CALMet Commons

 

OEB Conference 2025: From AI Experimentation to Practice-Ready Learning Design

 

In December 2025, we attended the Online Educa Berlin (OEB) conference, one of the key international meeting points for digital learning, instructional design, and educational technology. This year’s conference was framed by the overarching theme “Humanity in the Intelligent Age: Empathy, Responsibility, and the Duty of Care.”

Figure 1: Opening Plenary at OEB (source: oeb.global)

Rather than treating generative AI as a purely technical innovation, OEB 2025 consistently positioned it as a human and ethical challenge for learning professionals. Across keynotes, workshops, and panel discussions, the focus shifted towards questions of responsibility: How do we design AI-supported learning experiences that respect learners’ agency, diversity, and cognitive limits? Where does human judgement remain indispensable? And what does duty of care mean in an educational landscape increasingly shaped by intelligent systems?

Figure 2: When to use AI, Opening Plenary at OEB (source: oeb.global)

Against this backdrop, the conference offered numerous practical examples showing that meaningful use of AI in education depends less on the tools themselves and more on the pedagogical, organizational, and ethical frameworks in which they are embedded.

 

Selected Conference Highlights

Vibe Coding and Generative AI for Learning Design

The workshop on Vibe Coding explored how educators can prototype functional learning applications using generative AI. Participants were encouraged to start with a very narrow scope, iterate quickly, and refine outputs through continuous interaction with the system.

Several AI-supported tools were tested, including a custom GPT designed to support educational app creation and low-code platforms such as PartyRock. A key insight was that successful results depended far more on pedagogical clarity than on technical expertise. AI proved most useful when treated as a design companion rather than an automated solution.

Figure 3: PartyRock is a space where you can build AI-generated apps in a playground powered by Amazon Bedrock.

 

Experiential Learning and Roleplay Generation

One session connected AI-supported design to Kolb’s Experiential Learning Cycle, particularly through AI-generated roleplays and scenario-based chatbots. Prompt builders for roleplay design demonstrated how clearly defined roles, objectives, and reflection phases significantly improve learning quality.

The main takeaway: AI-generated roleplay is only effective when the experiential learning structure is designed first.

Figure 4: Kolb's Experiential Learning Cycle
(source: https://www.earlyyears.tv/david-kolb-learning-styles-cycle/)

Educator AI Literacy and Participatory Design

Teachers and trainers often approach AI outputs with skepticism, concentrating on their limitations while underusing the tool’s interactive potential. The reflections and discussions that emerged during this workshop highlighted the need to raise awareness and encourage deeper dialogue with AI tools. Through an iterative process of refining chatbot outputs, teachers and trainers can achieve results that not only meet their expectations but are also richer, as they emerge from a collaboration between the speed and power of the machine and the human capacity for conscious knowledge and reasoned judgment.

Lessons Learned for Practice

Expertise Remains Central

AI does not replace subject-matter or pedagogical expertise—it depends on it. Educators with strong domain knowledge were consistently better at evaluating AI outputs, identifying distortions, and refining results.

Motivation Is a Design Responsibility

Several panels highlighted that low learner engagement is rarely caused by a lack of motivation alone. Engagement increases when learning experiences:

• are grounded in realistic scenarios,
• respect cognitive load (avoiding “fire hosing”),
• focus on one essential learning takeaway, and
• create emotional relevance through storytelling.
  

Figure 5: Elements to consider for learner motivation (source: chatGPT)

Implementation Requires Organizational Support

Learning cafés and discussions confirmed a familiar pattern: when institutionally provided tools are underused, the issue is often communication, perceived relevance, or missing management backing.

Successful implementation of educational technology depends on:

• interdisciplinary collaboration,
• visible and measurable value,
• agile pilot phases,
• communication of success stories, and
• clearly defined roles for learning experts as connectors between stakeholders.

Figure 6: a framework to support reflection on EdTech implementation
(source: https://te-learning.nl/impression-what-really-makes-edtech-2-0-stick-in-practice-oeb25/)

Inclusion and Accessibility Must Be Designed In

Sessions on inclusive and equitable AI emphasized that generative AI can meaningfully support neurodiverse learners—but only if accessibility is addressed from the outset. AI tools showed promise in supporting reflection, personalized pacing, and social learning, provided that human oversight remains central.

Relevance for the CALMET Community

For training organizations in meteorology and climate services, OEB 2025 reinforced that AI should be approached as a design material rather than a shortcut. The most convincing examples used AI to:

• prototype learning scenarios efficiently,
• support reflective and experiential learning,
• personalize feedback without lowering standards, and
• strengthen professional judgement rather than automate it away.

The overarching lesson is clear: start small, stay critical, iterate deliberately, and retain responsibility for learning outcomes.