Please note: Quick Write is still in BETA testing. If you’d like to be a beta tester, email us at info@myshortanswer.com
What is Quick Write?
Quick Write is Short Answer’s newest activity (use the clickable walkthrough below to preview). Students submit a response andm then receive a proficiency score and AI feedback for each success criteria set. The class earns points based on proficiency scores, with a cooperative goal of hitting a target class score. Students are then prompted to pair up and discuss the results together. Quick Write is Short Answer’s first activity to feature student-facing AI. In this post, we want to explain more about how and why we’re using AI.
How does Quick Write use AI?
AI is used provide one of three proficiency scores and feeback to student writing based on the success criteria assigned by the teacher:
- Advanced: Success criterion is fully demonstrated with strong execution
- Intermediate: Success criterion is partially demonstrated but has gaps
- Beginner: Success criterion is absent, misunderstood, or incorrectly executed
Is it safe to use AI with my students in Quick Write
Yes. Our team has taken several steps to ensure student’s are safe, including:
- Students sign in using a one-time join code, so no personally identifiable information (PII) is required (e.g. full names, contact info, demographic data, etc.)
- Writing is anonymous when passed to AI
- Amazon Bedrock is used for AI services, creating an effective wrapper to ensure no data is shared with AI model providers
- Amazon Bedrock enables guardrails to help unintentional sharing of PII, hallucination, etc.
- Short Answer uses the best available version of Anthropic’s Claude for scoring, a model that addresses some of the challenges of Large Language Model (LLM) powered AI
- Short Answer does no fine-tune models based on student data
- Students are prompted to think critically about AI’s ouputs throughout the activity, including messaging that includes:
- “AI is tool, not a replacement for your own thinking”
- “AI’s can make things up. Verify info with your teacher, textbook, or reliable online sources”
- “AI gives suggestions, but you make the final choices”
- “Your teacher is your best resource for writing guidance”
- Students can flag feedback for review by the teacher, who can update scores accordingly
- Students cannot “interact” with a model (i.e. go back and forth via a chatbot interface) and the AI outputs will only provide feedback based on criteria set by the teacher
Can AI really "score" writing and provide feedback?
Research is ongoing, but the answer is a tentative “yes” in the context of low-stakes formative assessment:
- A 2024 study compared a leading AI model’s ability to score student essays against human graders. On average, AI scoring was not statistically significantly different from human scoring. The researchers suggest AI scoring could be valuable for low-stakes, formative assessment and early draft feedback (i.e. how it is used in Short Answer).
- Another 2024 study found “providing feedback in the early phases of writing, where students seek immediate feedback” to be a “plausible use case for AI.”
- The two studies above support the opinions of other researchers who argued in 2023 that AI tools can help students become better at both traditional literacy and the new digital skills they need in today’s world.
Out team closely monitors this research to ensure we’re following educational best practice in the implementation of AI in Short Answer. Short Answer CTO Alexa Sparks is even working on original research in this area at the Stanford Graduate School of Education.
Have more questions?
Our team would love to chat. Feel free to use the blue button below to setup time to talk.
