ºÚÁÏÉçÇø

• Composing Effective Writing Assignments
  • Examples of Effective Writing Assignments
• Scaffolding Writing to Support Student Learning
• Creating Assignments for Miami Plan Capstone Courses
• Teaching Literature Reviews
• Using Threshold Concepts to Design Assignments and Courses
• Teaching Grammar Rhetorically 
• Structuring Purposeful Group & Team Work
• Mentoring Graduate Writers

• Using ePortfolio Assignments 
• Giving Feedback to Writers
• Facilitating Meaningful Online Discussions
• Engaging Students in Effective Peer Response

• Providing Online Writing Support
• Integrating AI to Guide Learning
  • Teaching Resources and Lesson Plans

Introduction

The Research as Inquiry frame looks to help show that research is a process where you can develop a research  question and begin to ask more complex questions as the question is reframed after finding gaps in the  current information or research (Association of College and Research Libraries, 2016, p. 18).  Researchers have first to develop a foundational understanding of the topic before they can then identify  those gaps, arriving at a basic, fundamental, researchable question they can then use going forward.  

Other frames from the ACRL can also be folded into the lesson for this topic. One example would be with the  Authority is Constructed, discussing how LLMs and other AIs and how they reinforce bias along with how  sources are evaluated. Looking at how to create and modify prompts for these systems also gives a lead-in to  conversations around Information Creation as a Process and Scholarship as a Conversation.  

This lesson plan gives guidance on working with ºÚÁÏÉçÇøs to develop better skills in the formulation of  questions for research as well as the process of asking questions in a way that better promotes understanding  of the issues as well as further areas of study and research.  

Looking at what ChatGPT generates, starting with a simple prompt and then using the results to refine the  questions and prompts, allows ºÚÁÏÉçÇøs to gain a better understanding of how research is an iterative process  that requires evaluation of not only the original question but how that changes as you refine your sources  along with the discoveries you make during this research process.

Teaching Notes

While example prompts are provided for subjects of programming, math education, and astronomy,  you can use or adapt prompts as needed for any subject.  

Just remember to keep the initial prompt simple so that ºÚÁÏÉçÇøs can revise and reflect on how they  could be made better. 

• ChatGPT and other LLM tools are always in the process of being modified and improved. Expect that any prompt you give to an LLM could generate a different response each time the question is asked.
• This lesson plan is optimized for ºÚÁÏÉçÇøs in introductory courses but could be adjusted for any course  level.
• Access to the free version of ChatGPT and other LLMs could be limited so having a set of pre-made  prompts and responses could help limit downtime.

Learning Objectives

• Students will formulate research questions related to coding by engaging in iterative dialogue with  ChatGPT, allowing them to approach projects from a research perspective. 
• Students will communicate their coding solutions clearly and effectively by engaging with ChatGPT to review and revise their code and by using ChatGPT to identify potential areas for improvement and evaluate the generated code for completeness.  
• Students will understand the ethical considerations of using ChatGPT for research around issues such  as data privacy, security, and intellectual property rights.
• Students will understand the importance of ongoing learning in coding to discuss best practices and how they may be affected by ChatGPT.

Lesson Plan

Materials Needed 

Either have each group use ChatGPT to generate an answer of their choosing in the subject being discussed or  provide example code and the prompts used to generate those responses.  

Three example ChatGPT prompts and responses are provided below to be used to help with prompt creation  and provide pre-made examples for use.  

Introduction (15 minutes) 

Intro to the frame 

• Main idea - Research depends on complex questions that we may have to ask in different ways, so how  can ChatGPT and other Large Language Model (LLM) AIs help us to better focus on the concept or  problem we are trying to research or solve? 

What are the steps in this process? 

• What is the area we are trying to investigate, and what is the scope of the question?
• How can we break that question into more manageable parts that may be easier to understand and help  refine our question? 
• How can we use a tool like ChatGPT to help us with this process to make an excellent researchable  question or usable code?  

Considerations for using the Tool 

• Discuss an ethical framework for using this tool: When would it be helpful or not helpful to use this  tool? Who benefits from using it, and who is harmed by it? 
• Where does the tool get its sources from, and what is the best way to evaluate them? 

Working with the generated response (30 minutes) 

Using either a custom prompt or one of the example prompts provided below:  

• Learners work in groups/pairs/individuals to read the generated responses and evaluate them for  completeness and usability.  
• Learners will use their evaluation to prepare a brief overview of what was found and what needed to do to correct or improve the prompt. 

Questions to ask learners after generating responses to discuss together (15 mins discussion)

• What was the initial prompt you gave to the system?  
• Did ChatGPT generate the answer correctly? For example for a programming question, did it run  without any errors? Was generated answer correct for the topic asked?  
• What were the errors?  
• How could you revise the prompt for a better answer with fewer errors?  
• Could you tell where the system got its information from? Were any sources cited? What are the  implications of this?

Example Prompt 1

"Can you provide code for creating a bar chart in Python?" 

ChatGPT: 

"Certainly! You can create a bar chart in Python using various libraries, such as Matplotlib or  Seaborn. Here's an example using Matplotlib:

In this example, we define two lists: x for the categories and y for the corresponding values. Then, we use the plt.bar() function to create a bar chart by passing the x and y values as arguments. 

After that, we customize the chart by adding a title, x-label, and y-label using plt.title(),  plt.xlabel(), and plt.ylabel() functions, respectively. 

Finally, we use plt.show() to display the chart on the screen. 

You can modify the data and customize the chart further based on your requirements.  

Additionally, you may need to install the Matplotlib library if it's not already installed. You can use pip install matplotlib to install it."

Example Prompt 2

"I want you to act as a math teacher. I will provide some mathematical equations or concepts, and it will be  your job to explain them in easy-to-understand terms. This could include providing step-by-step instructions  for solving a problem, demonstrating various techniques with visuals or suggesting online resources for  further study. My first request is "I need help understanding how probability works." 

ChatGPT: 

"Sure, I would be happy to help! Probability is a fundamental concept in statistics that expresses  the likelihood that a specific event will occur. 

Understanding Probability:

1. Imagine you have a fair six-sided die. If you were to roll this die, the probability of any one side  coming up (say you were interested in rolling a 3) is 1 out of 6, often expressed as a fraction (1/6)  or about 0.1667 in decimal form or 16.67% as a percentage. This is because there is one desired  outcome (rolling a 3) and six possible outcomes (the six faces of the die). 
2. Basic Principles: 

• Probability always falls between 0 and 1 (or 0% and 100%). A probability of 0 means the event will  never happen and a probability of 1 means the event is certain. 
• The sum of the probabilities of all possible outcomes is always 1. In the die example, the  probabilities of rolling a 1, 2, 3, 4, 5, or 6 are each 1/6, and if you add these up, you get 1. 

3. Types of Probability: 

• Experimental Probability is based on the results of actual experiments. For example, if you flip a  coin 100 times and it lands on heads 55 times, the experimental probability of flipping heads is  55/100 or 0.55. 
• Theoretical Probability is what you expect to happen, assuming that all outcomes are equally  likely. The theoretical probability of flipping a coin and it landing heads is always 1/2, or 0.5,  because there are 2 possible outcomes and both are equally likely. 

4. Compound Probability: 

• If you want to find the probability of two independent events happening together, you multiply  their probabilities. For example, if you want to know the probability of flipping a coin and getting  heads (P=0.5) AND rolling a die and getting a 3 (P=1/6), you would multiply (0.5 * 1/6) to get  about 0.0833, or 8.33%. 
• If you want to find the probability of at least one of two events happening, you add their  probabilities. For example, the probability of rolling a die and getting a 3 (P=1/6) OR a 4 (P=1/6) is  (1/6 + 1/6) = 1/3 or about 33.33%. 

I hope this gives you a basic understanding of how probability works! Please let me know if you  want more detailed explanations or if you have other topics or problems in mind." 

Example Prompt 3

"What would happen if two supermassive black holes merged?" 

ChatGPT: 

"When two supermassive black holes merge, it results in a cosmic event known as a binary black  hole merger or coalescence. The process of black hole mergers is a fascinating and complex  phenomenon that involves the interaction of gravity and the distortion of spacetime. Here's an overview of what would happen during such a merger: 

1. Inspiraling Phase: Before the actual merger occurs, the two supermassive black holes start orbiting  around each other. This phase is called the inspiral phase, during which the black holes gradually  lose energy due to the emission of gravitational waves. These waves carry away energy and angular momentum from the system, causing the black holes to spiral closer together over an extended  period of time, which could be millions or billions of years. 

2. Merger Phase: As the black holes approach each other, their gravitational attraction intensifies.  Eventually, they reach a point where their event horizons, the boundaries beyond which nothing  can escape the black hole's gravitational pull, overlap and merge together. The merging process  releases an immense amount of gravitational wave energy in a very short span of time. 
3. Gravitational Wave Emission: The merger of two supermassive black holes creates intense ripples  in the fabric of spacetime, known as gravitational waves. These waves propagate outward in all  directions, carrying away a significant portion of the combined mass-energy of the black holes. The  emitted gravitational waves are so powerful that they can be detected by gravitational wave  observatories, such as LIGO and Virgo. 
4. Recoil and Ringdown: During the merger, the combined black hole undergoes a phase called the  "ringdown." The resulting black hole settles into a more stable state by emitting additional  gravitational waves. In some cases, the merger can also impart a gravitational recoil or kick to the  newly formed black hole, causing it to move through space at high speeds. This recoil is a result of  the asymmetric emission of gravitational waves during the merger process. 

The consequences of a supermassive black hole merger have profound implications for the  surrounding environment. The gravitational waves generated during the merger carry away a  tremendous amount of energy, which can impact the distribution and motion of matter in the  vicinity of the merging black holes. The release of energy from the merger may also lead to the  formation of jets and outflows of high-energy particles, which can have far-reaching effects on  the surrounding galaxy, including the regulation of star formation and the distribution of gas.  Understanding the dynamics and aftermath of supermassive black hole mergers is an active area  of research in astrophysics. These cosmic events provide valuable insights into the nature of  gravity, the evolution of galaxies, and the role of black holes in shaping the universe."

Introduction 

The "Scholarship as Conversation" frame focuses on understanding the discourse by which we come together  in the research community to negotiate meaning (Association of College and Research Libraries, 2016, p. 20).  Researchers are expected to seek out, critically evaluate, and correctly attribute previous perspectives on a  topic while also contributing their unique perspective at a level appropriate for the mode of communication  and audience. Researchers are expected to have “a fluency in the language and process of a discipline,” and  that lacking that understanding “disempowers their ability to participate and engage” in a scholarly  conversation.  

The "Information Has Value" and "Authority is Constructed" frames are closely related to this topic and can  also become part of class discussion. Information systems privilege authoritative voices, and that authority is  constructed (“various communities may recognize different types of authority”). Some communities and  scholars are underrepresented or entirely marginalized in academic discourse. This is carried over into the  databases used to build LLMs, reinforcing existing bias and inequities (Alba, 2022).  

The goal of this lesson plan is to work with ºÚÁÏÉçÇøs to increase understanding of the scholarly conversation  in their discipline of study, the rhetorical conventions, citation practices, etc. by comparing a standard  disciplinary article with an essay generated on the same topic by ChatGPT. Students will have the opportunity  to reflect on the conventions of their discipline and also the larger issues of bias, privilege, whose voices are  included in academic conversations and whether these same phenomena appear in text generated by  ChatGPT.  

Teaching Notes 

• Adapt prompts to your specific discipline/learning goals. 
• ChatGPT and other tools are being trained every time somebody gives it feedback on generated text.  Expect that over time it will be better able to replicate disciplinary scholarly conventions.
• Designed for ºÚÁÏÉçÇøs in introduction to the major courses (or the first class required of all majors).  

Learning Objectives 

• Students will critically evaluate content generated by ChatGPT in order to discuss the accuracy of the  information and disciplinary conventions of its presentation.  
• Students will examine modes of discourse in their discipline in order to evaluate evidence cited by  ChatGPT.  
• Students will understand how LLMs generate text in order to critically examine how individuals or  groups of individuals may be underrepresented or systematically marginalized within the system.  

Lesson Plan 

Part I 

• Break ºÚÁÏÉçÇøs into small groups 
• Introduce the concept that scholarship is a conversation among researchers that is ongoing and can be contributed to. Each discipline approaches this conversation differently based on: who the author is  addressing (audience), the subject of the conversation, and the intent of the conversation.  
• Provide groups an article from disciplinary journal, a class reading, or a seminal article in the field.
• Ask groups to analyze the article for disciplinary conventions.
  • Direct ºÚÁÏÉçÇøs to look for conventions of your discipline. [Citation style, in-text citations,  general structure of the article, rhetoric, etc.]
• Come back together to brainstorm what academic writing looks like in the selected discipline.
• Discussion question:
  • How do X [humanists, social scientists, scientists] communicate their research?
  • Where does bias appear in this system, and how might it get reinforced? 

Part II 

• ChatGPT prompt: Write an academic analysis of [topic of article used in Part I] citing 5 sources. Print  the generated text for ºÚÁÏÉçÇøs (avoid requiring ºÚÁÏÉçÇøs to sign up for ChatGPT, and to ensure that it  is optional).  
• Break ºÚÁÏÉçÇøs into small groups. 
• If doing this in a new class session, remind the class of the academic conventions discussed in Part I.
• Ask ºÚÁÏÉçÇøs to analyze text generated by ChatGPT.
• Discussion questions:
  • What scholarly conventions can ChatGPT follow? Which are missing?
  • Follow a citation. [This might require instruction in library resources, how to tell a book  citation from a journal from a book chapter].
  • Is the information cited presented correctly for the content it contains and for the disciplinary  conventions?
  • Where does bias appear in this system, and how might it get reinforced?

Introduction

The frame "Searching as Strategic Exploration" focuses on ensuring learners understand that searching is a  complex task that requires persistence and flexibility. Learners are expected to be able to use a variety of  tools, resources, and skills to help them find information relevant to their needs, as well as understand the  highly contextualized nature of information sources (Association of College and Research Libraries, 2016, p.  22) . 

This lesson plan leads learners through the keyword brainstorming and database searching processes, using  ChatGPT as a tool for topic exploration. Instructors will discuss the differences between GPTs and research  databases, the information they do and do not contain, and how they can be used (and misused!) to assist in  

the research process. Learners will use ChatGPT to generate a keyword list for a topic, and compare this list to  a human-generated keyword list. They will then have the opportunity to search a research database using  both lists and discuss how the search results differ. 

Due to the opacity in the way that ChatGPT and other large language models are trained, the frames  "Authority is Constructed and Contextual", "Information Creation as a Process", and "Scholarship as  Conversation" are highly connected to this topic and can easily be addressed as part of this lesson plan. A  question that is likely to come up is why learners shouldn't simply ask ChatGPT to provide a list of relevant  information sources (instead of searching in a research database), which will give instructors an opportunity  to discuss ChatGPT's "hallucinations" as well as the currency of its training dataset. If desired, instructors can  also use this as an entry point to talk about algorithmic biases and the white, Western-centric nature of its dataset.

Teaching Notes

• This lesson can be used in either introductory classes with a basic topic to help ºÚÁÏÉçÇøs better  understand how these research tools work, or in an advanced class with a more detailed topic to help  ºÚÁÏÉçÇøs gain a deeper insight into keyword generation.
• Instructors should test a sample prompt in ChatGPT (or other chosen AI tool) prior to the session to  ensure the tool behaves as expected and to scout potential issues and discussion points ahead of time.
• Instructors should not require ºÚÁÏÉçÇøs to sign up for an account with any specific tool due to ºÚÁÏÉçÇø  data privacy concerns. Instead, they should plan to either use their own account if desired, or create a  dummy account with shared credentials.

Learning Objectives

Students will be able to: 

• understand how library databases categorize article information in order to generate a basic list of  search terms relevant to a topic of their choice  
• understand the basics of how GPTs are trained and generate responses in order to identify gaps in the  search term suggestions they return 
• compare search results from AI-generated keywords and human-generated keywords in order to  critically evaluate the information and gaps returned from each

Lesson Plan

• Overview (2-3 mins)
  • Lesson should ultimately help you quickly and efficiently find relevant sources for your  research assignments
  • GPTs and research databases — uses and misuses for getting started with research
• What are GPTs? (5-7 mins)
  • What information do they contain?
  • What information do they NOT contain?
  • How is the information categorized?
  • Compare and contrast to search engines
  • Best way to retrieve information: targeted prompts 
• What are research databases? (5-7 mins)
  • What information do they contain?
  • What information do they NOT contain?
  • How is the information categorized?
  • Compare and contrast to research databases
  • Compare and contrast to search engines
  • Best way to retrieve information: keywords 
• Keyword generation (20-30 mins)
  • Basic process
    • Identify topic
    • Map out themes and synonyms
    • Use search engine to identify additional terms as needed
    • Connect with Booleans 
  • Basic process using GPT 
    • Identify topic
    • Create targeted prompt
    • Analyze response
    • Refine prompt if needed, repeat analysis
  • Activity
    • Divide into groups
      • half the class creates human-generated keyword list
      • half creates prompt for ChatGPT to generate keyword list
    • Discussion: compare the process and results from each group
      • What terms are the same across groups?
      • What terms are missing from either?
      • What challenges did you have and how did you overcome them? 
  • Discussion: Why can't we just ask ChatGPT to generate a list of references?
    • Frequent hallucinations 
    • Doesn't contain current information
    • Algorithmic biases 
• Database searching (7-10 mins)
  • Pick a database relevant to the topic
  • Activity
    • Compare actual results from each generated list
    • How relevant are the search results?
    • How could the search be improved? 
• Wrap-up discussion and questions (time remaining)

Subject-Specific Sample ºÚÁÏÉçÇøics

ChatGPT can be most easily queried using some variation of this basic prompt: 

" Please create a list of keywords that I might use to find research articles on the topic [insert  topic]."

Users can create more detailed and specific prompts if they like, but this simplistic prompt is more likely to be  similar to what undergraduates might use. 

Some possible topics, depending on the class: 

• graphite mineralogy (chemistry; geology) 
• ion transport in frogs infected with chytridiomycosis (biology; physiology) 
• mental health programs in k-12 schools and their impact on ºÚÁÏÉçÇø wellbeing (nursing; public health;  education)
• political commentary in the novel One Hundred Years of Solitude (literature; Latin American studies)
• political commentary in the play El Gesticulador by Rodolfo Usigli (theatre; Latin American studies)