Franklin Sayre

I’m the Makerspace Librarian at Thompson Rivers University and also Co-Chair the Librarians’ Department. These days I am particularly interested in creating active learning spaces that foster belonging and encouraging sustainability and critical thinking about technology and information.

Collaborating with AI Meta Workshop - Outline

The following is an outline for an experimental workshop on collaborating with AI for faculty, staff, and administrators at Thompson Rivers University.

[!NOTE] I used AI extensively in the creation of this workshop using more or less the same processes used in the workshop. Because of this, many parts of this workshop are purposely left at a draft stage a draft because I didn’t want my brainstorming, learning objectives, or initial prompts to be too polished, since that wouldn’t be an accurate set up for how most people will work with AI. AI was used to test the initial prompt, ensure it would consistently produce a similar outline, and then test the subsequent activities to ensure they would be stable enough to be cohesive while allowing for experimentation. At various points I used AI to get feedback and to find potential problems or inconsistencies.

Resources

Set-up

Individuals:

  • Name tags
  • Laptops or signed into a desktop
  • Signed into an AI model
    • ChatGPT (Free or paid)
    • Microsoft Co-Pilot (sign in with work account): https://copilot.microsoft.com/
    • Claude (Paid)
    • Local model via LM Studio (free and local install on Makerspace computers)
  • Everyone has a copy of this set of notes

Facilitator

  • Screen set up:
    • ChatGPT or other model on half the screen
    • A copy of this outline
  • Visual clock is set up and ready to go to help with time keeping

Learning Objectives and Introductions (~15 Minutes)

During this workshop we will be learning about generative AI by using it to create parts of this workshop as we go. We will do this by starting with learning objectives and a basic prompt, and then working together (and with AI models) to collaboratively refine and adapt the workshop content, exploring techniques like brainstorming, personas, generating feedback, and transforming content. We will use what we learn during this process to discuss what AI is good and bad at, and how it might fit into our work.

That this workshop uses the meta example of creating a workshop is not because I think AI is uniquely good at that or this is something we should all be doing. One thing I do believe is that you learn the most about AI by working with it in areas you have experience. If you work in post-secondary, in any role, you have some experience creating, teaching, or attending workshops.

Finally, what we build today will likely not be perfect (or even good) because that really isn’t the point. The point is to start exploring the boundaries of AI and how it intersects with our work and expertise.

Learning Objectives

  1. Demonstrate different techniques for prompting GenAI models, including brainstorming and generating content, getting feedback from various perspectives, and transforming content from one format to another.
  2. Analyze and explain how different prompting techniques affect AI-generated outputs.
  3. Critically evaluate the strengths and limitations of AI models for different types of content and tasks.
  4. Critically evaluate the role of expertise in the use of GenAI models, including for prompting and evaluating content.

What we are not doing today

  • Ethical issues
  • How AI works
  • Discussing regulations or guidelines

All these are important! But we have 1.5 hours and we are focused on active learning. Some of these should come up during the discussion section.

Roundtable: What is Your Experience with AI?

  • Two questions (2 minutes per person)
    • Name, department, and what your role is
    • What is your experience with AI
  • My (facilitator) key points:
    • Interesting technology. Numerous ethical concerns. Lots of naive criticism and boosterism.
    • I’ve found it very useful in my own work, but not really for “doing” my work.
    • I don’t want to use AI to do my work, I want to use AI to do better and more interesting work.
    • Students don’t know what it is bad at, faculty don’t know what it’s good at
    • Expertise is key:
      • No instruction manual. Nobody knows what this technology is good or bad at (or what kind of good or bad it is) without using it.
      • Exploring AI works best when working within a domain you know well. Your expertise and judgment are needed to distinguish what is interesting from what is boring or wrong.

Bootstrapping the Session: Starting Prompt (15 minutes)

This section starts with the group running the starter prompt to bootstrap the session and then working as a group to iterate on the results until we have an outline and some questions that we can use for the rest of the workshop.

What We want from this workshop

  • active learning
  • could work with faculty, staff, students
  • starts with roundtable where people say who they are and what their experience is
  • meets learning objectives
  • includes activities for each of the techniques listed in the LOs
  • includes discussions at the end for the role of expertise, strengths and weaknesses
  • Not in prompt
    • Brief intro section with overview of general rules for collaborating with AI and strategies (not included in initial prompt, so will need to be added later)
    • Number of participants
    • Length of Session

Starting Prompt

I am creating an active learning workshop for university faculty, staff, and students on how to use artificial intelligence, and specifically text generation or GPT models. Attendees might come from across the university, work in teaching, administrative, and/or instructional support roles. Since we can't know ahead of time, the session must start with a roundtable where people introduce themselves and what they do at the university and talk for 1 minute at most about their experience with AI. The learning objectives are: 1. Demonstrate different techniques for prompting GenAI models, including brainstorming and generating content, getting feedback from various perspectives, and transforming content from one format to another. 2.Analyze and explain how different prompting techniques affect AI-generated outputs. 3. Critically evaluate the strengths and limitations of AI models for different types of content and tasks. 4. Critically evaluate the role of expertise in the use of GenAI models, including for prompting and evaluating content. The rest of the session will involve participants using this prompt to build the workshop they are in. It must include activities for each of the techniques in the learning objectives, and discussions of the role of expertise, strengths and limitations for different types of content and tasks. Generate a draft outline of the workshop.

Discuss, Run, Iterate on Prompt as a Group

  • How did the prompt do? How close was the result to what we were looking for?
  • What didn’t it include that we wanted? What didn’t we include in the initial prompt?
  • What did it miss, get wrong, etc?
  • Did it do anything that was unexpected?

The initial prompt was purposely missing a few key details, and there are likely a number of other changes we will want to make to the outline.

At this point you could copy and paste the results and make those changes. You could also use prompts to ask the AI model to make these changes. Often I will copy to output and make some changes, and then give this back to the model with a new prompt.

For now, lets use prompts to ask the model to make changes.

Iteration Examples

The following are example follow-up prompts. The actual follow-up prompts will depend on what the model generated from the first prompt, which will vary every time.

A version of this prompt will likely be needed by everyone, since we didn’t include the number of people or the length of the session:

a couple changes. the workshop will have 6 participants and will be 1.5 hours. the workshop will involve the group creating the same workshop they are in, thus it will be a meta workshop. the prompt provided will be the same one I just provided. as such, the first activity on brainstorming and generating content will involve discussing, analyzing, and iterating on that prompt. re-draft the outline and questions based on this information

This example is more specific to what the model generates:

at the end of section 2, include 10 minutes for me to review general rules and strategies, making the whole section 20 minutes. for section 3, combine exercise 2 and 3 into one activity that covers both. remove section 4 "building the meta workshop since that is what the whole workshop is actually doing. add that time into the reflection questions and getting feedback section

Tips and Strategies for Collaborating with AI (10 minutes)

Before moving on to our active learning sessions, lets talk about some general tips:

Tips

  • Be Specific and Unique: Use your knowledge and expertise. Best way to avoid boring content and clichés by providing lots of context
  • Work Within Your Knowledge Areas/Expertise: Expertise lets you refine your prompts, make judgments, and learn what is actually useful
  • Think Like a Director: Give AI a role/persona and clear instructions
  • Tell it what you want it to produce: including output format, audience, tone, language level, etc.
  • Iterate Often: Evaluate, Incorporate, Refine through multiple iterations
  • Be Critical AI content is very verbally fluent. Be careful not to get pulled in by this and to ensure you are reading critically.

Strategies

  • Break It Down: Ask AI to follow specific steps (e.g., “propose a method of analysis, then critique that analysis, then conduct that analysis”)
  • Ask AI to Create and follow its own plan
  • Change Perspectives: Switch up perspectives to get different results
  • Use Examples: Provide examples of desired outcomes
  • Ask for Explanation: Prompt AI to provide its reasoning in more detail

Activities: Exploring Techniques (30 minutes total)

We are now going to work through a series of activities to explore different prompting techniques. Remember that the purpose is to explore! Keep the momentum going by taking what is interesting and iterating as much as possible.

We will do this in small groups, as individuals, and together as a large group.

Capture (using stickies, your phone, etc.) anything you notice that is:

  • Interesting
  • Concerning
  • Good
  • Weird
  • Wonderful
  • Etc.

So we can discuss it at the end of the session.

Brainstorming and Drafting Content (small Groups - 10 minutes)

In small groups (facilitator assigns)

Group 1: Apply and integrate some content into the prompt / outcome

  • Download the UDL Guidelines plaintext file (find file linked under Resources at the top of this file and paste it after a prompt asking it to integrate or analyze it in relation to the workshop in some way.
  • You can also upload the file with your prompt if that is supported, either way works
  • Iterate and ask questions of the model about how the workshop could be modified to integrate UDL guidelines.
  • Find 1 way that we can modify either the activity on transforming content or the discussion section that integrates in the UDL guidelines.
  • Paste this suggestion into the google doc

Group 2: Generate Discussion Questions for the next Section

  • Prompt the model to generate questions for the group discussion activity about learning objectives 2-4
  • iterate multiple times, take what is interesting, probe and change your prompt to get something more specific.
  • Aim for 2-4 questions
  • Paste what you find into the shared google doc.

[!IMPORTANT] Facilitator captures changes recommended by groups to next sections and notes them in running edits file, uses group AI chat to apply them to next 3 sections

Use Personas to Elicit Specific Perspectives (small Groups - 10 minutes)

In small groups (facilitator assigns) use personas to generate content from specific perspectives.

Persona examples:

  • Use yourself! Write a one paragraph about who you are and what your professional role is and then ask a specific question or for general feedback. Make this rich! Where do they work, what is their focus, what are they known for, etc? This is a place where your expertise helps you explore the boundaries of these models.
  • First year domestic student in a specific program.
  • International student whose first language isn’t English
  • Librarian with critical perspective on technology who believed education should be for liberation and who doesn’t believe AI can be ethically included in any aspect of education.

Example:

You are an academic librarian at a small Canadian primarily teaching university who adopts a critical (#critlib) approach to your work, believes education should support liberation, and doesn’t believe that AI can ever be ethically included in education. Generate draft questions for learning objectives 2-4 for the discussion section that can guide participants discussion

Group 1: Generate Discussion Questions Based on the Content for the next Section

  • Prompt the model to generate questions for the group discussion activity about learning objectives 2-4 by asking the model to adopt specific personas
  • iterate multiple times, take what is interesting, probe and change your prompt to get something more specific.
  • Aim for 2-4 questions
  • Paste what you find into the shared google doc.

Group 2: Ask for Feedback on the Workshop or Some Aspect of the Workshop

  • Adopting a persona, ask for feedback on the workshop as a whole or for a specific part. Use multiple personas and imagine multiple scenarios, e.g. students, faculty from different disciplines, asking for feedback before the workshop, after the workshop, etc.
  • Use follow-up questions to probe deeper
  • Ask for specific outcomes, e.g. types of feedback, a bullet list, etc.

[!IMPORTANT] Facilitator prompts participants to capture interesting observations or questions on stickies for 1 minute before we move on

Transforming Content from One Form to Another (big Group - 10 minutes)

As a group, we use prompts to transform the content we’ve created earlier from one format to another, e.g. to a draft slideshow, a plain language version, an instructors guide.

Whole group: We are going to take the discussion questions we generated and selected during the previous activities and ask the model to turn them into a discussion guide.

Discussion about the Strengths, Limitations, and the Role of Expertise (20 minutes)

As a group, we will work through either the following questions or questions generated earlier by individuals in the group?

Refer to Learning Objectives:

Analyze and explain how different prompting techniques/styles affect AI-generated outputs.

  • Refer to strategies and tips
  • Find examples

Critically evaluate the strengths and limitations of AI models for different types of content and tasks.

  • evaluate the strengths and limitations of the AI-generated outputs.

Discussion: Critically evaluate the role of expertise in use GenAI models, including for prompting and evaluating content.

  • Consider how expertise influences the effectiveness of prompting and evaluating AI-generated content.
  • Where was AI useful and where was human expertise crucial.

Wrap up (10 minutes)

Workshop-specific feedback:

  • What went well
  • What could have gone better?
  • What other questions do you have?

Updated: