Rafael Calvo: affective learning and writing

Rafael Calvo’s research at the Learning & Affect Technologies Engineering (LATTE) group aspires to complement traditional teaching methods by developing tools that support different forms of learning. This includes tools that use facial recognition to automatically detect a student’s emotional response to a tutorial and software that improves a student’s writing by providing different forms of automated and semi-automated feedback.
By Tim Groenendyk

Rafael Calvo

"Writing activities are learning activities, and the rhetorical structure you follow actually has a lot to do with the conceptions you have about the activity." Associate Professor Rafael Calvo

“The way you write tells us a lot about the way you think, and the way you work,” said Associate Professor in software engineering and LATTE director, Rafael Calvo.

Reading and marking assignments is one of the most time-consuming tasks for a teacher, Calvo explained. It is essential in a 13 week semester to get extensive and timely advice.

“Timeliness is very important. You can’t take four weeks to mark and give feedback on an assignment because it’s often too late by the time you give it.”

To bring the feedback closer to the writing process LATTE uses Google Docs to generate data on how students go about writing, and analyse writing habits almost instantly.

“For a student, writing activities are learning activities, and the rhetorical structure you follow actually has a lot to do with the conceptions you have about the activity. There’s a lot of evidence for that.

“Are you the kind of person that sits down in a three hour long sitting to write a whole essay? Or do you write five minutes here and there?

“I can see where you started writing: Did you start writing the title? Or did you start with the conclusions? Different approaches to writing the article will tell me a lot about what you were thinking when you wrote the article.”

But Calvo can delve even deeper with iWrite and Glosser. These LATTE-created tools can determine the semantic relationship between two paragraphs using statistical techniques.

“When you put a bunch of words on the page you can mathematically convert that into being a point on a graph. If the paragraphs are semantically related, if you’re talking about the same things, the two points are close on the graph. If another person writes about completely different things in the document the vocabulary is different, so two points will be farther apart.”

For Calvo’s team reading the feedback isn’t enough, soon projects will study how the student responded to the feedback, if at all.

“Soon, we’ll be able to see if you’ve made a change to the assignment. Because reading the feedback, if you don’t do anything, is not very useful either. That’s what we’re hoping to start up in the next year. Instructors can learn about what are the most effective forms of feedback.”

The most advantageous part of being able to see the writing process, however, is being able to better understand the process of plagiarism.

“And for the university that’s a huge issue.

“Because what is plagiarism? Plagiarism is about skipping the process of writing.

“When you focus on the process and not just on the outcome, the process regains its value.

“So plagiarism is useless, because you can’t forge process.”

Another side of Calvo’s work uses affective computing to give real time responses from an intelligent tutoring system.

“Affect means emotions. So we’re building systems that can detect emotions automatically, and then we can use that as a way of giving you feedback.”

LATTE records your physiological signs such as heart rate, respiration, the position of the face and head, and text (whether from writing or speech) to measure if the student is bored, confused or curious

“The system recognises those affective emotional states and the tutor changes. If he notices you are bored it gives you questions that are more difficult. If it becomes too difficult for you it can give you further explanation or easier questions.”

Calvo said that the accuracy of the intelligent tutor’s responses to the student’s emotions are often better than their baseline expectations.

“Technologies are getting better, and our understanding of human emotions is getting better.

“In very specific situations you can get pretty close to what a human will say about certain emotions.”