Grade papers and build lesson plans for teachers
This Easter I ran a series of experiments with ChatGPT that ended up with the creation of a POC app that uses ChatGPT to grade papers and build lesson plans for teachers. Task: This Easter I ran a series of experiments with ChatGPT that ended up with the creation of a POC app that uses ChatGPT to grade papers and build lesson plans for teachers. While I haven't stood this app up where folks can test it out (sorry the OpenAI API fees from the development of the app are coming in at 20 bucks now!), I did drop the source code into github for the industrious programmer to have a look. Archimedes app uses ChatGPT to grade papers at the elementary, intermediate, high school, bachelors, masters, and doctorate level. Ultimately I use a prompt engineering command execution harness that gives ChatGPT the ability to read files into the prompt so that it can process them. I then have prompt templates for grading papers, generating lesson plans, and providing feedback on pretty much any paper. Ultimately I think the grading I had it do on several of my papers was not only accurate, but was in many cases more informative and critical than the feedback I received from the professors. Given the sheer amount of time educators spend grading papers, I fully expect this process to be highly automated in the next few years with a solution not all that dissimilar for the one built here. My write up on the experience is here The Archimedes Paper Grading App source is located here (its a React+Flask app); tools: ChatGPT. Human role: The original post does not say how a person stayed involved.; limitations: Prerequisites include OpenAI API access and development costs (approx. $20 incurred). The source is a Reddit post with no verified metrics or deployment status. Integration uncertainty exists regarding file handling and prompt stability. Human approval is required for all grading decisions before finalization..
Tools
Published on Agent Work Atlas:
Overview
This Easter I ran a series of experiments with ChatGPT that ended up with the creation of a POC app that uses ChatGPT to grade papers and build lesson plans for teachers.
Workflow
This Easter I ran a series of experiments with ChatGPT that ended up with the creation of a POC app that uses ChatGPT to grade papers and build lesson plans for teachers. While I haven't stood this app up where folks can test it out (sorry the OpenAI API fees from the development of the app are coming in at 20 bucks now!), I did drop the source code into github for the industrious programmer to have a look. Archimedes app uses ChatGPT to grade papers at the elementary, intermediate, high school, bachelors, masters, and doctorate level. Ultimately I use a prompt engineering command execution harness that gives ChatGPT the ability to read files into the prompt so that it can process them. I then have prompt templates for grading papers, generating lesson plans, and providing feedback on pretty much any paper. Ultimately I think the grading I had it do on several of my papers was not only accurate, but was in many cases more informative and critical than the feedback I received from the professors. Given the sheer amount of time educators spend grading papers, I fully expect this process to be highly automated in the next few years with a solution not all that dissimilar for the one built here. My write up on the experience is here The Archimedes Paper Grading App source is located here (its a React+Flask app)
Reported result
Actual outcomes are unmeasured; the author subjectively found grading accurate and feedback more critical than human professors. The app was not deployed for public testing due to development costs. Proposed acceptance check: Compare AI-generated grades and feedback against a control group of human-graded papers for consistency and depth.
Human role
The original post does not say how a person stayed involved.
Limitations
Prerequisites include OpenAI API access and development costs (approx. $20 incurred). The source is a Reddit post with no verified metrics or deployment status. Integration uncertainty exists regarding file handling and prompt stability. Human approval is required for all grading decisions before finalization.
Try it
I want to build a POC app called Archimedes using React and Flask that integrates ChatGPT for grading papers and creating lesson plans. First, set up a prompt engineering harness that allows ChatGPT to read [replaceable input: sample student papers] into the prompt context. Create templates for [replaceable input: grading rubrics] and [replaceable input: lesson plan structures]. Run a dry process on one sample paper to generate a grade and feedback. Show me the output for my review before proceeding. Do not deploy or spend API credits beyond the initial test. If the feedback lacks critical depth, adjust the prompt templates and retry. Confirm with me before generating any further outputs.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.