions about it (in a proctored quiz environment) that an AI author couldn’t.

Quiz Setup Flow

Student Assessment Flow

Instructor Review Flow

How generation works

The paper’s key claims and specific examples become the source material. The model generates 3–5 short-answer questions based on the content of the students submission, within the parameters that the instructor has configured. AuthorProof uses the Canvas Quiz API to create the quiz using the generated questions and assigns it to the associated test taker.

The paper’s key claims and specific examples become the source material. The model generates 3–5 short-answer questions based on the content of the students submission, within the parameters that the instructor has configured. AuthorProof uses the Canvas Quiz API to create the quiz using the generated questions and assigns it to the associated test taker.

The paper’s key claims and specific examples become the source material. The model generates 3–5 short-answer questions based on the content of the students submission, within the parameters that the instructor has configured. AuthorProof uses the Canvas Quiz API to create the quiz using the generated questions and assigns it to the associated test taker.

Example question

"Your essay cites a specific study to support the claim that misinformation spreads faster accurate news. Which of the following correctly identifies that study?"

The reviewer experience

Reviewers see the paper, the questions, the answers, and a similarity score all within the same ecosystem if SpeedGrader where they can easily access a direct link to the proctored session for the assessment attempt.

Pilot results

82% of pilot cases correlated with follow-up review outcomes. Zero automated academic-integrity decisions — this is a tool for a conversation, not a verdict.

Quiz Setup Flow

Student Assessment Flow

Instructor Review Flow

The hypothesis

Detection tools ask did AI write this and give a probability. That’s not proof. If a student wrote a paper, they should be able to answer questions about it that an AI author couldn’t.