Artificial Intelligence Helps Grade Exams 90% Faster

Four UC Berkeley researchers developed a program to help grade papers during their time working as teaching assistants – and now, they’ve added artificial…

NVIDIA logo
NVIDIA
•
Brad Nemire
•2 min read•intermediate•
--
•View Original
Artificial IntelligenceTensorFlow

Overview

Researchers at UC Berkeley have developed an AI-enhanced grading application called Gradescope, which significantly reduces grading time for instructors. The app, which has processed 10 million answers, aims to cut grading times by up to 90% by automating the identification of question types and recognizing handwriting.

What You'll Learn

1

How to leverage AI for automating grading processes in educational settings

2

Why AI can enhance the efficiency of grading in large classrooms

3

When to implement AI features in existing grading systems

Key Questions Answered

How does Gradescope utilize AI to improve grading efficiency?
Gradescope uses AI to automate the grading process by identifying question types, distinguishing written marks, and recognizing handwriting. This allows instructors to grade one answer and apply that grading to multiple students, significantly reducing the time spent on grading.
What technologies were used to develop the AI features in Gradescope?
The development of Gradescope's AI features involved using Tesla K40 and GeForce GTX 980 Ti GPUs, along with CUDA and cuDNN-accelerated TensorFlow. These technologies helped train the models for recognizing various answer types.
What impact has Gradescope had on grading time?
Gradescope has reduced grading time by 50% since its launch, and with the addition of AI features, it is expected to cut grading times by as much as 90%. This efficiency is particularly beneficial for large classes.

Key Statistics & Figures

Reduction in grading time
90%
Expected reduction in grading time with the addition of AI features
Number of answers processed
10 million
Total answers accumulated by Gradescope from various college courses
Initial reduction in grading time
50%
Reduction achieved since the launch of Gradescope two years ago
Grading time reduction for a large exam
75%
Reduction experienced by co-founder Pieter Abbeel when using AI for grading a final exam for 600 students

Technologies & Tools

Some links below are affiliate links. We may earn a commission if you make a purchase.

Hardware
Tesla K40
Used for training AI models in Gradescope
Hardware
Geforce Gtx 980 Ti
Used for training AI models in Gradescope
Software
Cuda
Utilized for accelerating deep learning processes in Gradescope
Software
Cudnn
Used in conjunction with TensorFlow for deep learning model training
Software
Tensorflow
Deep learning framework used to train AI models for grading

Key Actionable Insights

1
Incorporating AI into grading systems can drastically reduce the workload for instructors.
By automating repetitive tasks like grading similar answers, educators can focus more on providing quality feedback and engaging with students.
2
Utilizing advanced GPUs and deep learning frameworks is crucial for developing effective AI applications.
The use of Tesla K40 and GeForce GTX 980 Ti GPUs, along with TensorFlow, demonstrates the importance of selecting the right hardware and software tools for AI development.
3
AI-assisted grading can be particularly beneficial in large classroom settings.
As shown by the experience of co-founder Pieter Abbeel, using AI for grading a final exam for over 600 students resulted in a 75% reduction in grading time, highlighting its effectiveness in managing large volumes of assessments.