Understanding Generative Artificial Intelligence

The purpose of this guide is to help you as a student writer to understand how GenAI works and how to use it.

ChatGPT, Google Bard, and Microsoft Gemini are all examples of writing tools using generative artificial intelligence (GenAI)—technology that can perform language processing tasks to generate new texts. While GenAI has the potential to enhance your learning, it is important to: 

  • understand the risks
  • consider the ethical issues of using these tools
  • exercise your critical thinking skills
  • make good choices

GenAI can mislead you and might get you into academic integrity trouble if you misuse it; however, when used strategically and appropriately, it can also help spark ideas, save time, support your comprehension of difficult texts or concepts, and provide useful examples. It is better to use GenAI outputs as a starting place for your thinking, research, and exploration, rather than an endpoint.

GenAI can’t support the development of your writing processes as well as working with trained writing tutors who are available to you through UVM’s Undergraduate Writing Center. Writing tutors can support you at all stages of your process, with all types of writing projects, in all disciplines—even when you are also using GenAI!

Where to Start: Follow Instructors Policies and Ask for Guidance

Body

When considering using GenAI for academic work, pay attention to the policies your instructors have established in their courses.

  • If you are not certain what the expectations are, check the syllabus and ask for guidance.
  • Even when it is allowed, there may be limits or expectations about how to give credit to GenAI.
  • If your instructors haven’t clarified their expectations, your questions may be a welcome invitation to collaborate and discuss the role GenAI will play in higher education and beyond. How will these tools be used in medical research, in the social sciences, in the arts and humanities, in business, in engineering, or in legal and government systems?

How GenAI Works

The Basics

Body

GenAI relies on large language models (LLMs)—a type of trained machine learning that finds patterns in massive amounts of textual input taken from the internet and predicts sequences of words. The LLMs that GenAI tools use have been fine-tuned to generate natural-sounding language.

 

Limitations

Body
  • Since this language-generating process is based on statistical probability and not an understanding of the material, the output can include “hallucinations”—false information that sounds plausible and is presented confidently.
  • False information may involve fabricated sources, non-existent scientific theories, incorrect definitions of terms, inaccurate summaries, and more.
  • GenAI can mimic and repeat the distortions and biases it has found from the data on the internet that was used to train it.

These limitations highlight the importance of exercising critical thinking, using research skills, and checking facts and references if you choose to use GenAI tools.

Other Risks and Ethical Concerns

Body

Most GenAI tools require you to make an account to use them, and you may first want to research the concerns that have been raised about: 

  • exploitative labor practices
  • environmental impact
  • ownership of training material
  • security and privacy user data

Harvard’s AI Pedagogy Project provides a starter guide about some of these issues and provides a ChatGPT interface you can explore without having to create your own account.

Writing GenAI Prompts

Body

Prompt engineering can make GenAI outputs more useful and allow you to practice thinking critically about your own “rhetorical context”—which includes your purpose for writing, your intended audience, your message or meaning, and the appropriate tone and style.

Don’t expect it to produce its best result on a first try; rather, treat it as a collaborator and assume the process will require more than one attempt. Guide it by telling it what was wrong or asking it to narrow down or broaden the results. Prompt engineering involves the strategic development of requests. Some examples are:

  1. Give the model an identity (“Play the role of …,” “Act as if you are a …”)
  2. Be specific about the length, tone, format, and genre you want.

Ask for multiple versions to get examples you can learn from or select from, which encourages you to be discerning and develop your critical thinking. (“What are three ways to revise this thesis statement?” “Based on these four body paragraphs, provide two versions of a concluding paragraph that suggests directions for further research.”)

Potential Uses

Brainstorming a topic
Body

You might ask about ideas or topics and use the output to further narrow or broaden areas you are interested in researching further.

Generating outlines
Body

GenAI can create outlines for writing tasks, writing project timelines, slide presentations, and any number of things. Giving prompts that generate several different versions of outlines that include, exclude, or prioritize information, can allow you to use the output as a starting point for constructing an outline that best suits your purpose.

Providing models
Body

If you are uncertain about a type of writing, you can request an example or description that can inform how you approach the genre or type of writing and can point out characteristics or features.

  • However, it may be more helpful to look at actual examples of the genre or use vetted writing resources to get information about conventions and expectations, like
    • Purdue University’s Online Writing Lab (OWL)
    • University of North Carolina at Chapel Hill Writing Center’s Tips & Tools
    • UVM’s Undergraduate Writing Center Tutor Tips
    • UVM writing tutors, who can guide you toward good examples and help demystify genre expectations and disciplinary writing conventions.
Understanding challenging texts or articles
Body

If you ask GenAI to explain what a text says in plain English and provide the text, it will generate a version that may be easier to understand for a non-expert than the original.

  • However, since GenAI does not understand or comprehend texts, relying on the simpler versions it provides can be risky. Instead, use the version GenAI creates to help with your comprehension as you re-read or review. Take your own notes as you reread the original to deepen your understanding.
  • In addition, be mindful that feeding texts to a GenAI tool without the creator’s permission presents potential copyright issues.
Assisting with revision, editing, and style
Body

While often simplistic and unoriginal, texts produced by GenAI are generally free of grammar or usage issues. You can give it your text and ask it to offer sentence-level improvements.

  • However, make sure you compare its suggestions to your original and make revision choices based on your meaning and intent.
  • Other AI tools, like Grammarly and Microsoft Editor, are often much better than GenAI for this purpose because their suggestions include explanations and provide opportunities for you to learn more about sentence-level issues and patterns in your writing.
  • Even better is working with a writing tutor and discussing which choices suit your rhetorical situation (your purpose, audience, meaning, context, etc.).
Generating headings and titles
Body

Give it your draft or a section of your text and ask for title and heading suggestions.

  • However, before adopting any suggestions, again consider whether they suit your rhetorical situation and your intentions, and make choices about whether to accept them, just as you should if the suggestions were coming from human readers.
  • Working with a writing tutor can help you think through your choices while considering your rhetorical situation. 
Providing questions
Body

If you are going into a situation where you will be asked or will have to ask questions, GenAI can suggest questions to expect or ask as well as offer responses to consider. (“I’ve applied for [a job, grad school, an internship] and have been invited for an interview. What questions should I expect? What questions should I ask?” “Quiz me about [a topic] and then give me feedback on my answers.”)

Suggesting wording for transactional communications
Body

Ask it to generate drafts of polite emails, meeting agendas, project timelines, blurbs to promote events, etc. Make sure you review and refine the results to match your understanding of your audience and purpose.

Citing GenAI

Body

Some organizations that provide citation style guidelines are starting to address GenAI:

Additional Resources

Body