A Few Thoughts on Generative AI in Education from a Recent Luddite![]() Over the weekend, I had the pleasure of presenting at the AI for Education summit. My talk was titled “Keeping the Human in Human-Experience Design,” and I focussed on some of the work that I am doing in my English classes and Design Thinking Club as I try to integrate responsible generative AI experiences into my students’ learning. It feels strange to be on the early adaptor side of this conversation because for most of my teaching career I’ve been demonstrably low-tech. I was one of the last of my friends to buy a cellphone, I’m not on most social media, I only recently got a smartphone, and I once gave a PD session in defense of paper and pen. But when Chat-GPT3 launched, something felt very different to me. Like many, I was amazed at what a word predicting tool could conjure up, and then I was horrified to see that it could competently complete a good number of the writing assignments I gave. But then I returned to amazement at how it could use backwards planning to give me a solid work schedule, how it could create diverse personas to give feedback on a design proposal, and how it could help me outline the plot beats for a screenplay I was working on. And then I started feeling overwhelmed by both the possibilities and the number of AI tools that were entering the market, the prompt libraries, the ethical considerations. I wondered too whether in my excitement over what AI could do, whether I was losing sight over what was actually in the best interest of my students’ learning. And then I discovered that I was in good company with these feelings because as far as AI in education goes, there are no experts. As ChatGPT-3 was released just before my daughter was born, I’ll be watching them grow up together. It’s birth offers a schism to my work life that she offers to my family life. And for those of you with kids, you know full well that everything changes. Sometimes while rocking her back to sleep after a bad dream, I’m reminded that ChatGPT will be the lowest form of technology she will ever interact with, and then it’s my turn for a bad dream, trying to puzzle together what her life will be like, and then I give a chuckle at my seven-years-old self who was once mesmerized by the capability to manipulate crude white pixels on a screen the first time I played Atari’s Pong. What must my parents have thought as they watched me practice such sorcery? While we are still in the early stages of integrating this new technology into our workflows and lives, I thought I’d share a few observations I’ve made from early adoption. Large Language Models (LLMs) like ChatGPT, Gemini, and Claude offer vast opportunities for enhancing learning, but they also present unique challenges that need careful consideration. To address these challenges, I’d like to offer two problems that I’ve encountered and three solutions that I am currently working with. Problem 1: Attitudes Towards Generative AI Many people don’t really understand what this technology can do, nor do they understand what they can do with it. Part of the problem with this misunderstanding is the affordances that the tool suggests. It is such a general tool, and it looks like Google, so it is easy to just treat it like a search engine and dismiss it the first time an inadequate or hallucinatory outcome is received. Conversely, the magic of those initial outcomes is problematic because people tend to accept those initial outcomes simply because they are created much faster than they could create them themselves. Those early magical returns can be tempting to submit for assignments, particularly if the individual lacks the skill or motivation to actually produce the given outcome. Additionally, the quality and speed of the outcomes can demotivate the acquisition of the given skill because if a tool can do this for me, why should I bother learning how to do it myself? Lastly, when people use the tool, they are reluctant to admit doing so for fear of the attached stigmas. If I am using the tool to produce outcomes I did not produce myself, then that must mean that I do not have the ability to produce said outcomes. Also, since we are still wedded to the idea that good work must take a considerable amount of time, the fact that this tool allows us in many cases to work faster might create the impression to others that we are not working hard enough. Problem 2: Human Collaboration and AI Another significant issue is the impact of LLMs on collaboration. When individuals interact more with screens than with each other, the essence of teamwork is lost. The focus shifts from engaging with peers to interfacing with a screen, potentially stunting the development of crucial interpersonal skills. As the main purpose of project-based learning is meant to be process, not product, carelessly outsourcing that work to LLMs would mean the loss of the soft skills that are normally developed during collective problem solving. Additionally, the quick returns that LLMs can offer lead groups to settle on initial solutions without going further. And it’s not just students who are prone to this; a recent study out of Stanford University showed that professional designers also succumb to this trap. As two-time individual recipient of the Nobel Prize, Linus Pauling, argues that “the best way to have a good idea is to have lots of ideas,” this is a particularly alarming observation. Solution One: Demystifying AI To combat these challenges, a step recommended by Stanford Design School’s Jeremy Utley is to have an emotional, interactive conversation with the tool. I’ve demonstrated conversations I’ve had with ChatGPT to my students that have really impressed me. I’m further impressed by the capabilities of ChatGPT-4, particularly the whisper function that allows me to record a rant or have a conversation about a half-baked idea that I would normally make my wife suffer through. Whatever I communicate is recorded, and then I can have my rambling thoughts reorganized into something more coherent to explore later on. Seeking solutions for our daughter’s education, I can have ChatGPT take on the role of three different personas asking me questions to help me better understand our circumstances. And while having interactive conversations like this have their own rewards, the greater value is the epiphany that comes from discovering some of the capabilities of this tool. If it helped me better understand the circumstances of my daughter’s education, might it be able to help me brainstorm ideas for next week’s research project, give me a novel hook for tomorrow’s lesson, or break me out of a creative rut for tonight’s dinner. Perhaps a useful way of thinking of Generative AI is as a precocious intern who can take on various and multiple personas and who is always available, but whose work needs to be checked. Solution Two: The FIXIT Framework Focussed Individual ConteXt Interactive Conversation Team Adopting the FIXIT framework can help reinforce new habits with this new technology. Individuals start with a focussed problem or interest. They then work individually on that focus. This is to ensure that their unique, authentic and human perspective does not get lost in the dynamics of AI or group collaboration. After the initial individual work, they then provide the necessary context to the generative AI tool. For this, there are numerous resources to share (here’s one). Alternatively, students can experiment with a trial-and-error approach. Once sufficient context is established, they then share their individual work with the LLM and engage in AI-assisted ideation through interactive conversation in order to enhance their ideas. Following this, they take their findings to their team. And, as is the case with the design cycle, the process repeats. This method encourages individuals to engage deeply with the material on their own before leveraging AI into their own work and then offering that to the team. It fosters a balance between independent critical thinking, AI-assisted critical thinking, and collaborative thinking and innovation. Solution Three: Encouraging Exploration People need to play with the tool and discover both what it can do and what they can do with it. As such, I’ve introduced a weekly prompt-generating competition that acts as "show-and-tell". Students are encouraged to use the FIXIT framework in their approach. It is low-stakes fun that allows for student exploration, and because it is self-directed, students take ownership of the experience without being held back by their teacher. Continuing the Conversation To me, the rapid advancements in generative AI that I navigate in my working life parallel the milestones my growing daughter introduces in our family life. Both scenarios are filled with excitement and anxiety, and both force us to continuously adapt. Just as my daughter shows us that no barrier is too high and no object out of reach enough, we should recognize that AI will similarly stretch the boundaries of what we thought possible. In the case of our daughter, my wife and I strive to foster curiosity, confidence, and independence, balancing our own anxieties with the need to let her explore safely. I believe we must approach AI with a similar mindset: embracing its potential while cautiously navigating its challenges. As we continue to refine our position and policies that will infirm our practices with generative AI in the classroom, I invite feedback and questions on these reflections. I also encourage you to conduct your own experiments and to explore. Let’s engage in meaningful conversations about how we can integrate this technology into our teaching, ensuring it enhances student learning in a balanced and thoughtful manner. If you are interested in learning more, I recommend the following resources: -How to FIXIT -Don’t let GenAI Limit Your Team’s Creativity–Harvard Business Review -Beyond the Prompt Podcast–Available on Apple, Spotify, and other Podcast Platforms -A Suggested Rubric to Communicate AI Expectations on a Given Assignment -Try Ten–An AI-assisted Ideation Tool in 5-minute Daily Exercises -Ethan Mollick and Lilach Mollick’s Paper on how to leverage AI tools into personalized learning experiences for students
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. Archives
July 2024
Categories |