Can artificial intelligence be a useful tool for students with reading and writing difficulties?
- By: Anja Taugl
- Published: December 1st 2023.
This article has been translated from Norwegian using artificial intelligence.
At VID Bergen, I am responsible for BASP1010, which is the first course that nursing students start with in the bachelor's program. This fall, I have also been associated with VID's resource group for Artificial Intelligence. Here, I have been challenged to test out AI together with my first-year nursing students.
As a preparation, I asked the students if anyone had experience with AI, where over half answered that they would not consider using artificial intelligence for writing assignments or exams. Either because they were unsure of what would be considered cheating or which end of the gray area they would be in. I think this shows that AI is still new and untested for many.
AI has often been mentioned as a gamechanger for students with dyslexia, and by chance, early in the semester, I came into contact with a group of nursing students who have various forms of reading and writing difficulties in one of my learning groups. They were easily persuaded to join in testing AI in one of the work requirements in BASP1010, even though some in the group were a little skeptical about whether it could be of any help to use artificial intelligence.
Reading and writing difficulties have some common features, even though there are different degrees and challenges for each individual. Some of the challenges presented by my students are:
- Difficulty starting a writing process and coming up with ideas about the topic.
- Reading and understanding text, especially when it is complicated.
- Writing a good text that not only shows knowledge and is understandable, but where students should also start writing academically and use new words and technical terms.
What was the assignment?
In the work requirement in BASP1010, the students in part 1 must: find a self-selected article within one of the subject's nine main topics, with good help from the library and searches in databases. The selected article is presented to the other student groups in their own seminar. Then, in part 2, they must: answer a task individually where they must use the article they found together with relevant curriculum literature. Before final submission, they must give peer feedback. Students receive individual feedback on the final work requirement. In this project, the students with reading and writing difficulties were challenged to see if AI could help them achieve the learning outcome, and still stay within reasonable use of AI. It was of course completely voluntary to participate, and the teacher did not set any guidelines for which AI they might want to try.
The students had different experiences with AI from before. One had used ChatGPT, one had myAI from Snapchat, while the last two had never used any form of AI. Before they started working on the work requirement, we talked about what could be okay and not okay to use AI for, in accordance with VID's guidelines for cheating on exams and work requirements. This introduction corresponded with the information the rest of the cohort received at an information meeting about the exam later in the semester.
Experiences from the first part of the work requirement
The students who used AI to find suggestions for topics for the article could say that it was quite useful. Basp1010 is a large and complex subject where there are a number of different topics to choose from. Here, AI gave good suggestions that could be used when good instructions were given. After they had found the article, some used AI to summarize the article to better understand what it was about. Here, the experiences were mixed, where several noted that central parts of the article were omitted. What worked better was to ask AI to condense the text part by part, which also gave motivation to read and understand the article as a whole.
Experiences from the second part of the work requirement
There were three different approaches that were used to improve text. One asked AI to mark writing errors in the text and give suggestions on how the text could be improved. Another student asked AI to correct writing errors by rewriting it. One student asked AI to shorten the text and clarify the message while one student asked AI to write the text more academically.
I asked the students to reflect on the following:
- Is this still your text where you can vouch for the content?
- How can you assess whether a text has become better?
- What distinguishes between the text being written by you or by AI?
This was easy to answer for the students who used AI to find writing errors in the text. The student who had asked for a more academic text could, after a little reflection, answer affirmatively that it was still her text, but that the message came across better. In a work requirement where students are to learn to write academically, my assessment as a teacher is that this is a gray area, and the question of what the student has learned becomes decisive. Here, the student could have learned more by asking AI for suggestions to make the text more academic. Another student who asked AI about this received feedback that more use of curriculum literature would have elevated the text. This corresponded with final feedback from the teacher. The student who asked AI to make the student more academic missed this opportunity to improve his text.
What have we learned from our experiment with AI
Our attempt to use AI to improve work requirements for students with reading and writing difficulties gave different experiences. There were positive experiences related to asking AI to come up with suggestions for topics after working with the instructions and giving it what it needed. AI was therefore useful for the students to get started with the writing process and choose a topic. By dividing up the article and asking AI to summarize the main findings, several experienced that the text became easier to understand. The downside was that some main points were omitted. The students therefore could not avoid reading the article in its entirety. Using AI to write a good text without writing errors gave mixed experiences. Here we learned that it is the instructions you give AI that distinguish between whether you learn something, or whether the text is created by AI.
Several discovered that AI was not a shortcut, but that there was a need for subject knowledge both to give good instructions and assess the answers they received. "You need knowledge of the curriculum to use AI in a good way". When asked to give advice to other students who want to try AI, the students were unanimous: Be critical of what comes out. Try to learn something. AI can be a good tool, but is not a definitive answer. The students agreed that the experiences they have made can be relevant and useful for more students. Both for students with and without learning and writing difficulties, and for students in other studies.