Leon Furze

This is the second post in the series, focusing on how generative artificial intelligence tools, such as text and image generation, can support the writing process. It follows the writing cycle outlined in our book Practical Writing Strategies, emphasising how generative AI can enhance writing skills without relying entirely on technology. For the previous post on stage one: Purpose, click here.

Exploration

Before getting into the exploration stage, writers establish the purpose, audience, and context for their writing. This foundational step is crucial for moving into the exploration stage effectively.

In the exploration stage, writers begin looking at quality examples of texts in the forms and genres they aim to create. This stage can be fairly informal, with writers like magpies gathering shiny objects, collecting snippets and examples from other authors to learn tricks and techniques. Or, it can be more structured, with explicit instruction and modelling of key skills such as annotation and close reading.

It’s important to note that the exploration stage can begin well before the actual writing process starts and can continue throughout a writer’s career. This post focuses on how students and young writers can use this stage to develop core skills, such as annotation, identifying style and voice, and drawing out writing features and techniques from other authors’ work.

In Victoria, short excerpts of quality texts are referred to as “mentor texts.” Like a good mentor, these texts allow students to learn by example, offering techniques, best practices, and encouraging them to develop their own personal style and voice. It’s not about copying another author’s style but experimenting with imitation and pastiche during the process of finding one’s voice.

Generative AI and Mentor Texts

Generative artificial intelligence could be used to create mentor texts in the styles of other authors. However, due to legal and ethical considerations regarding intellectual property, we are not going to go down that road.

Legal arguments aside, the fact is developers of large language models have taken the intellectual property of hundreds of thousands of authors and used it in their models without permission or license. While it is possible to generate a text in the style of Margaret Atwood or Stephen King, there are many other ways of using this technology, which are more respectful of intellectual property rights.

The activities for this stage focus on using generative artificial intelligence as a tool to visualise and explore the textual features of mentor texts, such as the conventions of interesting uses of language.

Practical AI Strategies includes an entire section on GenAI policy and assessment. It is available from Amba Press

Activities for Exploration

Activity 1: GenAI Assisted Annotation

1. The teacher demonstrates how to annotate an extract of text, emphasising the importance of articulating choices and explaining the use of symbols and formatting (e.g., markers, circling, underlining, highlighting). Do this as a think aloud with one of your mentor text extracts, slowly going through the text with the class and explaining your choices.
2. Provide students with an unannotated electronic copy of the same extract of text and have them use generative AI tools to annotate this mentor text, focusing on language features, structure, and style. Try a few different models for comparison. Use a prompt like the following:

Suggest annotations for this text focusing on elements like word choice, language, structure and style. Use footnotes where you would ordinarily write notes around the text in the margins. After the annotations, explain your choices. <copy/paste text>

Examples from Copilot, ChatGPT and Google Gemini (click to enlarge):

Activity 2: What are we looking for?

Students are guided to identify techniques and features in mentor texts. This includes creating lists of features to look for while annotating, tailored to individual students or groups, to avoid overwhelming them with a generic list.

1. Discuss mentor texts: Share examples relevant to the topic being studied, such as extracts from news articles, blog posts, or social media for nonfiction and new media writing.
2. Features to generate: Students use variations of the following prompt to generate checklists of what they are looking for, focusing on techniques, style, voice, and other elements of quality writing. This list helps guide their exploration and annotation.
3. Annotation: Once students have created their checklist with GenAI, they annotate their mentor texts individually or in groups. After annotating for their chosen element(s), they share with the class.

We are using mentor or model texts as a way of learning the techniques and style of quality writing. Generate an annotation checklist of things we could look for in our mentor texts related to <style, structure, voice, tone, language use, word choice, etc.>

Examples from ChatGPT, Copilot and Google Gemini:

Enjoying the article and want to stay up to date with the rest of the series? Join the mailing list:

Activity 3: Visualisation Activity

This activity encourages students to visualise a mentor text based on their understanding and then use image generation tools for visualisation.

1. Read Through Evocative Writing: Choose an example of particularly evocative or descriptive writing from a mentor text. Discuss the original details that create atmosphere and sensory information. Annotate a passage as a think-aloud with the students.
2. Create Visualisations: Individually or in groups, students write or draw their visualisations by hand based on the mentor text.
3. Use GenAI: Use a prompt like the one at the end of this activity to generate visualisations in different image generation apps. In a chatbot based app like ChatGPT Plus or Copilot, you can use the entire extract as the prompt (these models rewrite your prompt into an image prompt). In other apps like Adobe Firefly students will need to write an prompt based on the extract (see example)

Use the following extract to create a visualisation of the text: <copy/paste extract>

Chatbot style image generator prompt, e.g., Microsoft Copilot

A vivid, detailed, digital illustration depicting a person sitting on a green bench in a subway station. The walls are covered in green tiles, and the atmosphere feels green-tinted. Above ground, Alexanderplatz is depicted as a vast, empty expanse of grey concrete, with tiny figures scattered around, feelings of insignificance. Sense of isolation and discomfort in an urban environment. Abstract elements, paint, digital artwork.

Adobe Firefly prompt example based on extract from Anna Funder’s ‘Stasiland’

Activity 4: Strong versus weak

Mentor texts don’t have to be fiction, and they don’t need to be limited to creative responses. We use mentor texts of quality student writing to teach essay skills, such as how to write a strong argument, or make appropriate language choices for an analytical response.

In this activity, adapted from Practical Writing Strategies, students compare the generated output for “strong” versus “weak” arguments, using them as mentor texts and annotating them in the same manner as previous exercises. This is a quick way of providing some “student” model texts. After a few rounds of writing from your students, you’ll also have plenty of real human examples to add to the mentor text pile.

1. Ask students to brainstorm a list of characteristics that they think make a paragraph strong or effective.
2. Discuss the responses as a class and identify the key features; these could include:
   1. A clear, main idea
   2. Supporting evidence
   3. Logical structure
   4. Appropriate language
3. Provide students with two AI-generated sample paragraphs and ask them to read through them carefully, or have students generate their own using a prompt similar to the one at the end of this activity.
4. As they read, have them annotate of the features that stand out in each paragraph.
5. Have them compare the two paragraphs and discuss in groups or as a class, considering the structure, language, tone and overall effectiveness of the writing.

Generate two exemplar pieces of student writing, suitable for <year level> students studying <analytical/persuasive> writing about <topic>. The first piece should be of an average quality and the second piece much higher. Do not explain the differences. Only present the examples.

The next webinar for 2024 is Introduction to Generative AI on March 4th. Join us for a 1 hour webinar exploring the fundamentals of multimodal generative AI and the implications for education.

Conclusion

These activities are designed to support students and beginning writers with exploring mentor texts. Often, that means some clear instruction around core skills like annotation and close reading. These skills can be developed throughout K-12, but I have often found in the past that even senior school students (and tertiary) lack a basic understanding of how to annotate texts.

Having gone through the purpose and exploration stages of the writing cycle, students should now be ready to get some ideas down on paper (or screen). The next post in this series will look at the ideas stage, and how GenAI can help students to extend both the depth and breadth of their initial ideas.

If you have questions or comments, or you’d like to get in touch to discuss GenAI consulting and professional development, use the form below: