Listening, Speaking and Writing (28/32) -- AI for Teachers: an Open Textbook

Listening, Speaking and Writing 28 Writing with AI Manuel Gentile and Giuseppe Città We have long been used to writing on computers with dedicated software that goes by the name of word processors (e.g., Microsoft Word, Google Docs, Pages, LibreOffice), taking advantage of the grammar suggestions provided by these tools. Raise your hand if you have never been saved by these tools from making glaring mistakes 🙂 But the transformation induced by these tools is not limited to correcting some typos; it has been far more profound in invoking a different way of writing. Digital writing allows us to return to what we have written and modify it to express more effectively what we want to convey. Using somewhat more technical words, we have moved from a linear approach to writing, to an iterative process. According to recent studies, the transformation of the writing process induced by digital tools has improved the quality of the texts produced. Writing in the AI era Anyway, the process of evolution of writing and related thought forms has not stopped. In recent years, with the explosion of AI, it has accelerated significantly. Tools such as Grammarly, Wordtune, Ludwig, ProWritingAid, and so on, are designed not only to provide grammatical correction of the text. They support the user throughout the writing process by stimulating the improvement of writing style, checking for plagiarism, and more. Recognising how the school world cannot be immune to such innovations is trivial. This is confirmed by the growing number of educational interventions, proposed in literature, that are designed to take advantage of such software. Some scholars propose using these tools to work on students’ skills in using external information sources to develop appropriate paraphrasing skills that can avoid plagiarism problems. Many of these tools can support the teacher in evaluating the texts produced by the students, providing timely analysis of individual student’s strengths and weaknesses. Moreover, these tools allow the student themselves to self-assess their own writing skills, thus enabling metacognitive processes and speeding up learning. All that glitters is not gold… Clearly, these innovations have potential problems. First, you have probably understood how, underlying all these deep learning mechanisms, is the source data on which the models are built. Limited or incorrect training data could cause bias. In addition, the risk of a general homogenisation of the texts produced/expected by those tools is likely. It could determine a consequent limitation (or penalisation in the case of assessment) of students’ creativity. Finally, these tools are primarily limited to managing the English language; thus, non-English-speaking contexts can be used in the L2 domain. That said, the speed of innovation is such that we will soon see similar tools emerge for languages other than English. A look to the future One of the main cognitive processes related to the writing process is