AI and the Writing Process – Getting Started
Getting Started on an Assignment
Starting an assignment can be overwhelming. It can be difficult to know if you’ve understood the assignment instructions and expectations, and you may not feel familiar with the expected content or choice of topics.
Generative AI, such as ChatGPT, can be helpful with this stage if you use it carefully and ethically. The following examples demonstrate some ways of using ChatGPT to get started with choosing a topic, brainstorming, and creating a research question.
1. Choosing a Topic
Whether you need to better understand the relevant themes or issues of a topic, or you need some help figuring out what topic or sub-topic to focus on, ChatGPT is a helpful tool for starting this process.
Understanding a topic or concept
You can prompt GenAI to give you an overview of a topic or concept. It typically generates a succinct and accessible overview. Use follow-up prompts to expand on an aspect of the response; to ask about other perspectives, counter-arguments or exceptions; and to explore new directions.
Content generated is typically provided in a themed and numbered list.
💬 “Summarize the research on time-release medication delivery technologies.”
Not knowing the proper terms for these technologies might trip you up in a library search, but ChatGPT will often clarify common terminology, like in its response to this sample prompt.
💬 “Summarize the research on the effectiveness of CBT in treating depression.”
This prompt generates information about CBT that is mostly positive. A follow-up prompt can round out the information you receive:
💬 “What are the criticisms of CBT?”
Narrowing down or focusing your idea
Sometimes you have a general idea for a topic for an essay or project, but you need help narrowing down your topic, identifying a problem, or writing a research question. Asking ChatGPT to offer a list of things to consider can help you to learn more about your topic and find a focus.
💬 “What are relevant historical perspectives or analyses to consider for a history essay on the Roma in Europe after WW2?”
Note how specific this prompt is. Naming the discipline (history) and the project (an essay) helps ChatGPT to focus on relevant perspectives. The rest of the question identifies who/what (the Roma), where (in Europe), and when (after WW2) for further definition.
2. Brainstorming or Ideating with ChatGPT
By blending your curiosity and critical thinking with ChatGPT, you can use ChatGPT to help you brainstorm a possible approach to a project, report, or essay. For example, if you wanted to focus your engineering or environment assignment on highway pavement, a series of prompts can help you think through various approaches or angles:
💬 “What pavement technologies are used on highways to decrease the need for repairs in extreme climates?”
In the responses from ChatGPT, you might notice that a lot of technologies either reduce or repair cracking in pavement. This might lead you to a follow-up prompt like:
💬 “How does cracking in highways lead to their deterioration?”
ChatGPT generates a list of items and their descriptions. Items like “water infiltration” and “material erosion” might lead you to consider the broader impacts of cracking on the surrounding environment. A follow-up prompt might ask:
💬 “Does highway cracking have environmental impacts?”
In response, ChatGPT lists environmental impacts caused by cracking on highways, including pollution, erosion, wildlife impacts, and greenhouse gas emissions. You might then be curious about ways to have the lowest environmental impacts in remote locations by mitigating highway deterioration:
💬 “What are the most environmentally responsible mitigations for highway deterioration in remote locations?”
The generated list could give you a starting point for library research or provide you with another idea or question to pursue. The sky is the limit. Keep thinking creatively and critically, and then turn to the research to confirm directions and dig into new ideas.
3. Writing a research question
Writing a research question involves identifying a problem or gap that you can address through research and analysis. A research question is not a thesis. It’s a question that gets to the heart of what you are investigating, and it guides your research. Research questions vary by discipline and type of research. In general, research questions may:
- Investigate a connection or relationship between factors*
- *Relationships between factors can also emerge as you investigate the next types of research questions, but you may not be able to identify factors or their relationship in your initial research question.
- Seek to explain an observation, phenomenon, event, or problem
- Compare one thing to another
- Describe quantity, proportion, likelihood, frequency, pattern, behaviour, consequence, characteristic
Developing a Research Question
You develop a research question by first learning more about a topic or relationship, then narrowing to a particular element you are curious about and can investigate. Your initial engagement with ChatGPT may look much like the series of prompts used to narrow a topic, but your goal is to decide on a question you can investigate.
For example, if you’re generally interested in how daycare subsidies impact families, you could start with:
💬 “How do daycare subsidies impact parents?”
The results include a long list, including parental well-being, economic mobility, and reduction of financial stresses. You might then be curious about the impacts of daycare on children and families generally.
💬 “What are the impacts of daycare on children from low-income families?”
The results list a number of items, like structure, emotional and behavioral development, and nutritional support. By linking nutritional support to the economic allowances of daycare subsidies on parents, you might start thinking about a research question involving long-term healthy eating habits and food security due to daycare subsidization.
The problem you’re starting to define is whether daycare subsidization has a long-term impact on children and family health through economic allowances and behavioral / eating-habit changes in low-income families, which might lead you to this prompt:
💬 “Does daycare subsidization have a long-term impact on children's and families' health in terms of eating habits and access to healthy foods?”
The response outlines potential positive impacts on children and on families in two lists, and it lists factors that may impede or help produce these impacts. These lists may provide some parameters for a research question or some potential research directions.
There are two ways to proceed from here:
1. Ask ChatGPT to give you multiple examples of research questions that focus on questions or problems that interest to you. Make sure you identify the discipline you are working in, your course level, and the assignment:
💬 “Give five examples of research questions that investigate the link between childcare subsidies and family nutrition for a second-year university economics essay.”
💬 “Give five examples of research questions I can use for a fourth-year health sciences research paper on social determinants of health. My focus is on the impacts of daycare subsidization on families’ nutrition and eating habits. I’ll be conducting interviews with families as part of my research.”
Even though you specified parameters, you likely won’t use one of the examples generated because they will not be quite appropriate. But the examples can help you craft your own research question, perhaps by mixing different elements from the examples or using one or two as models. The examples may also introduce elements you haven’t considered, for example, that there are different kinds of daycare subsidies.
2. Proceed on your own with brainstorming and research. From your prompts so far, narrow down the topic, and go to the library to scan the research literature to see what work has been done, what other approaches have taken, and possibly what primary data are available to you. Draft a research question that zones in on what you are investigating.
4. Other ways to use GenAI at the start of the writing process
Understanding the assignment instructions
If you don’t understand your assignment instructions, you can use ChatGPT to find detailed explanations and examples of how to proceed with your assignment. Here are two examples:
Example 1: A kinesiology assignment asks you to create a detailed exercise physiology case study, but you’re not sure what that is or where to begin:
💬 “What is a "detailed exercise physiology case study"? What components should it include?”
In response, ChatGPT generates an ordered list of elements that a case study might include, with details for each point. Compare this list to your coursework, textbook, and the assignment instructions to confirm.
Example 2: You are tasked with writing a literature review and instructed to synthesize the research. You don’t know what this means:
💬 “In a literature review, I have to "synthesize the research". What does this mean? Provide an example of how to do it.”
In response, ChatGPT generates a list of ways you can work with research in order to synthesize it. Using an example topic, it shares examples for each item on the list. However, it does not give a full example of a literature review that synthesizes evidence. To see synthesis in action, you can use the first list to follow up with a prompt such as:
💬 “Write an example literature review paragraph with in-text citations on cellular respiration that compares and contrasts the literature and describes a gap in the research.”
For this type of prompt, choose a topic that you know something about so that you can compare what you know with the GenAI output. Avoid using the topic you’re writing about. If you use your assignment topic, it will make the task of writing your own literature review difficult as you try not copy the example. There will also be issues with the accuracy of the paragraph generated and the validity of the citations.
Instruct ChatGPT to include in-text citations, which are necessary to a literature review, or it will not include them. Example citations demonstrate how you might refer to specific studies and research as part of your synthesis work, but they will likely not be accurate or real research sources. If they are real, ChatGPT may not accurately represent their findings or claims.
Again, resist asking for examples based on your actual assignment topic to help you manage your academic integrity.
Creating a list of keywords or search terms for research
Keywords are words or phrases that are central to the topic. For example, start-ups can also be called small businesses or micro businesses. Search terms are longer phrases that bring your key words together to get the specific search results you need, for example: “micro businesses and impact of government bursaries in India.”
When you conduct library research, keywords help find academic information and help you verify content produced by ChatGPT. Try a variety of keywords in different search terms to conduct a thorough search with relevant results. Keep track of the keywords and keyword combinations you use.
Different words return different results. Because it can be difficult to generate synonyms or alternate words on your own, GenAI can create lists of keywords for you:
💬 “Generate a list of alternative keywords for "start-ups" to use in a library search.”
Or you can use ChatGPT to help you consider new search terms:
💬 “For a library search, generate a list of search terms on the effects of business subsidies on micro start-ups in India.”
ChatGPT generates a long list of potential search words and phrases that you can use. Additionally, library databases have significant functionalities you can use to narrow down your search and exclude unrelated results. Ask the Library about strategies for conducting a search for your assignment.