Getting Started with Gen AITry Prompting On Your Own

Getting good results from a Gen AI tool starts with a few general tips for better prompting—like being clear, specific, and thoughtful about what you ask. These tips can go a long way in improving the quality of the answers you get. But if you’re still finding it hard to phrase your requests, getting irrelevant answers, or spending too much time refining your prompts, that’s when a prompting framework can help. Frameworks provide a clear structure to follow, making it easier to ask the right questions and get useful, accurate responses. In this module, enhance your prompting skills by mastering three keys to success, diving into two frameworks, and practicing with the examples below.

🌟3 Keys for Prompting Success🌟

When prompting, there are a few key elements to keep in mind:

1. Be specific, provide context, and include details

Specificity + Context + Detail = Better Output

Being specific in your instructions helps the chatbot generate more focused and higher-quality responses, rather than vague or general ones. A prompting framework can help ensure you add all the pertinent information you can in your prompt at the get go. Explore the CLEAR or RTRI frameworks below.

2. Always review the outputs

Recognize that Gen AI Chatbots may reflect biases and incorrect factual information or reproduce copyrighted material. Therefore, it is critical that you reflect on what those biases might be, cross-check results with trusted resources, and consider whether it is ethical to use, share, and distribute material that others may own.

Some questions to ask yourself when validating the output from a Gen AI Chatbot:

For Text Outputs:

Is the information factually correct and aligned with what I’ve found in reliable sources?
Does the output include proper citations for any facts or data presented?
Could the response be influenced by biases or stereotypes, and how might this affect its use in my work?

For Visual or Creative Outputs:

Does the image or visual content resemble copyrighted material, like famous artwork or logos?
Does the visual content appropriately reflect the cultural or historical context I’m working on?

For All Outputs:

Does the output represent diverse perspectives, or is it limited to the views of a specific culture, group, or ideology?
How does this output align with the professional integrity and ethical guidelines set by my organization or community?
Could using this output unintentionally mislead my audience or misrepresent the topic?

Remember…when using AI-generated content in your work, you bear full responsibility for its accuracy, quality, and appropriateness. This means you should carefully review and verify any AI-generated material before using it, just as you would with any other source. You are accountable for ensuring the information is correct and properly cited, and the fact that AI helped create it doesn't shift that responsibility away from you. It's important to treat AI as a tool that assists your work, not a replacement for your own judgment.

3. Provide feedback to the chatbot - Prompting is an iterative process

Give Feedback: Review and let the chatbot know what to improve by revising your prompts or asking the chatbot to revise the output. Tell it what needs to be added, adjusted, or removed.
Add Your Input: Incorporate your ideas, knowledge, or insights to make the response more complete and personalized.
Iterate as Needed: Remember this process is interactive, and mistakes will be made by both you and the model. Be inquisitive and critical, provide more feedback, keep trying, or try again until the output addresses what you are investigating.

Prompting Frameworks

We will introduce two prompting frameworks CLEAR (Concise, Logical, Explicit, Adaptive, Reflective) and RTRI (Role, Task, Requirements, Instructions) to help you develop prompts to explore.

⭐No matter which framework you choose to start with, both will help structure your prompts and ensure you include the essential information needed to get the best output from your Gen AI chatbot. Explore, play, and have fun!

CLEAR Framework

Introduction to CLEAR

One approach to prompting is to think about constructing your prompts through the CLEAR Framework illustrated below (Leo S. Lo (2023) The clear path: a framework for enhancing information literacy through prompt engineering, The Journal of Academic Librarianship, 49:4, 102720, DOI: 10.1016/j.acalib.2023.102720).

Getting Started with Prompting Using CLEAR Video

RTRI Framework

Another framework to try is the RTRI prompting framework.

As you begin prompting, your aim is to fill out each part of the acronym, answering the prompt guidance associated with the prompt component.

See the table below to explore each part of the acronym.

Prompt Components	Prompt Guidance	Example
Role	Who should the AI be?	“Act as a budget-savvy travel planner”
Task	What do you want it to do?	"Plan a fun and exciting spring break trip within the U.S.”
Requirement	What are the task parameters (i.e., include, remove, facts, etc.)?	"The budget is $1,000, including travel, accommodations, and meals. I prefer outdoor activities. I will be driving from Cleveland, OH. The trip should last 5-7 days. ."
Instructions	How should the AI act on the prompt?	"Provide the plan as a bullet point list, including sections for the destinations options, a sample itinerary for one destination, and a budget breakdown. "

Full Example Using RTRI Framework

Now we put it all together into a single prompt to give to the chatbot as:

Act as a budget-savvy travel planner. The budget is $1,000, including travel, accommodations, and meals. I prefer outdoor activities. I will be driving from Cleveland, OH. The trip should last 5-7 days. Provide the plan as a bullet point list, Including sections for the destination options, a sample itinerary for one destination, and a budget breakdown.

Using RTRI Framework with CWRU AI Demo:

Key Message

Regardless of the framework you use, any specific details you can provide will help Gen AI Chatbots produce an output that is more likely to address your prompt. The goal of these prompting frameworks, like CLEAR and RTRI, is to help us remember to add as much relevant detail as possible and make our prompting more efficient with few back-and-forth interactions.

Try On Your Own

Now that you have a foundation in prompting, explore and try the activities below.

Prompting for Planning

Using CWRU AI to try the below examples.

Attempt the cases below using your prompt. We have added our prompts as examples for you to compare.

As you work through each scenario, each introduces a more advanced prompting method. Read the tips in each example and remember to iterate your prompts as needed!

Scenario 1 (Beginner Level - Zero-shot Prompt)

You are an activities planner for a community center, and you want to brainstorm some activities for community members to do during the Spring. Use RTRI (Role, Task, Requirement, Instructions) to create the prompt.

Possible Prompt for Scenario 1

Role: Act as a creative activities planner for a community center.
Task: Brainstorm a list of engaging activities for community members to do during the Spring at the center.
Requirements: The activities should cater to varying skill levels, include educational and fun elements, and be feasible within a two-hour meeting timeframe.
Instructions: Provide at least five diverse activity ideas, each with a brief description, necessary materials or resources, and an estimated duration. Ensure the ideas encourage participation and collaboration among members.

Full prompt:

Act as a creative activities planner for a community center. Brainstorm a list of engaging activities for community members to do during the Spring at the center.. The activities should cater to varying skill levels, include educational and fun elements, and be feasible within a two-hour meeting timeframe. Provide at least five diverse activity ideas, each with a brief description, necessary materials or resources, and an estimated duration. Ensure the ideas encourage participation and collaboration among members.

Tip: Be as Descriptive as Possible

Notice that we can be very specific about the requirements here. For example, we wanted activities of varying skill levels to be educational and fun, fit within a two-hour meeting, etc. The point is you can add every little detail you think is relevant, and this will help the chatbot respond. This single descriptive prompt is called a zero-shot prompt. In other words, we are trying to complete the task in one single prompt.

Scenario 2 (Intermediate Level - One-shot/Few-shot Prompt)

You love traveling and want to create a personalized bucket list of destinations and activities to pursue during your college years and beyond. You want help in finding diverse locations, planning itineraries, and incorporating unique experiences that align with your interests and budget.

Use RTRI (Role, Task, Requirement, Instructions) to create the prompt.

Possible Prompt for Scenario 2

Act as a personal travel planner who specializes in creating personalized bucket lists. Help me create a personalized and comprehensive travel bucket list based on my preferences. Include a variety of locations and a mix of activities, and always consider budget and time constraints. Use the examples below to help generate the format of the bucket list, and before generating your suggestions, ask me for 3 interests, my budget, and the time I have available for travel.
Example 1: 
User Preferences:
Interests: Adventure, Hiking, Wildlife 
Budget: Moderate 
Time Available: 2 weeks 
Travel Bucket List:
Costa Rica – Arenal Volcano National Park
Activity: Hike the trails around Arenal Volcano, visit hot springs, and go zip-lining through the rainforest. 
Duration: 5 days 
Budget: Moderate accommodations and activities. 
New Zealand – Queenstown
Activity: Engage in bungee jumping, skydiving, and explore the scenic landscapes of the Southern Alps. 
Duration: 4 days 
Budget: Adventure activities priced accordingly. 
Kenya – Maasai Mara National Reserve
Activity: Safari tours to witness the Great Migration, visit Maasai villages, and enjoy guided wildlife walks. 
Duration: 5 days 
Budget: Safari packages with varied pricing options.

Tip: Give the Chatbot Examples to Enhance Your Answer

This is an example of a one-shot prompting strategy with RTRI. We have used one example to give the chatbot a better idea of the format for which I want it to construct its suggested travel bucket list. If you wanted to give the chatbot two or more examples, that is what is called few-shot prompting. This can be especially helpful if the output needs to be formatted in a specific way or if a chatbot struggles to solve a problem correctly. Giving it an example of what it should do or what the correct output is should help chatbots give better outputs!

Scenario 3 (Intermediate Level - Chain of Thought Prompt)

You are a member of a volleyball club, and you’re responsible for increasing the club’s visibility and membership through a strategic social media campaign. The campaign hopes to engage current participants, attract new members, and promote the club’s events and initiatives. Planning involves defining campaign goals, identifying target audiences, creating engaging content, scheduling posts, monitoring performance, and adjusting strategies based on feedback and analytics.

Possible Prompt for Scenario 3

Act as a successful and experienced social media marketer for a volleyball club. I'm responsible for increasing the club's visibility and membership through a social media campaign. Help me first by defining the campaign goals.
[Review and Validate response]
Great goals. Now, let's identify our audiences.
[review and validate response]
The audience for this campaign looks good. What are some ideas for creating content to engage these audiences?
[Continue prompting each part of the scenario and validating each out until you complete the entire task]

Tip: Break Down Complex Tasks Into Smaller Chunks

While you could prompt the chatbot to complete all the pieces of the entire task in one go, you would have little control if the chatbot made an issue in reasoning along the way. By breaking up complex tasks into simpler steps, you 1) can maintain control of the output of the chatbot 2) you can ensure the chatbot outputs are correct. This is an example of Chain of Thought prompting.

Tip: Prompt, Review, Provide Feedback

Sometimes, the output from a chatbot might not be exactly what you were hoping for. That’s okay—working with Gen AI is about providing more feedback. That is, refining and improving the output through this iterative process of prompting, reviewing the output, providing feedback to the chatbot, and prompting it again.

Prompting for Research

Take a look at these quick demos, follow along, and try prompting these with your topics

Prompting CoPilot to Refine a Topic Idea

Key Message

Crafting clear and concise prompts:
- highlights the practical application of the CLEAR and RTRI frameworks in crafting effective prompts for AI tools like Copilot.
- demonstrates how to combine these frameworks to create concise, explicit, and role-defined prompt.
Iteratively refining prompts based on initial results:
- an iterative approach to prompting, where the initial prompt generates a response that leads to further exploration and refinement of the research topic.
Combining AI with other research resources:
- the importance of combining AI tools with other resources, such as library databases, online encyclopedias, and expert consultation (librarians).
Critically evaluating and verifying information:
- the need for critical evaluation of AI-generated outputs, including verifying information and exploring the topic further using reliable sources.

Practice Your Skills > Research Rabbit

Most colleges and universities like CWRU maintain subscriptions to databases, which their students and faculty have permission to access while they study and work at CWRU. As a member of the community, you can visit the Kelvin Smith Library and access these databases while you are in our library, but if you are working or learning at home, these are resources that you would need to pay for to use. The good news is that some research is “open access”, which means that it is freely available online for anyone with an internet connection. One tool that the Kelvin Smith Library recommends to graduating students and members of the community, like you, is Research Rabbit. The prompting skills that you learned in the previous section can be applied when using a generative AI search tool like Research Rabbit.

Watch this short video to learn more about how to get started:

Key Message

Leveraging the strengths of library databases:

highlights the importance of utilizing library databases like Dimensions Research GPT for academic research.
emphasizes that these databases offer access to scholarly articles and resources not readily available through general online searches.
recognizes that library databases specialize in specific areas like scientific research, providing a more targeted and reliable source of information.

Accessing full-text articles through library subscriptions:

demonstrates the process of accessing the full text of an article through the library database, highlighting the benefits of library subscriptions for accessing scholarly content.

Next Step

With that, you have the basics you need to start prompting Gen AI Chatbot to empower your learning. Let us continue by considering the best ways we can use Gen AI chatbots responsibly in our learning.