• Nirmal Patel

Can AI do grading for teachers?

Artificial intelligence is growing in capacity every day. Novel AI systems are emerging that are doing what people thought wasn't possible. A decade or two ago, AI systems were more specialized. Now, they are becoming more generalized. For example, a chess-playing program was seen as a very successful demonstration of AI man years ago. But that program could only play chess and was not able to do anything else. People are now creating systems that can be used in a wide variety of tasks.


AI has been impacting education for some time now, but most of AI in education work has been centered around adaptive and personalized learning. Giving learners content that is at their level, giving students what they need based on their prior interactions and assessments. Adaptive learning technology is now part of a large percentage of online learning platforms. But AI in education doesn't have to stop there.


AI for grading


gif

Teachers spend an enormous amount of time grading student work on paper. We can capture those grading actions via SmartPaper technology aggregate them into AI models. When we have many examples of how teachers grade questions, then AI can start grading those questions instead of teachers. For example, if there is a trigonometry problem where we have ~500 examples of teacher grading, then we can build a computer vision model that can capture the grading intelligence of the teachers and start reducing the work that teachers have to do.


What can we do now?


Today's AI is capable of helping grade simple questions such as math fluency tasks, fill-in-the-blank worksheets, true-false and multiple-choice questions, etc. AI is also getting better at detecting math equations, chemistry equations, and different types of symbols. These functionalities can immediately help us automate trivial evaluation tasks that take many hours of work for teachers to complete. We described in another one of our articles how we can automate worksheet correction tasks.


Let's get started!


We are building SmartPaper in a way that can continuously capture the grading intelligence of the teachers. Simpler evaluation tasks are done by AI, and more complex work is examined by teachers or maybe other graders. Once sufficient examples of grading have been captured, AI can slowly start suggesting automated feedback. Once people have trusted AI's feedback enough times, we can let AI on 'loose' and start offloading complicated things to it! Although, we should be a little cautious and always keep an eye out for places where AI can make mistakes. Automation technology is not perfect, but with the human-in-the-loop approach, we can both benefit from it and avoid the pitfalls.

8 views0 comments