Human-Led AI: Designing the Road Without Dictating the Destination with Goldie Love

Generative AI can produce learning content in seconds. But producing content is not the same as designing a meaningful learning experience.What happens when AI provides an answer before the learner has had an opportunity to reflect? When generated activities assume a particular outcome? Or when efficiency begins to replace instructional intention?This session explores the instructional designer's responsibility to guide AI-assisted development without surrendering professional judgment.Drawing on practical experience developing reflective, self-guided learning experiences, Goldie Love will demonstrate how human-centered design principles can shape the way instructional designers prompt, evaluate, and revise AI-generated content.Participants will examine how sequencing, scaffolding, language choices, and learner autonomy influence instructional quality. Through practical examples, the session will illustrate how AI can support the design process while the instructional designer remains responsible for determining when to guide, when to question, and when to create space for the learner.The central message: AI can help us build the road, but instructional designers must ensure that learners retain ownership of their destinations.

Luis Malbas: Alrighty. Looks like the session is live. Welcome, everybody. Thank you so much for being here. This is the final session of our Human-Led AI for Instructional Designers event. We started the event on Monday, today is the final day, and I think it's been a fantastic event. I've learned so much from so many folks.

Before we get started, I want to mention a couple of things. Our next event is going to be taking place December 14 through the 17, and it's called the TLDC Change Summit. What I'm trying to do, as I navigate through L&D right now, is find answers about the overlap of AI, change management, and L&D. That is what the event is going to be about. I've posted the registration link and the Call for Speakers form in the chat. It'll be free, just like this one, and hopefully, it will benefit lots of L&D folks out there.

Moving on, Goldie Love, thank you so much for being here today to close out the event. Goldie is the founder of the Amore Group. Bringing over 18 years of experience across K-12 education and corporate L&D, Goldie is a learning strategist dedicated to human-centered instructional design and reflective learning. As generative AI transforms content development, she is a fierce advocate for preserving instructional judgment and learner autonomy. Guided by her philosophy, "I'm not the destination, I'm just creating the roads," she is here to help us pave the way forward. Everybody, please welcome Goldie Love. With that, I'll hide myself from the screen and let you take over, Goldie.

Goldie Love: Thank you, everyone. Welcome. We've had an amazing conference thus far. Luis, thank you for putting this together and offering our community of L&D practitioners a space to learn, grow, and share. I hope this afternoon I am able to share something you can use in your practice as you navigate new projects and incorporate into your workflow.

My name is Goldie Love, and my professional journey has taken me from K-12 education to corporate learning and development. Across these varying settings, one question has consistently guided my approach to learning: I focus on intentional learner outcomes. It is not simply about what information we're delivering, but what learners should be able to think through, decide, or do as a result of the experience.

We've all heard the conversation that AI is going to take all of our jobs. While AI is certainly changing how we work, I think that statement overlooks something important about the work we do as instructional designers. AI can generate content, suggest learning activities, develop assessments, and even recommend instructional strategies. But generating those elements isn't the same as exercising professional judgment about how they should come together to serve learners. AI can help us consider possibilities, but it doesn't independently carry the responsibility for understanding our learners, navigating organizational context, or determining the most appropriate learning experiences.

As AI becomes increasingly integrated into everyday life and instructional design, we need to ask ourselves: are we applying the same level of intentionality and instructional judgment to AI-generated content? Or are we allowing the efficiency of generating content to overshadow the purpose of the learning experience? I can't tell you how many times someone has said, "Just put it in chat, and it will tell you what to do." When we focus solely on crafting the perfect prompt, we risk bypassing critical thinking and accepting AI-generated responses without thoughtfully evaluating them.

How do we take advantage of what AI makes possible without becoming overly dependent on it? Have you ever prompted an AI tool, looked at its response, and thought, "That's pretty good, I can use it, adjust a few words, or prompt it again?" I want us to consider whether we are treating generative AI like a slot machine. Are we putting in a coin, pulling the lever, accepting the output, and designing plug-and-play learning? Efficiency is valuable, but what happens between receiving a polished output and deciding it's ready for learners?

When I talk about human-led AI, I'm not suggesting that an ID has to manually complete every task. There are real benefits to AI. I'm suggesting that we remain responsible for the instructional decisions we make. AI might generate five learning objectives, three activities, and an assessment in seconds, but we still have to determine whether those objectives reflect the performance needed, whether the activities provide appropriate practice, and whether the assessment gives us meaningful evidence of learning. That's instructional judgment. The prompt generates possibilities, but our judgment determines what serves the learner.

When evaluating an AI-generated learning activity, I start with alignment and meaningful practice. Does the activity support our instructional objective? If learners need to demonstrate a skill or make decisions, are we giving them opportunities to do so, or are we simply asking them to recall information? Are learners applying what they learned, thinking through situations, or receiving useful feedback? We're not just evaluating what AI generated; we're determining whether it actually serves the learner.

My philosophy is that I'm not the destination; I just create roads. My role isn't to dictate every conclusion a learner must reach; it's to intentionally design pathways that encourage exploration, critical thinking, and informed decision-making. AI should not replace our instructional judgment about when learners need support, where they should have choices, and which boundaries must remain in place. AI changes how quickly we can generate content, which requires humans to be intentional about evaluating what AI produces.

This brings us to the three actions that help us apply instructional judgment when working with AI: Reflect, Revise, and Release. These actions aren't intended to replace the ADDIE process; they offer a practical lens for evaluating and refining AI-generated learning experiences. It's essentially checking AI's blind spots.

First, Reflect. When AI gives us an activity, our question isn't just "Do I like it?" but "What is the learner actually being asked to do?" Imagine a module for a newly promoted manager conducting a difficult performance conversation. AI might offer five communication strategies and a knowledge check. Knowing five tips is not the same as conducting difficult conversations. Our design action is simple: compare the learning objective with the behavior the activity requires. If they don't match, we have an instructional gap. We're not simply examining what AI has generated; we're weaving what learners already know into what we want them to know. Reflection is about recognizing what AI may have overlooked, misunderstood, or omitted.

Next, Revise. Once we've reflected, we revise. We don't need to discard the response or start over, because starting over can confuse AI. Instead, we build on what's useful while addressing what AI missed. Did it address everything we asked? Did it maintain the intended learner outcome? Revisions give us an opportunity to redirect AI, reiterate omitted outcomes, and refine the response. The goal isn't simply to get a different response; it's to guide AI toward a response that accurately reflects our instructional decisions.

Then there's Release. Release is where we begin to trust the intentional work we've done. We're the experts. We've taken the time to examine what AI generated and clarify assumptions. Now we're in a better position to move forward. Release is about trusting the intentional decisions we've made and recognizing when we've provided sufficient direction, refinement, and structure to move the learning experience forward.

Let's see how these three actions work together in a scenario. Imagine we're developing a 20-minute eLearning module for newly promoted managers on addressing employee performance concerns. Our prompt asks AI for communication strategies and a knowledge check, and we include specific context about the work environment. AI gives us five relevant strategies and a knowledge check, but it misses the specific context we included.

Instead of starting from scratch, we use Reflect to catch what was missed and Revise to clarify our expectations. We give AI a follow-up prompt that reinforces those expectations, keeping what works and clarifying what doesn't. We use keywords like "retain," "revise," and "preserve." Then, Release means we review the new response to verify it is ready to move forward. We didn't change our instructional objectives; we simply used our professional judgment to guide the AI.

Before we release AI-generated learning, we have to ask ourselves five questions:

  1. What assumptions has AI made? (Identify missing context or oversimplifications).
  2. Where is human judgment essential? (Identify decisions requiring professional expertise or ethical consideration).
  3. What needs clarifications or revisions? (Guide AI with specific instructions rather than starting over).
  4. Are we preserving meaningful learner choices? (Ensure pathways support autonomy).
  5. Have we verified the revised output before release?

AI can help us create the experience, but we are responsible for determining whether it's ready for the learners. Remember, we want to let AI accelerate the work, but human judgment guides the learning. Are you pulling the lever and hoping for a win, or are you designing the learning? We must reflect on the purpose, revise what needs strengthening, and release an experience that allows learners to grow. AI may help us build more roads, but the responsibility for meaningful learning still belongs to us. Thank you!

Luis Malbas: Thank you so much, Goldie. Do you have time for a couple of questions?

Goldie Love: I absolutely do.

Luis Malbas: This one is from Rachel: "When I prompt Gen AI for larger and/or more complex assets, I find that the output tends to be very repetitive to the point of sounding like fluff. Have you noticed this, or do you have any advice on how to avoid this?"

Goldie Love: Yes, you have to make sure that you break it down. Generative AI is much like consuming food. If you take something very large and stuff it in, you're going to choke. If you feed it a bunch at one time, it's going to give you anything to get it out because its core goal is efficiency and speed. Break it down into smaller components to ensure that, as you build upon your final objective, those pieces have been truly evaluated and understood by AI.

Luis Malbas: Here's one from Geralyn: "As we start to use AI agents, do you see this approach on outlining being tweaked at all?"

Goldie Love: I have concerns with agents. Most people use an agent to quickly mass-produce outcomes for repetitive tasks. What happens is it removes the nuance, and efficiency becomes more important than the value of the information requested. Even with agents, it needs to be constantly revisited and checked to ensure the agent you built is still useful for where you're going.

Luis Malbas: Is there any one thing that really stands out to you as a failure that made you need to look at AI in the way you are now?

Goldie Love: Absolutely. I was working with school districts to teach teachers how to use AI to generate prompts for learning. I realized I was putting together presentations very quickly, but it completely left out the most vulnerable set of students—students with special needs, disabilities, and multiple learning styles. It was an eye-opener because I was presenting information without a true human component, not looking at assumptions and misunderstandings. I remembered what an instructor told me in undergrad: "You don't teach computers, you teach kids." I was instantly teaching computers.

Luis Malbas: Interesting. It's great to hear the background to how you came across all of this wisdom. Thank you so much for this presentation. Everyone, I pasted the link to Goldie's handout in the chat. We've had a nice long session, Goldie. I really appreciate your time. For everybody here, don't forget our next AI-related event happens in mid-December. With that, I'm going to end the session.

Goldie Love: Thank you all. Please connect with me on LinkedIn. Have a great week!

‍

Key takeaways

Similar videos