Embracing Generative AI in Education

Exploring GAI with students is crucial for their future success

Generative AI in Education from Stable Diffusion

With each passing week, the disruption posed by generative AI to education accelerates. Microsoft's announced integration of AI into Office Suite, the launch of GPT-4, and OpenAI's ChatGPT becoming available through their API are just a few examples of rapid deployment and adoption. Attempting to ban generative AI technology isn’t possible. Instead, we’re left with little choice but to adapt our educational practices and prepare students for the inevitable changes in writing, art, and our digital reality.

To do this, we need to develop our own AI and information literacy, while also investigating how AI can impact student writing, research, and reading. Here's a framework I am attempting to use for my first-year writing classes. It’s evolving in real-time and this Spring marks the second semester I used generative AI with students.

• Frame AI Assistance as an Emerging Practice

  An emerging practice is fluid and adaptable, allowing writers to integrate or discard aspects of AI assistance as needed. While writing will always require a structured process, framing AI assistance as a practice empowers students to thoughtfully combine human creativity with AI capabilities. We can do this by providing students with a clear framework, carefully selected AI tools, and thoughtfully scaffolded assignments.

• Recognize the Value of Human Writing

  I require my first-year writing students to start with their own writing before seeking AI assistance. This helps emerging writers establish a solid foundation, while more advanced students can gradually unlock AI's full potential in upper-level courses. I think this is a healthy boundary to set with students, one that shows the value of their own human writing and knowledge before seeking any form of AI assistance to augment their writing.

• Encourage Students to Identify AI-Generated Content

  I likewise see some value in asking students to clearly indicate AI-generated text in their work. This exercise helps them analyze how AI influenced their writing, and I encourage them to do this for that reason rather than policing their work. How long will this stay relevant given Microsoft’s integration of AI in their Office Suite is anyone’s guess. For now, I think it’s helpful for students to draw significant distinctions between their ideas and work and what was generated via an algorithm.

• Promote Metacognitive Reflection

  One of the main benefits of having students identify AI-generated text is that it easily allows them to reflect on how using generative AI affected their writing process or assignment outcomes, enabling them to better understand the technology's impact on their writing and learning.

• Choose the Right Tool

  Different AI writing assistants cater to various needs, so it's essential to select the one that best supports the student's specific assignment or learning outcome.

• Use AI Assistance to Augment Learning

  AI assistance should be treated as a learning aid, similar to how artists use tracing to develop their skills. While controversies will arise regarding the legitimacy of AI-generated content, using AI to augment human writing can offer valuable learning experiences.

Generative AI in Education

The above framework is sure to change—all emergent practices are bound to. The point is it is a starting point, a foundation, one that’s easy to modify based on how one uses it. Below are some examples of existing generative AI tools and some possible use cases for them in education.

---

AI Writing Assistants

While ChatGPT has its advantages, it may not be the best tool for introducing generative AI to developing writers. It functions primarily as a raw text generator with limited human input, unlike writing assistants that incorporate AI more effectively to augment natural writing. AI-integrated writing assistants provide students with the flexibility to either use the AI or rely on their own writing skills. This flexible approach can help students refine their writing through practice.

Lex

Wordtune Spices

Fermat

AI writing assistants have a range of use cases in education, helping students improve their writing skills and supporting faculty in tons of tasks. Here are a few:

1. Drafting Assistance: AI writing assistants can provide suggestions, ideas, or even complete sentences, helping students overcome writer's block and develop their thoughts more effectively.

2. Grammar/ Mechanics/ Rewriting: AI writing assistants can identify and correct grammar, punctuation, and spelling errors, helping students produce polished and professional writing. Students already have Grammarly for some of these features, but the ability to rewrite sentences is really amazing.

3. Style and Tone Suggestions: AI writing assistants can analyze the style and tone of a piece of writing, providing suggestions for improvement and helping students develop their unique voices. This can also function in the opposite and rob a student of their voice. This is why pedagogy matters!

4. Paraphrasing and Summarization: AI writing assistants can assist students in paraphrasing and summarizing complex texts, making it easier to understand and incorporate source material into their work.

5. Essay Organization: AI writing assistants can help students structure their essays, providing guidance on organizing ideas, creating outlines, and building logical arguments.

6. Citation Assistance: AI writing assistants can help students with formatting citations and bibliographies, ensuring they follow the correct style guidelines and properly attribute sources. Of course, any user will need to worry about the hallucination issue.

7. Feedback and Revision Support: AI writing assistants can provide targeted feedback on students' writing, identifying areas for improvement and suggesting revisions. Think about this as a powerful personal peer review agent.

8. Writing Prompts: AI writing assistants can help faculty generate writing prompts, and short assignments, or spur creative thinking, inspiring students to explore new topics and develop their writing skills.

9. Language Learning Support: AI writing assistants can help students learn a new language by providing translations, explanations, and suggestions for improvement in their writing.

10. Faculty Support: AI writing assistants can be used by educators to create lesson plans, generate exam questions, provide feedback on student work, and even automate administrative tasks like grading.

Reading & Research Assistants

Incorporating generative AI into PDFs or databases is really something. It enables users to employ the technology as a research and reading assistant, allowing them to find sources, receive reading assistance, and translate complex terms and concepts into plain language using natural language processing. This is particularly helpful for students with disabilities or those who are non-native speakers.

Elicit

SciSpace

ExplainPaper

AI research and reading assistants have numerous use cases in education, enhancing learning experiences and making information more accessible. Here are a few use cases:

1. Personalized Learning: AI research and reading assistants can tailor content to individual students' needs, adjusting the level of difficulty and presenting information in various formats to accommodate different learning styles.

2. Source Discovery: AI research assistants can help students quickly find relevant articles, books, and other resources for their research projects or assignments, saving time and effort.

3. Summarization: AI reading assistants can provide summaries of complex texts, helping students grasp the main points and better understand the material.

4. Translation: AI reading assistants can instantly translate content into different languages, making resources more accessible to students who are non-native speakers or learning a new language.

5. Vocabulary Enhancement: AI reading assistants can provide definitions, synonyms, and explanations for unfamiliar terms, helping students expand their vocabulary and improve comprehension.

6. Reading Comprehension: AI assistants can generate questions and quizzes based on the content, helping students test their understanding and retain information more effectively.

7. Accessibility: For students with disabilities or learning difficulties, AI reading assistants can provide additional support by offering text-to-speech features, visual aids, or other accommodations tailored to their needs.

8. Collaborative Learning: AI research and reading assistants can be used in group projects, helping students divide tasks, share resources, and synthesize information more efficiently.

9. Text Analysis: AI reading assistants can analyze texts for themes, sentiments, and patterns, helping students develop critical thinking and analytical skills.

Transcription Assistants

OpenAI recently introduced Whisper, a speech-to-text transcription neural network that transcribes speech with near-human accuracy. Although students have used various technologies to transcribe lectures for years, Whisper stands out. Once an audio file is uploaded, Whisper transcribes the lecture and processes it through OpenAI's GPT-3 powered Playground. Users can then write simple commands to summarize, organize, prioritize content, and even remove irrelevant material. Additionally, the transcribed lecture notes can be synthesized and connected with notes from previous classes, allowing users to create practice questions or even real exams based on the lecture notes.

Whisper, OpenAI's speech-to-text transcription neural network, has numerous use cases in the field of higher ed. Some of the potential applications include:

1. Lecture Transcription: Whisper can transcribe lectures, allowing students to review and study the material at their own pace, as well as share notes with their peers.

2. Accessibility: For students with hearing impairments, Whisper can provide real-time transcriptions of class discussions, making the learning experience more inclusive.

3. Language Learning: Whisper can be used to transcribe audio from language classes, providing written material for students to practice listening, speaking, reading, and writing in their target language.

4. Meeting Minutes: Teachers, administrators, and other staff can use Whisper to transcribe meetings, ensuring that accurate records are maintained and important information is not lost.

5. Study Groups: Students can record and transcribe their study group sessions, allowing them to review their discussions and better prepare for exams or assignments.

6. Oral Presentations: Teachers can use Whisper to transcribe students' oral presentations, providing a written record for grading and feedback purposes.

7. Tutoring: Tutors can record and transcribe tutoring sessions, helping students review the material and track their progress over time.

8. Interview Transcription: Whisper can transcribe interviews conducted for research projects or assignments, allowing students to analyze and reference the content more easily.

9. Podcasts and Videos: Educators can use Whisper to create transcripts of educational podcasts and videos, improving accessibility and making it easier for students to study the material.

10. Multimodal Projects: Students can incorporate Whisper's transcription capabilities into their multimodal projects, such as documentaries or video essays, to provide accurate subtitles or transcriptions for viewers.

These lists were generated using the aid of GPT-4. I’ve attempted to edit them, but I think these are a decent starting point in understanding at least some of the potential use cases generative AI may have in higher education.

---

AI Assistance is like Tracing in Art

When searching for an analogy to try and convey this particular approach to using generative AI, I think the most logical example is how tracing is used by artists as part of their practice. Most of us learned to draw in some way by tracing images, shapes, or designs that were not our own. For me, it was hands and fingers. I would trace the lines needed and incorporate someone else’s version of a hand into the larger drawing I was working on. In this way, having a writer call upon generative AI to help them with an issue in their writing is very much so like how artists use tracing within a drawing.

However, tracing isn’t without controversy and I imagine these debates will mirror criticism of AI assistance. For centuries, artists have argued about the legitimacy of tracing someone else’s work and passing it off or incorporating it into something new. Charges of theft, hackery, and incompetence abound, yet, millions of artists use tracing as part of their daily practice. We are seeing similar debates about the use of generative AI in written work now and these should engage nuance. A student using a small text generation to help with a research question or counterargument or rewording a clunky thesis is a far cry from prompting ChatGPT for an entire essay. The former is like tracing a few lines to augment a greater artistic work, while the latter is no different than going into a second-hand store, buying someone else’s painting, and putting your name on it.

The Debate—How Much Assistance is Too Much?

AI assistance is controversial, and like tracing, may always be so. It is very likely to reveal our own moral, political, social, and cultural biases about how much technology should be used to augment human creative work. No writer lives as a hermit within a cave. We learn from reading countless works and having others read and respond to our own output. In many ways, well-prepared students have repeatedly received outside assistance with their writing through tutors, intervening parents, and well-funded and resourced education. But less prepared students often have received far fewer benefits. These students likely stand the best chance of benefiting from structured AI assistance. It may well help bridge gaps in equity that higher education has been pursuing for decades.

While there are exciting possibilities to help students, there is no consensus about how much assistance should be allowed, nor is there clear agreement about what learning is lost when a student relies upon AI. When you think about the benefits of tutors, peers, parents, and resources, most of us tend to frame these as a helpful support mechanisms to guide emerging writers to develop the skills needed to be successful. Will AI assistance prove to be the same? For educators and students to begin to answer that question requires careful exploration.