EPFL School of Computer and Communication Sciences
@icepfl.bsky.social
640 followers 26 following 110 posts
Posts Media Videos Starter Packs
icepfl.bsky.social
🔬 Dr. Zhao’s team iteratively developed and refined a working prototype of multimodal learning analytics, testing it in classrooms across 56 sessions and planning further improvements in the next iteration to better support student reflection.
icepfl.bsky.social
🔬 It was a Multi-Year Research Project!

💡 The goal: give teachers and students evidence-based tools to reflect deeper and learn better together ✨
icepfl.bsky.social
🚨 The problem: teachers often lack evidence to guide student reflection during debriefs.

✨ Solution: Multimodal Learning Analytics (MMLA) automatically captures students’ actions and creates analytics to support reflection.
icepfl.bsky.social
💡 He talks about how reflecting on collaboration in classrooms is hard. Multiple events happen at once, and it’s easy to miss what’s important.
🎓 Teachers rely on experience, students rely on memory 🧠 both can be subjective and incomplete.
icepfl.bsky.social
🎙️ The second talk for Day 2 is given by Dr. Linxuan Zhao from Monash University, titled Designing Multimodal Learning Analytics to Enhance Face-to-Face Collaboration. ✨
icepfl.bsky.social
Don’t miss the chance to explore 23 years of groundbreaking research from the CRAFT and CHILI labs, built around two pillars of education research: discovery and invention.

📍 EPFL | PO 01
📅 Thursday, October 9
👉 memento.epfl.ch/event/lecon-...
Honorary Lecture – Prof. Pierre Dillenbourg - EPFL
memento.epfl.ch
icepfl.bsky.social
At EPFL, we don’t just study what works in education... we engineer new ways to make it work better. In his honorary lecture, Prof. Pierre Dillenbourg explores how the engineering mindset shapes education research, and how the complexity of human learning challenges even the most advanced algo. 🧑‍🏫✨
icepfl.bsky.social
By reinventing human-AI co-creation, Frederic Gmeiner discovered thhat even small interaction techniques showed how this can change how AI feels👩‍💻💫
icepfl.bsky.social
For blending learning with workflows, Frederic Gmeiner created ‘PointAloud’, a digital platform for architects.

This platform helped them transcribe their thoughts, and summarise them🚀✨
icepfl.bsky.social
He continues by proving that reflective questions help work more effectively, and could be designed with AI.
This AI could be developed with different agent styles such as asking questions, planning or sketching with suggestions.🤖✨
icepfl.bsky.social
First, Gmeiner shares to us how AI could help education and human learning in terms of interface🖥️📊

He does this by stating 3 grand challenges:

- Cultivating Cognitive Growth
- Blending Learning
- Reinventing Human-AI co-creation paradigms
icepfl.bsky.social
Now, Frederic Gmeiner from CMU will tackle how to design interfaces with AI for education!🎤🎙️
icepfl.bsky.social
Next, he focuses on workload 🏋️‍♀️. ECTS don’t tell the full story, he says, so he suggests combining surveys and ML to build better, more realistic course metrics. 📊
icepfl.bsky.social
After conducting a study, they observed a 25% increase in engagement and a 40% improvement in skill mastery. They also found that adding feedback further enhanced students’ motivation. 🌟📖
icepfl.bsky.social
He explains that in many fields, we already use data to make recommendations and drive improvement (for example, connected watches). His goal is to apply the same idea to education📚, using data-driven goals and rewards to encourage persistence and boost motivation. 🏆
icepfl.bsky.social
Coming up next in our workshop 🎤a fascinating talk by Conrad Borchers from CMU Effort-Sensitive AI for Learning: Measuring and Shaping Persistence. 💫🤖
icepfl.bsky.social
Metacognitive cues can boost critical thinking & engagement with AI tools
But timing and personalization matter ⏰✨
icepfl.bsky.social
Some learners didn’t find cues useful because of:
🙋‍♀️ High confidence in their own strategies
🤖 Preconceptions about GenAI’s limits
🌀 Cues disrupting thought flow

💡 Learners were more receptive when cues appeared:
🕐 Before asking a new question
📝 During pauses in note-taking
icepfl.bsky.social
✨ Learners who received cues showed:
📊 More active + critical engagement
🔎 Searched more topics (p=0.006)
❓ Asked more questions (p=0.053)
🔁 Showed more persistent inquiry, relevant follow-ups (p=0.047)
icepfl.bsky.social
To support learners, researchers developed Metacognitive Cues 🧩:
🔹 Orienting
🔹 Monitoring
🔹 Comprehension
🔹 Broadening Perspectives
🔹 Consolidation
icepfl.bsky.social
👩‍💻To use GenAI effectively, learners need Metacognition, the ability to:
Prompt thoughtfully, Evaluate outputs critically, Detect inaccuracies, Adjust strategies with confidence and flexibility
icepfl.bsky.social
💭 Dr. Singh highlights how Generative AI gives students rapid access to information, but also leads to Cognitive Offloading 🧠
icepfl.bsky.social
📚✨ This morning we’re kicking off our workshop:
Learning Sciences & Learning Technologies: Research Directions for the Next Decade 🚀
🎤 First up: Dr. Anjali Singh, University of Texas at Austin
Topic: Designing Technologies for Critical & Active Learning in the Age of AI 🤖✨
icepfl.bsky.social
This year’s Magistrale ceremony closed on a high note with an inspiring speech from Alex Margot, EPFL Alumni mentor, the perfect way to end such a meaningful day. Heartfelt congratulations to all IC graduates; your achievements are something to be truly proud of! 🎓💛

#epflalumni#closingnote#congrats
icepfl.bsky.social
And last, but absolutely not least, massive congratulations to Pedro Almendros (CS) for achieving EPFL’s 2nd highest master average grade, and to Salim Najib (DS), who received the Commune de Chavannes-près-Renens price for an outstanding student! 🤩💫