AID-E Magazine

AID-E BACHELOR’S PROGRAMMES MAGAZINE ACADEMIC YEAR 2024 A MAGAZINE FROM THE- 2025 ARTIFICIAL INTELLIGENCE AND DATA IN EDUCATION NETWORK

IMAGE GENERATED BY ADOBE FIREFLY

FOREWORD Data and AI in education have been receiving a lot of attention

With the AID-E network we aim to cross the boundaries that

for several years now, (even before the launch of ChatGPT). In

exist between different stakeholders and between different

the Netherlands, several different programs and organizations

disciplines. We aim to connect different types of knowledge

focus on the use of data and AI education (e.g., NPuls, NOLAI,

and expertise with regard to data (science) and human

NLAIC). At the University of Twente, data and AI in education

decision-making, and to inspire, innovate and accelerate the

are also an important focus, taking into account various different

use of data and AI for learning and development. We also

perspectives: from a research perspective (e.g., assessing its

aim to disseminate the knowledge we gained here at this

effects on teachers and students), from a teaching perspective

university, to have an impact on our society. We want to

(e.g., using AI and data in courses), from a professional

support people in leveraging the potential data and AI have

development perspective (e.g., supporting lecturers in the use

to offer, whereas at the same time be cautious about the

of AI and data), from an ICT perspective (e.g., managing the

possible risks involved.

use of data and AI within our infrastructure), and from a policy perspective (e.g., understanding policies that need to be in

In this online magazine, you will find stories about the use

place for data and AI in education).

of data and AI in education from our network members.

We hope to add new stories every month. For each of the

In this context, the AID-E (AI and Data in Education) network

stories, you will also find the contact information of the

was launched over a year ago to learn from and integrate

authors and we encourage you to contact them if you want

these different perspectives. A central focus of the network

to know more. We hope you will find as much inspiration in

is how to use data and artificial intelligence (AI) to improve

reading these stories as we got from collecting them.

education decision-making. Lecturers, for example, might use data to boost individual and group learning performance;

Kim Schildkamp,

policy advisors and managers may use data to make proactive

Bernard Veldkamp,

decisions based on forecasting and analysis; professional

Maurice van Keulen,

development providers may support staff by using data,

Adelson Dias De Araujo

learning analytics, and artificial intelligence; researchers may develop educational tools using artificial intelligence or study how to enhance their data literacy and artificial intelligence literacy.

Coordinators of the AID-E network

DATA TEAMS AS MYTH BUSTERS Educational organisations are confronted with a lot of challenges: How can we improve the reading and math skills of our students? How can we improve our final examination results physics? How can we reduce grade repetition? How can we improve the wellbeing of our students?

FOR QUESTIONS PLEASE GET IN TOUCH!

WHAT IS IT ABOUT?

Cindy Poortman c.l.poortman@utwente.nl

Each day, educators in primary, secondary and higher education are confronted with numerous decisions they have to take to confront these challenges and improve the quality of education.

Often these decisions are based on the basis of anecdotal information, assumptions and intuition. However, taking these decisions based on data can improve the quality of these decisions, and ultimately the quality of education.

Kim Schildkamp k.schildkamp@utwente.nl

WHY IS IT IMPORTANT? To support educators in the use of data to improve education the data team intervention has been developed. The data team intervention focusses on the professional development of educators (e.g., teachers, school leaders, managers) in the use of data for school improvement.

Examples of data educators can use include assessment results, lesson observations, surveys, information on background of students, and interviews.

A data team consisting of six to eight educators are coached in doing research and solving concrete problems with the help of data. Using an eight-step plan, they define problems, develop hypotheses, gather and assess data, analyze it, draw conclusions, take data-informed actions, and evaluate outcomes. A coach and resources support this process, helping them debunk misconceptions, identify causes, and enhance education quality through evidenceinformed actions.

KIM SCHILDKAMP

The data team is supported by a coach, a practical manual and a data analysis course. Based on data, myths in the educational organization about the causes of a problem are debunked and the real causes of the problem being investigated are discovered. A team can then implement action based on data to solve their problem and improve the quality of education.

WHAT CAN WE LEARN FROM IT? The data team intervention has been designed at the University of Twente and has been studied extensively in primary, secondary and higher education over more than a decade. Research results show that working with data teams can lead to an increase in data literacy, data use, and achievement, and that the work is also sustainable in the schools

WANT TO KNOW MORE? On the data team website an overview of available publications and resources can be found: https://www.utwente.nl/nl/bms/elan/datateams/

CINDY POORTMAN

CONVERSATIONAL AGENT FOR COLLABORATIVE LEARNING The Collaborative Learning Agent for Interactive Reasoning (‘Clair’) is a conversational agent that aims to support students in small group discussions in science learning to be more productive.

WHAT IS IT ABOUT? Clair was developed at the University of Twente to address some limitations that similar technologies have. Our first prototype has been tested in secondary classrooms in Brazil and the Netherlands. Among the main challenges for designing collaborative conversational agents are “what to say?” and “when to say it?”. To address what, Clair has a set of “talk moves” or questions to ask to students. These talk moves are derived from a classroom talk framework called Academically Productive Talk (APT). To decide when, we designed fuzzy logic-based triggers for each of these talk moves. These triggers are designed to be interpreted and configured by humans while still relying on patterns from data.

FOR QUESTIONS PLEASE GET IN TOUCH! Adelson de Araujo a.dearaujo@utwente.nl Pantelis Papadopoulos p.m.papadopoulos@utwente.nl

WHY IS IT IMPORTANT? As students are working together online in small groups and using chat discussions to understand these concepts, having a productive discussion is critical. Teachers can assist students by asking questions to make them deepen their reasoning or clarify the ideas at hand.

This is usually very time consuming for teachers, and in some cases, they do not guide student chats at all. In addition, the quality of discussions can greatly vary, from engaging and productive to distracting and unproductive. For example, a more verbal student will contribute more while the partner just agrees, or another student may barely build on the partner’s contributions. Conversational agents for collaborative learning are seen as a promising approach to scale productive talk support among student groups, particularly in cases where a teacher cannot be present.

ADELSON DE ARAUJO

WHAT CAN WE LEARN FROM IT? From this project, we have learned that recent advancements in machine learning can support the design of conversational agents that can work across various science topics and languages while also providing flexible triggers for a range of talk moves. By having a conversational agent that is reliable across multiple contexts, educational researchers can further study how to dynamically provide productive talk supports to student groups, teachers can have more flexibility with online classroom orchestration, and students can reflect on their discussion and internalize practices of productive talk into their lives.

WANT TO KNOW MORE? de Araujo, A., Papadopoulos, P., McKenney, S., & de Jong, T. (2023). Supporting Collaborative Online Science Education with a Transferable and Configurable Conversational Agent. CSCL 2023 Conference Proceedings.

PANTELIS PAPADOPOULOS

BOOSTING PRE-U EDUCATION WITH LEARNING ANALYTICS The University of Twente’s Pre-U programme offers high school students a range of activities, including scientific education, UT study program exploration, and skill development (e.g. teamwork, critical and creative thinking). Activities vary from one-day workshops to multi-day deep dives into subjects.

WHAT IS IT ABOUT?

FOR QUESTIONS PLEASE GET IN TOUCH! Maschja Baas m.i.a.baas@utwente.nl Sander Wenderich s.wenderich@utwente.nl

Currently, we are running pilot projects to assess the impact of learning analytics on educational outcomes, focusing on data from e-learning modules. One specific pilot seeks to understand student behaviour and enhance education quality. Furthremore, we are testing a blended flipped classroom approach to boost student engagement and higher-order thinking skills. Using McKenney & Reeves’ (2019) educational design research method, we have created a blended learning masterclass with four in-person sessions and four e-learning modules. We gather data on student behaviour in e-learnings and evaluate learning outcomes in in-person sessions, combining online and face-to-face data.

WHY IS THIS TOPIC IMPORTANT? Globally, we are evolving into a digital society, and online learning is becoming integral to our future, whether combined with in-person education or not. In the education sector, rapid technological advancements allow us to enhance educational quality through data analysis, potentially with the help of machine learning. Learning analytics offer insights into learning processes and behaviors, helping Pre-U improve online programs and focus in-person time on higher-order thinking skills.

WHAT CAN WE LEARN FROM IT?

MASCHJA BAAS

With the pilot project we hope to gain insight in: • What kind of patterns we can find in the learning process as well as the learning behavior of students in our online programs; • How to use these insights regarding learning analytics to increase the quality of our education; • How blended learning can enrich Pre-U’s educational programs and the engagement and learning outcomes of students within these programs.

WANT TO KNOW MORE? We hope to share the results of our first pilot, mainly executed by EST-graduate student Eliza Vermare, in autumn 2023. Simultaneously, we will initiate additional pilot projects, including online skills education, in collaboration with local high schools in the Twente region. All projects are executed in co-creation with Peter Groothengel (TELT department) to test new e-learning software and evaluate the use of learning analytics. For more information and questions please get in touch with Maschja Baas (online learning specialist at Pre-U).

SANDER WENDERICH

VOICE VOICE, a Virtual voice-based artificial intelligence conversational agent, addresses the vital need for students to enhance their verbal communication skills. This innovative online system provides a platform for individualised practise in oral examinations, offering flexibility and convenience

FOR QUESTIONS PLEASE GET IN TOUCH!

WHAT IS IT ABOUT?

Simone Borsci s.borsci@utwente.nl

VOICE serves as a digital space where students can independently practise oral exams, tailored to their schedules and preferences. It alleviates the burden on educators by automating the process of formative assessment.

Johannes Steinrücke j.steinrucke@utwente.nl

WHY IS THIS TOPIC IMPORTANT? Oral examinations are essential for evaluating knowledge and communication skills, but teachers often struggle to provide sufficient practise and personalised feedback. This limitation can lead students to undervalue verbal presentation skills. VOICE aims to bridge this gap by allowing students to practise without fear of judgment, fostering inclusivity. Additionally, VOICE aligns with the Dutch knowledge agenda, promoting technology in education and enhancing organisational approaches. It ensures equal opportunities for students to engage in additional practise and feedback, potentially gamifying the learning experience for improved inclusivity. In essence, VOICE provides non-judgmental, individualised formative assessment, enabling students to assess and enhance their knowledge at their own pace, using their preferred devices.

WHAT CAN WE LEARN FROM IT? VOICE explores the advantages of voice-based AI in enhancing students’ learning performance and engagement. Additionally, it provides valuable insights for educators on optimising educational organisation, formative assessment, and system practicality, with a focus on reducing workload and increasing flexibility. The research identifies crucial components for the system’s effectiveness, assesses user-centric design, and emphasises the importance of inclusivity. Furthermore, it delves into the accountability of AI in educational assessments, examining the roles of teachers and developers, and considers the potential of voice-based AI for summative assessments, providing a comprehensive evaluation of its applications.

JOHANNES STEINRÜCKE

WANT TO KNOW MORE? Currently, we are exploring potential grant opportunities. For more information please contact the in VOICE involved AID-E members:

• Johannes Steinrücke, j.steinrucke@utwente.nl • Simone Borsci, s.borsci@utwente.nl • Maryam Amir-Haeri, m.amirhaeri@utwente.nl • Maurice van Keulen, m.vankeulen@utwente.nl • Pantelis Papadopoulos, p.papadopoulos@utwente.nl

SIMONE BORSCI

THE BLESSINGS OF AI FOR EDUCATIONAL ASSESSMENT Educational measurement is a critical field of study within the Cognition, Data, and Education (CODE) group at BMS. This group has evolved over time, leveraging AI to enhance the assessment process, but also encountering new challenges in terms of validity, ethics, and authenticity.

FOR QUESTIONS PLEASE GET IN TOUCH! Bernard Veldkamp b.p.veldkamp@utwente.nl

WHAT IS IT ABOUT? Educational measurement is a key area of interest for the CODE group at BMS. Initially, our focus revolved around psychometric models for assessing latent abilities in students through exams and surveys. We delved into topics like computerised adaptive testing, multi-level modeling, and detecting unusual response patterns. However, the landscape of data evolved significantly, shifting from simple binary matrices to include verbal responses, response times, process data, and sensor data. To adapt to this changing data landscape, we turned to AI for assistance.

WHY IS IT IMPORTANT? Within the CODE group, we applied AI in various ways, such as stealth assessment, essay scoring, fraud detection, and analyzing response behavior. This transformation enhanced the quality of assessments substantially. Nevertheless, the use of AI in assessment also presents challenges. Questions arise about the validity of AI-based scoring, particularly when using complex algorithms that operate as black boxes. Ethical concerns emerge regarding the outsourcing of critical teaching functions like assessment to algorithms. Additionally, applications like ChatGPT can generate content that closely resembles human work, raising questions about authenticity and trust. Within CODE, we like to focus on these challenges.

Topics like explainability of AI and trust are high on our agenda.

WHAT CAN WE LEARN FROM IT? When it comes to assessment, AI has enriched the teacher’s toolkit tremendously. But in order to benefit, increase of AI literacy is really needed. Besides, instead of focusing on the end products, i.e. checking the correctness of the response, focus has to shift to the process of coming to the product, which can reveal so much more about the true mastery level of the students. As AI researchers, we can provide the tools. Now the teaching community has to embrace them. Only together, education can truly benefit.

WANT TO KNOW MORE? For further questions or discussions, please reach out to Bernard Veldkamp

BERNARD VELDKAMP

AI AS THE TEACHER’S ASSISTANT: ADAPTIVE MATH APPLICATION This project explores AI’s role in enhancing education. It focuses on the development of an AI assistant designed for primary school math, aiming to optimise student learning.

WHAT IS IT ABOUT?

FOR QUESTIONS PLEASE GET IN TOUCH! Maurice van Keulen m.vankeulen@utwente.nl

Our project focuses on integrating AI to support teachers in the classroom, with a specific AI assistant tailored for primary school math education. It’s main goal is to enhance learning by delivering math problems that are neither too easy nor too difficult, adapting to each student’s evolving needs. While current practice categorises students into 3 to 5 levels, AI promises close monitoring and real-time adjustments, accounting for diverse student preferences and sudden leaps in understanding.

The AI serves as an assistant, operating under the teacher’s guidance to address individual student needs. Teachers have access to a monitoring dashboard to track student progress and intervene when necessary. Through our research, we identified the challenge of meeting the diverse needs of students. For instance, highly gifted students may struggle with simple exercises while excelling in more challenging ones. The AI’s ability to differentiate between insufficient understanding and frustration, and adjust difficulty accordingly, relies on various data inputs, including sensors to detect signs of boredom, frustration, distraction, overwhelm, and fatigue.

WHY IS THIS TOPIC IMPORTANT? Optimal learning necessitates a blend of confidence-building through simpler problems and deeper learning through more challenging ones. The conventional 3-to-5-level approach falls short, especially for students at the extremes of the learning spectrum or those with unique learning needs. Moreover, it places the burden of close monitoring and level adjustments on teachers, whose dedication varies. AI can individualise this process, adapting to different problem types in real time. On the teacher’s side, adopting this technology relieves them of administrative and monitoring tasks, allowing for more meaningful interactions with students.

This, in turn, enhances both the student learning experience and teacher job satisfaction.

LOTTE VAN DIJK

WHAT CAN WE LEARN FROM IT? This project highlights the potential of AI in education and the benefits of collaborative AIhuman systems in various domains. It also showcases how teachers can maintain control, reduce administrative burdens, and enhance the quality of their service by working in synergy with AI.

WANT TO KNOW MORE? Please contact Maurice van Keulen or have a look at Lotte van Dijk’s Masters thesis via https://essay.utwente.nl/89015/

MAURICE VAN KEULEN

AI FOR STANDARDISING TEST SCORES This project seeks to ensure fair grading across different exam versions using machine learning, enhancing educational measurement and data harmonisation.

WHAT IS IT ABOUT? To prevent fraud, exams often have different versions, containing different sets of questions.

However, we want to have the same type of assessment. For a student and their exam results, it should not matter what test version was used. The question here is thus how to make grades comparable across test versions? Somehow, we need to translate scores from one test version to another test version. We propose to use machine learning algorithms to do just that.

Currently we are exploring methods like K-nearest neighbours, ordinal regression, and support vector machines, adapting them to the specific case of exams. We then compare them to traditional methods from educational measurement.

FOR QUESTIONS PLEASE GET IN TOUCH! Maryam amir Haeri m.amirhaeri@utwente.nl Stéphanie van den Berg stephanie.vandenberg@utwente.nl

WHY IS IT IMPORTANT? Exam grades should be the same, regardless of what specific test version was used.

Students get points for the items on a test, and these total scores have to be translated into grades. If one test version is slightly easier than another version, the translation of a total score to a grade should be different for the two test versions. We need a method to do this fairly.

WHAT CAN WE LEARN FROM IT? The current methods use item-response theory models and other methods. In practice however, these models do not always show a very good fit to the actual test data. From this project we can learn what AI can do in educational measurement. We can also use the same knowledge to harmonise data from health and behaviour questionnaires.

MARYAM AMIR HAERI

WANT TO KNOW MORE? For questions please get in touch via Maryam Amir Haeri or Stéphanie van den Berg

STÉPHANIE VAN DEN BERG

INTERACTIVE INCLUSIVE HERITAGE The VU-UT Smart Societies impact program explores the integration of Virtual Reality (VR), Conversational AI, and Knowledge Graph technologies in Dutch museums and cultural heritage institutions.

WHAT IS IT ABOUT? This project aims to enhance visitor experiences, improve accessibility, and promote diversity and inclusion. Collaborating with museums and artists, the research prioritises representing diverse perspectives, innovative interaction designs, and technology to engage audiences, especially adolescents.

FOR QUESTIONS PLEASE GET IN TOUCH! Shenghui Wang shenghui.wang@utwente.nl Carolien Rieffe c.j.rieffe@utwente.nl

WHY IS THIS TOPIC IMPORTANT? This research addresses the need for diverse representation and inclusion in cultural heritage.

By using technology, it breaks down traditional museum limitations and fosters equity, diversity, and inclusion beyond physical boundaries. It also provides interactive experiences that attract a broader audience, including tech-savvy adolescents.

WHAT CAN WE LEARN FROM IT? 1.

2.

3.

4.

5.

Importance of Representation: This research highlights the significance of diverse representation to challenge biases, promote inclusivity, and move beyond a single dominant narrative.

Technological Advancements: The integration of VR, Conversational AI, and Knowledge Graph technologies enhances the museum experience, providing immersive, interactive, and personalised exploration.

Accessibility and Engagement: It emphasises the importance of making cultural heritage accessible through technology, engaging younger generations and a diverse audience.

Equity, Diversity, and Inclusion: Fostering equity and inclusion within cultural heritage is crucial, challenging stereotypes and creating inclusive spaces.

Enriched Understanding: Incorporating Knowledge Graph technologies deepens our understanding of cultural heritage by contextualising artifacts and narratives within a broader historical and social framework.

SHENGHUI WANG

WANT TO KNOW MORE? This project line within the VU-UT Smart Societies impact program is formed by experts in the fields of Psychology, Human-Media Interaction, Game Design, Communication Science, and Sociology. Collaborating partners are Nationaal Archief, Museum het Rembrandthuis, and visual artist Remy Jungerman.

For more information please contact Shenghui Wang

CAROLIEN RIEFFE

ONLINE CHATTING AND 21ST CENTURY READING TRENDS A thorough analysis of PISA data spanning from 2000 to 2018 reveals a noteworthy correlation between the rise of online chatting and changes in reading literacy, offering valuable insights into the evolving reading habits and educational landscape.

FOR QUESTIONS PLEASE GET IN TOUCH! Hans Luyten j.w.luyten@utwente.nl

WHAT IS IT ABOUT? Secondary analysis of PISA data (2000-2018) reveals a strong correlation between the growth of online chatting and changes in reading literacy. In Europe and North America from 20002009, online chatting gained popularity, accompanied by a slight dip in reading proficiency.

From 2009-2018, online chatting continued, though at a slower pace, coinciding with improved reading skills. Conversely, in other countries, moderate online chatting in 2009 was linked to improved reading, but post-2009, the surge in online chatting corresponded with declines in reading literacy.

Between 2000 and 2018, 39 countries participated in PISA surveys, evaluating cognitive skills in reading, mathematics, and science among 15-year-olds, alongside collecting data on family backgrounds, attitudes, and extracurricular activities.

Both during 2000-2009 and 2009-2018, there was a clear correlation between increased online chatting and decreased reading literacy. This connection was mediated by changes in reading fiction and awareness of reading strategies.

WHY IS IT IMPORTANT? While controlled experiments show screen reading’s adverse effects on comprehension, large-scale surveys contradict this. Intense ICT users are not necessarily poor readers.

Examining per-country trends shows the real-life relevance of screen reading’s impact. At present, traditional reading skills require more effort since reading outside school has diminished. Educators are concerned about ChatGPT’s authenticity in written assignments.

The research on the link between online chatting and reading skills suggests that even seemingly trivial ICT tools like Whatsapp can significantly affect reading skills.

WHAT CAN WE LEARN FROM IT? Declining student achievement does not necessarily reflect deteriorating education quality.

Instead, this study highlights the connection between the rise in online chatting and the decline in reading literacy.

WANT TO KNOW MORE? Contact Hans Luyten or read his paper: • Luyten, H. (2022). The global rise of online chatting and its adverse effect on reading literacy.

Studies in Educational Evaluation, 72.

HANS LUYTEN

USING AI-GENERATED CASES TO ENGAGE STUDENTS MORE What can a teacher do to make the curriculum more appealing? And how do you ensure that students can apply the knowledge they have acquired in practice? These are questions that many university teachers grapple with. The introduction of AI chatbots offers new possibilities for educators. This article elaborates on one example of how chatbots can be used to make lectures more engaging.

The introduction of AI chatbots has caused a stir worldwide. This is certainly true for universities. Since then, lecturers, students, and policymakers have been reflecting on the implications of AI for education and the work of academics1. And it doesn’t stop at reflection, as many are already using chatbots like ChatGPT2. As a result, lecturers need to adapt their teaching methods. Some universities and schools have chosen to ban the use of chatbots3, while others see potential benefits4.

I align myself with the latter group. Chatbots are not just a risk; they also offer an opportunity. Lecturers can choose from a multitude of electronic tools, such as interactive quizzes and videos. Chatbots add a valuable tool to this arsenal.

AI-GENERATED TEACHING CASE I teach at the University of Twente within the Health Sciences programme. One of the courses I teach is a first-year course called “Change Processes and Management of Innovation in Healthcare”. What could be better than having students apply change models and innovation processes in practice? As a student, I experienced the most enjoyable and educational lectures when we had to analyse cases using theory. More importantly, this aligns with the learning objective of applying insights about change strategies and approaches to describe innovation outcomes in healthcare organizations. However, developing a good case study is time-consuming, especially the writing process.

Enter ChatGPT. The tool had been available to the general public for just a few weeks, and I had long been toying with the idea of developing a case study for my students. The case would focus on a significant organizational change in a hospital. I

experimented with ChatGPT to write the text and used DALL-E to create accompanying images to make the case appealing.

THE “ALEXANDER HOSPITAL” Meet the “Alexander Hospital”, a fictional hospital facing several significant challenges. First, I asked ChatGPT for suggestions on what the case should revolve around. It gave me ten suggestions, from implementing a new electronic patient record to using a surgical robot. Pretty good and relevant examples! I then used several “prompts” to further develop the case, such as developing a case about option 1 (implementation of EPR) and adding a persona for the hospital manager responsible for this change. Students read the case from the perspective of Karen, who, according to ChatGPT, has 20 years of experience in the hospital world and a reputation for effectively implementing change! I then had ChatGPT elaborate on several other ideas: background information about the hospital, the hospital’s financial data, and a dilemma.

SOURCES 1

Renkema, M., & Tursunbayeva, A. (2023). Demystifying the future of knowledge workers: implications of Artificial Intelligence in academia.

In European Academy of Management 2023 Conference proceedings.

2

rtlnieuws.nl/tech/artikel/5378644/chatgpt-app-nederland-chatbot

3

ny.chalkbeat.org/2023/1/3/23537987/ nyc-schools-ban-chatgpt-writing-

4

utwente.nl/nieuws/2023/7/1052023/universiteit-twentebenut-kansen-ai-en-is-

artificial-intelligence zich-bewust-van-risicos#ai-vaardigheden

Colleagues from the University of Twente have developed several teaching cases for bachelor students over the past few years. I studied these – along with the cases I still had from my studies and some real examples from hospitals – to understand what makes a good case.

There need to be several personas playing the main role, a choice or dilemma to be made, some events, and a lot of background information. Dialogues were missing in my case, so I instructed ChatGPT to present Karen with a dilemma and generate a dialogue. It emerged that Karen is dealing with a financially troubled hospital – and she has been tasked with solving these problems while the hospital needs to become more innovative.

With all this input, I got to work. I used Karen’s case as an introduction and supplemented it with additional information. I had to adjust and check the organization description somewhat. The financial results turned out to be unusable after checking the annual accounts of several local hospitals. In other words, the writing process was a cocreation between ChatGPT and myself, where ChatGPT was asked to come up with ideas, generate texts, and provide options – while I checked the texts for usability. This ultimately resulted in a 16-page teaching case about the Alexander Hospital.

the case, and although it was a bit daunting to perform a role play, they said they learned a lot from it. They also recognized certain aspects of the Alexander case in guest lectures by innovation managers from healthcare institutions, which were also organized in this course. In the course evaluation, students appreciated the chosen approach: “During the tutorials, often the same list of stakeholders, etc., so you became well acquainted with the case, and this was good preparation for the exam. Tutorials very interactive, and you benefited a lot from them.”

The positive side of AI chatbots AI chatbots pose real challenges for teachers. However, they can also be used to improve education and facilitate work. If I had had to develop the teaching case myself, it would have taken me a lot of time, and I probably wouldn’t have done it. Nevertheless, the chatbot is not a panacea, and you need to check everything carefully.

However, chatbots provide quick suggestions and a starting point for further elaboration of (new) educational ideas.

APPLICATION OF THE CASE I used the case in the three tutorials of the course – where students had to formulate answers to questions such as which change management approach did the Alexander Hospital choose? Which aspects of this process are good, and which could have been better?

Some groups were asked to present, and then the answers and theory were discussed in class. We also did a role play several times, where students had to make suggestions from different roles in the hospital to solve the organizational problem. This enabled students to practice applying theory in practice and to practice dilemmas where difficult choices have to be made. They found it educational and fun to work on

FOR QUESTIONS PLEASE GET IN TOUCH! Maarten Renkema m.renkema@utwente.nl

MAARTEN RENKEMA

This image was generated by midjourney where the original photo from dall-e was used as source material and the following prompt was added: “imagine this image but more realistic with the power of midjourney.”

LEARNING ANALYTICS WITHIN THE UT Learning Analytics, at UT, involves continuously gathering data throughout a student’s academic journey to create a comprehensive picture of their performance, strengths, and weaknesses. It goes beyond merely assigning a final grade, offering students valuable insights into their progress and opportunities for improvement.

In 2020, the UT and TELT (Technology Enhanced Teaching and Learning) initiated a Learning Analytics Proof of Concept, in collaboration with the UTLC, to explore the potential of Learning Analytics in education. The initial test in 2020/2021, involving Canvas, H5P, and authoring tools, yielded promising results.

Subsequently, in Spring 2022, a successful pilot took place in the TNW faculty for Chemical education, paving the way for future developments.

PILOT CHEMICAL EDUCATION The data collection in this project relied on three content sources: • H5P interactive content creator; • Dominknow ONE, authoring module for content creation; • Canvas LMS, for hosting content and discussion platform.

These sources effectively utilised the xAPI international standard for collecting learning data, which formed the project’s essential foundation.

DATA TRANSMISSION AND TRANSLATION When students interacted with educational content from H5P, Dominknow, or Canvas, the generated data was sent in real-time to a Learning Record Store (LRS). In the case of Canvas data, a translation script was employed to bridge the gap between Canvas and the LRS. This script was necessary because Canvas could not produce xAPI data on its own. The script was implemented to convert Canvas data into xAPI data.

comparing this data with the video’s content, it became evident which segments were engaging or challenging for students. Light colors indicated highly-watched segments, while dark colors signified less-viewed segments. This information empowered teachers to finetune their lectures and better align with students’ needs.

BENEFITS OF LEARNING ANALYTICS The pilot’s results demonstrated that learning analytics enabled students to identify their strengths and weaknesses, leading to targeted self-improvement efforts such as additional practice or seeking assistance from teachers or peers. Teachers, in turn, gained deeper insights into their students’ progress and how they engaged with course materials. These insights guided instructors in focusing on specific aspects during lectures.

FUTURE PLANS For the rest of 2023/2024 we are planning new pilots with Learning Analytics in the UT’s education. TELT is also working closely together with SURF, various Dutch and international univeristies, and internal companies.

PETER GROOTHENGEL

DATA PROCESSING AND DASHBOARDS The Learning Record Store collected and processed all the data, presenting it in dashboards designed for students and teachers.

These dashboards were then integrated with Canvas, allowing students and teachers to gain valuable insights into the learning process of the educational module.

VIDEO ANALYTICS FOR ENHANCED TEACHING An example of in-depth video analytics revealed where students faced difficulties while watching instructional videos. By

FOR QUESTIONS PLEASE GET IN TOUCH! Peter Groothengel j.g.p.groothengel@utwente.nl

BLINDED BY THE LIGHT In April 2023, two University of Twente students wrote op-eds on Generative AI in education for U-Today. As an educational consultant, it reminded me of Springsteen’s ‘92 song “Human Touch,” with its line “I just want someone to talk to, and a little of that human touch.” However, in this op-ed, I’ll explain why we should be cautious not to be blinded by the light on this topic.

By all means, ChatGPT ticks the boxes of being a disruptive technology. One that shocks us and forces us to adapt in education.

But let’s not forget that this is still the tip of the iceberg in terms of this specific technology. Shortcomings and limitations of today may be gone in months and years to come. So, while the strings of words may be nonsensical at times now, we may laugh at the foolish quality of this in years to come.

Before working here, I worked for years as a teacher in special needs education. A job that, above all, requires focus on being connected with your students and colleagues. An experience that taught me that there’s not a single tool, AI model or whatever now or ever that’ll replace the connection between a teacher and their student. Any educational tool can and should only augment this connection.

There’s already many ideas and options explored in both op-eds on how you can use this type of technology. Why not go and do the exact thing Niels criticises: try and let your essay be written by an LLM. But instead of handing it in you can choose to reflect on it, and learn from the different approaches to write your own piece in a new manner.

To close off with another terminology that’s being thrown around: rewarding and recognition. Teaching is and has always been more than just transporting information on a specific field. It’s about helping others find their way far beyond this specific field. Let’s keep recognising our human touch here.

And as both gentlemen work out perfectly: this type of technology requires critical thinking. Critical thinking in the prompts you use; critical thinking in the output you get. And this is where my op-ed comes in, to quote one of the sayings around my fellow educational consultants: self-regulated learning cannot be taught self-regulated.

For many years terminologies like ‘21st century skills’ have been thrown around, but I believe that more than ever we’ll need those in our education. Critical thinking, problem solving, creativity and digital/information literacy: examples of skills that aren’t always as explicit in education as your regular learning objectives. But this is the exact type of toolbox we need to deal with new technology like this.

And how do we support others get to this toolbox? Just a little of that human touch. Like Fritdjof’s call to community, I subscribe to the idea of working and learning from others. One thing my time as a special needs educator taught me is that even your small course, of just 50 minutes per week, contributes to this toolbox of your students.

ROBIN VAN EMMERLOOT

FOR QUESTIONS PLEASE GET IN TOUCH! Robin van Emmerloot r.h.m.vanemmerloot@utwente.nl

NPULS MAGAZINE ‘SMARTER EDUCATION WITH AI’

The Smarter Education with AI magazine takes an in-depth look at AI in education and highlights UT’s innovative approach. It includes tips, inspiring examples and interviews with experts.

SCAN TO READ THE MAGAZINE!