Using Generative AI and prompt engineering tools such as ChatGPT to perform various L&D processes and tasks and to analyze data has been all the buzz in recent years. Several trailblazers in the L&D community have conducted intriguing experiments recently with ChatGPT specifically. For example:
- Josh Cavalier created 250+ prompts (and growing) to increase L&D creativity and productivity. Get the prompts at his website.
- Julian Davis created a tool to fetch and cleanse xAPI data before using it with his custom xAPI Report Assistant. Get the tool at the xAPI AI website.
- Bil Arikan created several prompts with ChatGPT to support his workflow for generating SMART KPIs that align with business goals and learning goals, then used those KPIs help inform his xAPI data collection strategy during an xAPI Cohort project we worked on together on in the spring of 2024. He also created an xAPI Statement Crafter using ChatGPT.
While these are all exciting uses of AI technology, the topics we've been most interested in are using AI tools for content generation and using AI tools for learning analytics and generating graphs and visualizations of xAPI data.
Using AI for learning analytics
We are curious about what the future may hold for AI's role in learning analytics, so we conducted several experiments. We will share an example of one of the prompts we used and the corresponding graph output from ChatGPT below. For the dataset, we exported a sample xAPI dataset from Veracity Learning and then uploaded it to ChatGPT. For the first prompt, we simply wanted to identify the 10 most active learners who were generating the most data.
ChatGPT prompt
Analyze this dataset and generate a bar chart showing the top 10 agents who most frequently appear in the xAPI statements, along with the count of how many times each agent was found. The count for how many times each agent appears in a statement should be indicated by the height of the bar, which corresponds to the y-axis on the bar chart. Each agent's name should be displayed in text beneath each bar in the graph. Use the value of the "actor.name" property to display each actor's name.
The result
Pretty impressive, right? This graph was generated in less than 20 seconds, but it is likely that the results were so good because we had an understanding of what to ask ChatGPT and we also have a solid understanding of the xAPI data model and the dataset. We also knew how we wanted the data to be visualized as a graph.
So back to the big question: Can you use AI to generate graphs & visualizations of xAPI data?
The answer is yes—but there are many caveats.
Here are a few simple tips for those caveats and effectively using AI tools with xAPI data:
- Garbage in, garbage out. When you have a quality xAPI dataset, the graphs rendered by generative AI tools like ChatGPT may closely match the ones provided by learning analytics platforms from Veracity and Watershed. The results of your AI prompts will only be as good as the veracity of your data. This caveat also highlights the importance of using an xAPI conformant LRS that passes ADL's LRS Conformance Test Suite.
- Know your data and your prompts. You must be extremely familiar with the data and the data model. You also need a solid understanding of how to craft detailed questions pertaining to the data in your prompts. And, you must have an idea of how you want that data represented statistically or visually.
- Size matters. Dataset file sizes are restricted in some AI tools. That means that the size of the dataset that you can import into a generative AI tool might be limited.
The other big question: Can you use AI & xAPI to create content?
When the question refers to using AI and xAPI for content creation, the answer is also yes.
This approach has been discussed extensively, but it's essential to focus not just on the technology, but also on the people you're training.
Technology is only one piece of the puzzle. The true key to success is understanding your learners—knowing their context, goals, and job responsibilities. By doing so, you can create more effective and relevant content, whether or not you're leveraging AI.
Today, advanced learning ecosystems integrate mobile learning, HRIS data, xAPI, and even sensor-enabled tools that give us a complete picture of our learners' daily activities. This rich data can serve as valuable input for AI-powered content creation, making the process more precise and personalized.
Use available data
Instead of relying solely on traditional methods like interviews or surveys, consider the wealth of data already available within your organization.
- Your HRIS system knows each employee's role, department, and reporting structure.
- Point of Sale (POS) systems, CRM tools, and ERP software track task durations and workflows.
- Support systems highlight common troubleshooting approaches.
These data sources, often accessible through APIs, provide essential context that can inform AI tools, enabling them to create content that is both timely and relevant to your learners' needs.
In this digital age, even human challenges can have technological solutions. Creating "one-size-fits-all" learning should be a thing of the past. Your existing xAPI datasets offer insights into strengths and areas for improvement, which can shape more effective content strategies.
Leverage modern tools—responsibly
When evaluating AI-powered platforms, the best tools today have memory—allowing them to remember key details about learners and organizations. This memory helps refine content delivery, adjust recommendations, and transform existing content to keep it relevant. Modern platforms also offer reusable prompts and AI assistants that simplify the more complex aspects of working with AI, like crafting prompts, managing workflows, and fact-checking content.
However, it's crucial to address concerns around data privacy and ownership. Make sure to review your software agreements to ensure you retain control over the data you're generating. If your systems store learner data externally, now might be a good time to reassess your terms to ensure you own and control your training content.
Here are a few tips to consider for effectively using AI tools with xAPI for content creation:
- Use existing data to enhance AI-powered content creation. Leverage HRIS, CRM, and ERP systems to provide real-world insights that inform training content, making it more relevant and targeted.
- Move beyond one-size-fits-all learning. With AI and xAPI, you can create personalized, dynamic learning experiences that respond to your learners' specific needs and behaviors.
- Be mindful of data privacy and ownership. Ensure you maintain control over your data by reviewing software agreements and understanding how AI platforms handle learner information.
AI and xAPI together offer a powerful combination for scaling and personalizing learning content. By leveraging the right data and AI tools, you can break free from outdated, generic training approaches and deliver highly relevant, engaging learning experiences.
Get hands-on experience using AI with xAPI at DevLearn
Hungry for more tips, findings, and or even some AI tips, tricks, and prompts that you can reuse for your own xAPI projects? You'll need to catch "Charting a Path to the Future of Learning Using xAPI & AI" with Chad Udell and Jason Haag at DevLearn 2024 Conference & Expo. DevLearn offers more than 100 sessions, plus pre-conference learning, DemoFest, and more. Hope to see you there!