Brainstorming benefits of an AI writing assistant app
User needs, smart prompts and functionality were central to creating an AI tool that supports student writing and thinking. Here, Joanne Chia and Angela Frattarola share the pedagogical and technical considerations involved in the process
Created in partnership with
Created in partnership with
It was the limitations of existing tools that inspired us. We wanted an app that would allow students to receive immediate feedback as they brainstormed ideas throughout the writing process. We envisioned a tool that would encourage self-evaluation and, where appropriate, draw on GenAI to help students develop ideas.
Back in 2022, colleagues from the Language and Communication Centre were chatting about their experiments with the tool Mentimetre and its then-new asynchronous feature that could offer students set explanations based on responses to questions. We created the app Waai the following year, with the aid of an EdeX teaching and learning grant, and working collaboratively with student assistants.
Finding our target audience
We decided that the students who most needed such an app would be the wide range taking a general writing class that is compulsory for all Nanyang Technological University (NTU) students. This course begins with students closely observing a lived experience, analysing their observations to formulate a question and then engaging with sources to develop a compelling argument that sheds new light on their topic. This design, which roots student writing in experience and enquiry, has become even more relevant with the pervasiveness of GenAI.
Before we began designing the app, we surveyed about 240 students from the writing class. This survey showed that students would be excited to try an app that could assist them throughout the researching and writing process. Yet they were also worried about losing their voices if they relied on the seeming convenience of large language models.
Pedagogical framework
The course is premised on the idea that writing is a tool for thinking, so students follow scaffolded steps that take them from fieldwork to an exploratory argumentative final essay. We strive to help each student discover their argument through the messy process of researching, drafting and revising.
We designed the sequence of pages in Waai to reflect the order in which the scaffolded assignments are tackled. Students are guided from the initial stage of making observations all the way to reflecting on their overall argument. The flow chart (below) shows how the stages in Waai complement the course curriculum:
In addition, design thinking was adapted as a non-linear framework for our app to make the writing process more relatable to STEM students, who might enter the course with this prior knowledge. The objective of the iterative process is for students to test out concepts to arrive at a possible new insight for their chosen topic.
To emphasise the non-linear aspect of the writing and research process, a side panel or menu allows students to move forward and backwards while brainstorming.
Incorporating AI into the app
Based on the coursebook, pre-designed prompts were created to mediate students’ engagement with AI. We paid particular attention to the kinds of information students could receive through the app to avoid limiting or overly influencing their ability to generate their own ideas. When designing the app’s dropdown menu of prompts, establishing backend prompts was equally instrumental to the success of the AI tool. Our app developer, Yajat Gulati, a Year 3 student from the College of Computing and Data Science, was central to coming up with the initial backend prompts through, perhaps ironically, the help of ChatGPT.
- Spotlight guide: Bringing GenAI into the university classroom
- Resource collection: AI transformers like ChatGPT are here, so what next?
- Campus podcast: The pros and cons of AI in higher education
The AI chatbot named Nudgy was designed from scratch by our user interface and user experience designers, Putri Azra Besim Kukuljac and Alicia Ng Ying Xuan, Year 3 students from the School of Art, Design and Media. Its dropdown menu allows users to accumulate the “chats” they have with the chatbot, based on the prompts they choose on the respective pages. The chatbot has a “save” function that stores all information provided by users, from their areas of interest to their source summaries, and feeds this information to ChatGPT-4 with the help of a system prompt.
Prompt engineering
Our goal was to avoid having the GenAI chatbot evaluate a student’s writing in a way that might contradict the teacher’s assessment. To that end, we presented students with front-end prompts that focused on helping them go deeper into their observations and analysis, such as: “Help me uncover useful definitions of concepts relating to my AEIOU observations”; “Identify any patterns or trends in my observations that I might have missed”; “Give me phrases that can emphasise the five senses from the AEIOU observations”. These prompts are found in the Introduce Op-Ed page, which includes pre-designed prompts that address both explicit and implicit observations based on students’ fieldwork experiences using an activities, environment, interactions, objects, users (AEIOU) framework.
For the back-end prompt engineering, we included a system prompt that instructs the Nudgy about its role as well as to clarify that it should not generate writing for the student. The Nudgy prompt templates ensure that the Nudgy identifies concepts and makes suggestions in terms of phrasing. This also avoids one of the issues students face when using GenAI for a written assignment – the potential for plagiarism. The back-end prompts thus define everything from phrases to patterns or trends. These prompt templates were tested, refined and modified before being implemented in Waai.
Optimising for mobile accessibility
Waai, which continues to be supported by NTU’s Institute for Pedagogical Innovation, Research & Excellence and Application of Teaching and Learning Analytics for Students, has most recently been optimised for mobile accessibility, which should increase frequency of use. While Waai surpasses writing apps designed for editing, it remains experimental. Our research suggests that it can enhance the writing and thinking process only if students are able to decode the feedback and apply parts of it to the revision of their drafts. As with any writing app, feedback is most beneficial when partnered with a human reader response.
Joanne Chia is lecturer and Angela Frattarola is senior lecturer and director of the Language & Communication Centre, both in the School of Humanities at Nanyang Technological University, Singapore.
This is an edited version of the blog post “Empowering student writing: how we built an AI writing assistant app”, which was first published by NTU’s Institute for Pedagogical Innovation, Research & Excellence.
If you’d like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.
