← AI for Educators Daily with Dan Fitzpatrick
AI for Educators Daily with Dan Fitzpatrick
It's all about trust for both students and teachers
00:15:29
Loading summary
Transcript1 lines
- [00:00]
A
If this episode makes you think, please let us know in the comments and support us by subscribing and leaving a review. Thank you. Today we are exploring a fascinating study that uncovered significant mismatches between what students and teachers in Germany think about artificial intelligence in their classrooms. Imagine a research paper from Tomohiro Nagashima, Lisa Sigrist, Nicholas Schultz, Shintaro Sato, Martina Vincoli, and Mansu titled Mind the Trust Gap Identifying Misalignments in Teacher Student Views toward Control and agency in K12 classroom AI published in the Proceedings of the ACM on Human Computer Interaction. This team representing Saarland University, the University of St. Gallen and other institutions found that while students and teachers share some common ground, there are critical areas where their perspectives on AI powered tools just don't line up. In fact, their detailed speed date and study with 16 school students and 15 teachers in Germany revealed distinct differences, especially regarding how much they trust AI and the social and emotional aspects of learning with AI. That's a huge finding, isn't it? The core claim is that understanding these different teacher student AI views is absolutely critical for designing AI that actually works for everyone in a K12 classroom. The researchers used a clever method, a speed date and study with storyboards. They presented realistic scenarios of AI use in the classroom. Things like an intelligent tutoring system, assigning tasks or grouping students for a project. Now for those unfamiliar, Intelligent tutoring systems or its are AI powered tools designed to provide adaptive scaffolding, personalized feedback and content assignments to individual students responding in real time to their learning needs and even their emotional states. Think of it as a super smart digital tutor. The storyboards show different levels of control. Sometimes the student made the decision the sometimes the AI did and then explored both positive and negative consequences. This wasn't about what AI could do, it was about what teachers and students preferred and why. And what they found points directly to the complexities of stakeholder interaction with AI in classroom settings. Let's dive into some of the key findings. First, the trust gap. This is probably the most striking misalignment. The study showed that students across all scenarios had mixed feelings about trust in AI's decisions. On one hand, they acknowledged that AI could be helpful, perhaps by suggesting topics they hadn't considered, but on the other, they seriously doubted AI's ability to truly understand their needs. One 14 year old student talking about AI making content decisions, worried about trivial mistakes skewing the AI's assessment leading to it choosing the next task automatically after like 10 seconds, even if the mistake didn't reflect their actual knowledge. Crucially, students emphasize that AI can't see my emotions or understand social dynamics like friendships. They stressed their trust in human teachers, believing teachers have a better connection and understand students as whole individuals. Here's where it gets really interesting. Teachers, while also not fully trusting AI's assessments, they worried about AI assigning inappropriate tasks and stress their need to intervene and override the system, had a completely different concern about student trust. Many teachers actually feared that students would trust AI more than them. One teacher with a year and a half of experience observed, I think that there is even more acceptance among students when the computer says here, work with this because it can help you than when the teacher says something that the teacher has consciously decided. They worried that students might perceive AI as fairer or less biased in its decisions than a human teacher. This is a profound disconnect, isn't it? Students say we trust you teacher over the AI. Teachers worry students will trust the AI more than Mises. This isn't just a miscommunication. It has deep implications for how we implement AI in classroom settings. If teachers believe students will automatically prefer AI guidance, they might inadvertently step back, creating a vacuum that students don't want. My core philosophy, Human in the Loop, is about teachers remaining the decision makers with AI as the tool. But for that to work, teachers need to understand their unique irreplaceable value, especially in areas like relationship and care, which AI simply cannot do. This trust gap directly impacts the capacity for human connection, which is paramount in K12 education. The second thing teachers should know is about social dynamics and emotional aspects of student learning with AI. Both students and teachers agreed on this one. AI falls short when it comes to understanding social dynamics and personal feelings. Students were really clear, especially in scenarios involving AI generated group work that AI can't grasp. Things like how well you get along with these people, they said. This personal connection that I think is important is just lacking for students. Human to human relationships trump performance based grouping. Teachers echoed this worrying that AI might create dysfunctional groups or not account for a student's mood. But teachers went a step further. They expressed a deep concern that an over reliance on AI or digital tools in general could reduce social interactions in the classroom. One teacher knowed that virtual hand raises, while efficient, feel artificial and a bit of normal interpersonal communication would be good. Another, a teacher with 15 years of experience mused, do I now work with a robot that I can somehow optimize the emotion is an important thing that needs to be there for the learning success. This emphasizes that for many educators Learning is deeply interwoven with human connection and empathy. It's a powerful reminder that machines can compute, they cannot wonder, they cannot care. When we design learning, we need to ensure it's design and learning that cannot be faked because it demands depth, care and imagination. AI in Education Germany, just like everywhere else, needs to grapple with this Third let's talk about AI monitoring and the feeling of judgment Students consistently raised concerns about the pressure and discomfort caused by AI monitoring. They worried about their data being shared with people beyond their teachers, parents, peers, even the wider school community. One 14 year old expressed a feeling of existing pressure at school. You're told you're going to school under pressure and the only thing you're told by the teacher is that you can't do this and that well and that you should do this and that better. Adding constant AI surveillance to that they felt would be too much. Teachers shared these concerns, acknowledging the potential negative impact on student feelings, but they also highlighted how AI monitoring might change to their own behavior. One teacher with four years of experience worried that if AI alerted them to a struggling student, it would look as if you are coming to the student because you are now placing them under general suspicion. This demonstrates a deep concern for the teacher student relationship and a desire to create a safe learning environment where mistakes aren't cause for embarrassment. In fact, some teachers with more experience, like one with 28 years under their belt, suggested that strong trust between students and teachers could actually mitigate the fear of AI monitoring. This is where ethical non negotiables like data privacy and transparency become really important. It's not just about what the AI can do, but how it feels to the people using it. The fourth major area of misalignment revolves around data sharing and its pedagogical benefits. Students, while prioritizing privacy and wanting control over who accesses their data, showed a willingness to share information specifically with their teachers, but only if they understood the purpose and benefits. A 13 year old student put it simply, if you leave everything to yourself, then no one can help you. But if you just share the biggest part with the teacher, then the teacher can also help you. The key here was transparency, knowing what data was shared Shushu who she saw it and how it would be used. Teachers, on the other hand, had a strong desire for full access to student data, seeing it as essential for effective instruction. They were generally reluctant to give students full control over data sharing, fearing it would limit the system's usefulness, especially for students who most needed help, but might be hesitant to share one teacher with 15 years of experience said, I strongly assume that the low performing students in particular don't want to share their data and that is not an overview for me. If only the good ones share it or those who want to share it. Interestingly, some teachers even considered not being fully transparent about monitoring, believing students might prefer not to know. This is a classic tension point for AI in classroom implementation. We have to address these concerns directly. Ensuring transparency about AI use is paramount. It's about building know like trust. First awareness, then credibility, then confidence. Fifth, let's look at students autonomous decision making and help seeking Students generally desired a degree of control over their learning choices. They wanted the freedom to deviate from AI recommendations, perhaps choosing easier tasks on a tough day, as one 16 year old student put it. However, they also showed a surprising metacognitive awareness, admitting that they might not always make the best choices, recognizing their own tendency to pick the easier way just because they want more free time, they look to human teachers for guidance in these situations. Teachers had a similar perspective. While they believed giving students choices could be motivating, they also acknowledged that some students would indeed avoid challenging work simply practicing what they already knew. This, as one teacher with two years of experience pointed out, would mean no learning gain. They recognized that AI could enhance the efficiency of providing help, especially for shy students who might be uncomfortable asking questions aloud. But teachers firmly believed their human intervention and support were irreplaceable. One student speaking about teachers over AI said, I think talking with a human can help you better because it's more of a dialogue and they can get if you're understanding. This aligns perfectly with the idea of outsource the doing, not the thinking. AI can handle the repetitive tasks of delivering hints, but the nuanced empathetic dialogue and the judgment of a human teacher are what drive real learning and productive struggle. This is key to developing AI literacy, which isn't about memorizing tool features but about collaborative reasoning ability thinking with AI, not just using it. The findings here resonate with challenges observed in AI education Germany, where PISA data has shown varying levels of support provided in classrooms compared to other OECD countries. So where do these misalignments come from? The researchers suggest two main factors. First, bere developmental and metacognitive differences. K12 students are still developing their perspective, tatum skills and self regulated learning strategies. They might prioritize comfort and ease in the moment, whereas teachers are looking at the long term learning goals and the broader classroom environment. Secondly, there are asymmetric roles and power dynamics. Teachers have the responsibility for student success and curriculum alignment, often needing data to monitor progress. Students are the subjects of this monitoring, with understandable fears about surveillance and judgment. To address these gaps, the study offers compelling design implications. We need to scaffold students metacognition within AI based systems. Instead of just giving an AI recommendation, the system could ask do you agree with this recommendation? Why or why not? Inviting students to reflect Furthermore, it's about fostering transparent communication and trust building, perhaps through perspective taking exercises where students are prompted to think, if I were a teacher, how would I like to use this tool? This isn't about skipping stages in the know like trust progression, it's about actively building them. Giving students more agency, even if just by allowing them to share preferences on how AI parameters are adjusted can make a huge difference. It's an evolution, not revolution approach to integrating AI. The real value of AI is not in what the machine produces, but in how the student responds and how the teacher leverages it to deepen learning and enhance human connection. We need to remember that AI is helping us hold the complexity so we have capacity for creativity. We're not teaching students to outsmart machines, but to outthink them. If you're finding these insights helpful and want more discussions on the practical applications of AI in education, please make sure you're following and subscribe to the podcast. Ultimately, this study reminds us that when we talk about AI in schools, we're not just talking about technology. We're talking about people, students, teachers, and the intricate human relationships that form the very foundation of learning. Mind the trust gap, acknowledge the social and emotional elements and build solutions that respect the intelligence and unique perspectives of everyone in the classroom. That's all for today. Thanks for listening.