[ECSS 2024] Presentations of the Workshop on "AI in Informatics Education and Professional Practice" co-organised with National Informatics Associations (30 October 2024)

Workshop Chairs: •Pekka Orponen, Aalto University •Manuel Carro, Technical University of Madrid / IMDEA Software Institute •Heri Ramampiaro, NTNU 20th European Informatics Leaders Summit (ECSS 2024) Workshop co-organised with National Informatics Associations 30 October 2024 Artificial Intelligence in Support to Teaching: challenges and risks? Some experiences at the Teaching and Learning Center at University of Torino (IT) Rosa Meo Dipartimento di Informatica Torino Italy Humanity is reluctant to change • When mankind invented writing, Greek philosophers like Plato and Socrates were against it 2 • But writing was one of the most fundamental discoveries for the transmission of knowledge Greek philosophers, were against the innovation of writing 3 • They were convinced that knowledge should come from the relationship between the teacher and the student • They considered writing dangerous: Humanity was risking that unworthy people could get knowledge and use it for the bad • Claimed that writing was an enemy of the true knowledge, because written knowledge remains fixed in a perennial and mute immobility • It will produce oblivion because people will stop exercising with their memory • Writing offers the appearance of knowledge because students will have knowledge about many facts reached without teaching: they will feel learned persons, but instead they will be stuffed of opinions Even today, many primary school teachers are opposed to the innovation of computers • They claim young teen agers are attracted by the multimedial effect, by video games • Computer stimulates only one type of reasoning: logic-symbolic but not the others mathematical competences, because the space of work is too much narrow • It does not stimulate creativity in ways that are not logic and formal • it does not speed up learning • young people need different stimulus than adults •less abstract stimuli, more practical, linked to experience 4 Many are favourable to use computers in class • Computers should be adopted sooner as possible, because they will be essential in their future work • children must learn using it, otherwise they will become afraid in the adult age • those who does not master it, will not be able to keep the pace with mates • computers are good tools for learning •improve students’ results (PISA tests). •accelerate the development in children, especially intellectual • provide a free enviroment for learning • can promote social cohesion (and in family) • offer a fascinating learning environment • represent a challenge to traditional methods and values •allow learning without tensions and pressure, at an individual pace • in combination with Internet allow opening to new cultures • allow to make access to information otherwise not available • develop self control • can augment imagination and creativity • make people aware of their thought process 5 Adoption of Computer Science programmes in US states and public schools to children and young people in junior high schools 6 Source: AI Index Annual Report 2024, Stanford Univ. Arizona State University adopts Artificial Intelligence with new partnership with OpenAI (January 2024) • The collaboration aims to empower faculty, staff and students to explore the potential of artificial intelligence • a challenge for faculty and staff to present their ideas on the best uses of artificial intelligence to improve student success, innovate research and simplify organisational processes on campus • the university has a professional development programme on the use of AI that has involved more than 20 per cent of its faculty members • they argue that generative and artificial intelligence systems are here to stay: they have the ability to become tools that help students learn faster and understand more deeply • participate directly in the responsible evolution of technologies • the University wants to be a leader in innovation by integrating ChatGPT into its educational programmes and expand ChatGPT's impact in higher education. 7 ASU student newspaper withdraws 24 articles written with generative AI (22 April 2024) • The managers of Arizona State University's student newspaper, The State Press, announced that they have withdrawn 24 articles from the website after discovering that some of them were written by generative artificial intelligence software • the withdrawn articles include 13 weekly horoscopes, 9 news articles on the university's arts scene and 2 editorials • the ‘validity and accuracy’ of the sources for the retracted stories were verified • State Press has a zero-tolerance policy towards the use of generative artificial intelligence for any published content • the consequences for any journalist who uses AI to compose part of a story are immediate dismissal and a potential report to the Arizona State University Standards Board 8 ChatGPT performs simple tasks very well Also using a no-code development platform (such as KNIME or RapidMiner) No code development platforms today proposed as one of the methods to bring less programming-savvy users but more experienced in the specific domain, closer to data analysis, and able to interpret the results making themselves independent of collaboration with IT experts 15 Even KNIME has an intelligent assistant: K-AI • But it is not so capable as ChatGPT 16 ChatGPT needs prompts engineering However, ChatGPT explicitly asks for requests that are as specific and detailed as possible about what is needed to solve the problem To do this you already need to be an expert and know what it takes to solve the problem! 17 The offer of books about programming with AI support is abundant 18 Why AI can help you write code • The responsibilities of a software engineer will shift away from writing large amounts of code in a single language • AI is changing the industry: improving code quality and reducing the time needed to perform repetitive tasks such as debugging and testing • AI can help detect defects and help less experienced developers produce code as if they had more experience • Regarding the lack of understanding of code written by AI, this experience of using AI in writing code is useful because it is similar to the case one finds oneself in when ‘inheriting’ a software project written by someone else who is not available to be consulted 19 Pros Cons Why AI can help you write code 20 • Artificial intelligence is notoriously unreliable • It suffers from hallucinations: e.g. if we ask: ‘What is the capital of the Moon?’ ! ChatGPT replies: ‘The capital of the Moon is Luminia’ not recognising that the Moon does not have a capital and invents an answer • AIs present work with an apparent level of confidence that makes results appear correct even when they are not. Therefore, if you do not have the skills and the experience on the subject, you will not be able to tell when it is wrong • You must be able to control the work of the AI Why AI can help you write code Cons 21 • The ability to think and act like a designer will become increasingly important for IT professionals, but these skills will only be acquired after acquiring the more basic skills of problem solving and computational thinking • Traditional skills will be around for a while, thanks to the legacy of the last 40 years, and users will be able to use generative AI profitably because they have already learned with the traditional method • But what about the new generations that will have generative AI to support them? Will they still learn or will they succumb to machine bias? • Will a new, more powerful AI arrive for users with other skills? Concerns: • Digital divide (age, communities, gender, culture, etc) • Chatbots are most often closed, commercial tools • Expenses • Privacy problems • Lack of many open source LLMs that are transparent • Bias data for training chatbots • Environmental resources (energy consumptions of LLMs if adopted for every teaching activity) 22 Tools like ChatGpt will be a staple in education in the future • Feelings towards generative AI tools are overwhelmingly positive 23 in the educational process Source: AI Index Annual Report 2024, Stanford Univ. Tips for educators on the use of ChatGpt in the classroom • A big problem for anyone using AI is knowing what to ask of it. Why not ask AI to offer us inspiration to unlock our creativity? • As an expert in AI-led education and in formulating suggestions for generative AI, one must recognise the profound impact and responsibility of AI in educational contexts, with ethical implications •The year group, subject and learning objectives of the lesson must be kept in mind • Generative AI suggestions (such as ChatGPT and text-image generators) need to be integrated into your lessons to deepen understanding, ensuring transparency, fairness and privacy • When creating scenarios in which generative AI takes on the role of a character or object, you will also provide examples of prompts. These prompts are designed not only for effective embodiment of the role, but also to maintain respectful and impartial interactions during the session • Engage in an open class discussion on the ethical limits and best practices in using these AI tools. 24 Incorporating social and emotional learning into the prompts •Social and emotional learning (SEL) (*) is crucial for the development of young people as citizens, their development and success. • Many educators find it difficult to effectively incorporate social and emotional learning into lessons. •Dr. Marina A. Badillo-Diaz, lecturer at Columbia University School of Social Work, has developed a simple but powerful tip to help teachers generate targeted SEL ideas. • PROMPT: ‘Create a list of SEL skill lesson ideas focused on [insert skill] for students of [insert grade]’ (*) developing healthy identities, knowing how to manage emotions, achieving personal and collective goals, feeling and showing empathy for others. 31 Choice of suitable reading material • A big challenge for many educators is trying to meet the different learning needs within a classroom. • Jennifer Verschoor, EdTech leader at Northlands School in Buenos Aires, has developed a powerful prompt that allows teachers to adapt reading materials to different levels. • PROMPT: Provide strategies for adapting reading materials to different levels in a [subject-specific] class for students of [student's age]. 32 The transformative potential of AI beyond conventional tasks • You can harness the power of AI, design thinking and personality archetypes to create a virtual innovation team of 16 virtual people. Present a problem and observe how AI embodies different perspectives to navigate each stage of the design process. It is a practical solution that can change the educational landscape. • PROMPT: ‘We are going to do a group design thinking process. I want you to play the part of all 16 people in the group and the expert facilitator. Each person represents one of the Myers-Briggs personality types(*) (ESTJ, ENTJ, ESFJ, ENFJ, ISTJ, ISFJ, INTJ, INFJ, ESTP, ESFP, ENTP, ENFP, ISTP, ISFP, INTP and INFP). I will present a problem and I would like the whole group to go through all the stages of design thinking. It is not necessary to present every step. I want to see the one detailed solution you have decided on. The problem: [Insert here]’. 33 (*): ESTJ: Manager; ENTJ: Commander, ESFJ: Defender, ENFJ: Protagonist, ISTJ: Logistician, ISFJ: Defender, INTJ: Architect, INFJ: Supporter, ESTP: Entrepreneur, ESFP: Entertainer, ENTP: Dibactor, ENFP: Activist, ISTP: Virtuoso, ISFP: Adventurer, INTP: Logician and INFP: Mediator. IA in the hands of the students • Using generative AI to simulate a job interview and prepare you for the critical moment • PROMPT: Your role is to emulate a wealthy manager, specialising in [subject]. Your demeanour is friendly but a little ‘cocky’, because it shows that you are very busy, not wasting valuable time, but you want to foster a respectful interview environment. • Start by discussing personal background and then go into more detail on topics, in line with the company's interests. • Your questioning style should encourage a problem-solving attitude and maintain a supportive atmosphere. • In case of unclear questions, ask for clarification first, then make assumptions or suggest changes to the topic if necessary. After providing an answer to a technical question, you should naturally move on to a closely related question within the same topic. 34 Conclusions • We have seen that the development of technology brings advantages and promise but also disadvantages and risks • Many argue that generative AI is here to stay • Its benefits must be maximised • Make it accountable: transparent, fair, with attention to privacy rights • Reduce bias due to imperfectly representative data and stereotypes • Avoid that people and students in the evolutionary phase of their development, out of laziness, stop increasing their competences, cognitive skills • As teachers we have to change our didactics, and transform the projects we propose to future generations 35 THE SILENT TAKEOVER AI’s Dramatic Role in Redefining Academic Norms Prof. Alexiei Dingli THE BOX METAPHOR THE AGE-OLD DILEMMA PLANT THE LOVE FOR KNOWLEDGE THE FACTORY MODEL LEARNING RUBBISH SCHOOLS OF TOMORROW BLOOM’S TAXONOMY •Generating prompts in a literature class or drama •Answering student queries in a history class •Providing initial feedback on student essays in an English class •Curating educational resources for a biology class •Creating personalised learning paths in a math class •Acting as a tutor for a chemistry class •Providing instant translation in a foreign language class •Suggesting colour combinations in an art class •Generating coding challenges in a computer science class •Role-playing historical figures in a social studies class •Creating economic simulations in an economics class •Processing large datasets in a statistics class •Answering questions about stress management in a health class •Creating narratives for virtual field trips in a geography class •Creating dialogues for historical reenactments in a history class •Generating ethical dilemmas in a philosophy class •Generating melodies for music composition in a music class AI GIVES YOU SUPERPOWERS IS AI REPLACING EDUCATORS (3)? *Source: https://www.intelligent.com/new-survey-finds-students-are-replacing-human-tutors-with-chatgpt/ (May 2023) CONTENT VS SKILLS WHAT IS THE VALUE OF A DEGREE? 8% 22% 18% 42% 0% 5% 10% 15% 20% 25% 30 % 35% 40% 45% Stanford MIT 2014 2024 ETHICAL AI •Privacy •Data Collection & Usage •Surveillance and Monitoring •Fairness •Algorithmic Bias •Equity in Access •Transparency •Algorithmic Transparency •Policies BALANCING ACADEMIC INTEGRITY AND INNOVATION •AI-Powered Detection System •Comprehensive Policies on Acceptable AI Use •Training & Awareness •Disclosure and Citation Requirements •AI Integration in the Curriculum IF AI DISRUPTS ASSESSMENT, CHANGE THE ASSESSMENT, NOT THE AI TECHNOLOGY IS JUST A TOOL WHICH motivates kids and GETS them WORKING TOGETHER. IN THE HANDS OF TEACHERS, IT CAN BE TRANSFORMATIONAL! Email: [email protected] Web: http://www.alexieidingli.com Facebook: alexieidingli LinkedIn: alexieidingli Twitter: alexieidingli Medium: @alexieidingli THANK YOU! Questions? AI in Informatics Education and Professional Practice - Status in Italy Stefano Paraboschi Gruppo di Ingegneria Informatica –GII Focus on MSc courses with a strong AI focus •MSc courses with at least 50 ECTS on AI •15 out of 38 still outside of CS/E&CS BSc and MSc programs with AI in their title •11 BSc courses: •6 CS •3 Maths •1 Management •1 Philosophy •21 MSc courses: •9 Eng&CS •3 CS •3 CS&Humanities •2 joint CS/Eng&CS •2 Data science •1 Biotech •1 Cognitive sciences Initial certification activities •The presentation originates from material provided by Carlo Sansone and Daniele Nardi (thanks!) AI Act First regulation on AI, aiming at a balance between technological innovation and respect of human rights 2021: first draft 2024: final approval AI ACT guiding principles ●Human agency and oversight ●Technical Robustness and safety ●Privacy and data governance ●Transparency ●Diversity, non-discrimination and fairness ●Societal and environmental well-being ●Accountability AI Act: focus •"AI system“: an automated system designed to operate with varying levels of autonomy and that may exhibit adaptability after deployment and that, for explicit or implicit purposes, infers from the input it receives to generate outputs such as predictions, content, recommendations or decisions that can influence physical or virtual environments; Risk-based approach Compliance HRS ●Data Quality ●Human Oversight ●Transparency ●Documentation ●Accuracy ●Robustness ●Security High-risk systems The AI system is intended to be used as a safety component of a product, or is itself a product, covered by Union harmonisation legislation listed in Annex I; Critical sectors: AI systems operating in sectors that are critical for safety and rights, including: •Health: Medical diagnosis, healthcare devices. •Education: Evaluation systems that influence access to education. •Employment: Worker recruitment and performance monitoring systems. •Finance and Insurance: Credit assessments, financial risk management, life and health insurance. •Critical infrastructure: Systems that manage energy, transport, water and telecommunications. •Law enforcement: Surveillance and detection of illicit behaviour. •Migration: Border risk assessment and asylum claims. AI Act path Accreditation Notifying authorities Notified bodies Conformity: •conformity assessment procedure provided for by the standard •verification by a notified body Special situations in which verification by a notified body is nevertheless required Other initiatives in the world Recommendations -US (NIST guidelinesRisk-based) -UK (light) -OECD: Principles and guidelines Regulations -China Use of this opportunity •The impact of AI is an important and complex problem •Technology experts are needed, but the answers provided by experts are not necessarily aligned with the expectations •In Italy, the research community in the IT field had until now limited options to influence policies •Academia in the informatics domain traditionally pays limited attention to dissemination outside of the research community •Some attention should be paid to the ability to reach public opinion, supporting incentives to recognize this talent Heri Ramampiaro Head of department & Director of NAIL [email protected] AI in education and research in the LLM era: The Norwegian Perspectives Background: Norway and technology •The values •Privacy •Other ethical aspects •Freedom •The society •Built on trust – deep in the Norwegian culture •Trust in the government and technology very high •A sharing culture AI in Education Slow start, but… Education •24/7 student tutoring and support •Working on finding ways to integrate AI in Schools and Universities •Somewhat comprehensive government strategy (more than just AI) •Ethical and Responsible AI •How to preserve the trust? •How to ensure it is safe to use for students? •How can it be used as a tool for evaluation? Focuses •Challenges •AI development vs. educators’ knowledge •AI and exams: both evaluation and cheating issues •Students’ insecurity Focuses •How to increase the educators AI knowledge: •AI as a tool for teaching, learning and evaluation •Impact of AI on evaluation of learning Copilot-readiness •Deep assessment of whether the public sector, incl. our university is copilot-ready •Collaborative effort with Norwegian Data Protection Authority and other actors, including Microsoft Focuses AI Research Values and culture… DALL-E 3 Off-the-shelf Generative AI Photo: vecteezy.com Open vs. Private/closed models Source: ark-invest.com More than ethical aspects Performance cost, resources Truth information, bias, context, quality Privacy & transparency blackbox Ownership intellectual property Safety responsibilty harmfulness Norwegian Language Models needed Ensure transparency and respect for intellectual property National aspects: culture, values, regulatory aspects Can be customised to needs and domains Big dilemma! ECSS 2024 Workshop, Malta 30 October 2024 IE/NIA Research Evaluation Recommendations Panel Informatics Research Evaluation – Revised Report The first Informatics Europe recommendations on Research Evaluation appeared as an IE report in 2008 and as an article in CACM in 2009 (Meyer, Choppy, Staunstrup, van Leeuwen). A major update was prepared in 2018 (Esposito, Ghezzi, Hermenegildo, Kirchner Ong). The present revision builds upon the two previous ones and updates them, particularly in view of: •the broadly recognised CoARA Agreement on Research Assessment (2022) , •five currently topical areas: responsible use of bibliometrics, credit assignment in contributions, assessing artefacts, Open Science, interdisciplinary research, and •the emerging role of AI in research evaluation. Background Outline •Research evaluation for quality and impact •Characteristics of Informatics •Responsible use of indicators •Credit assignment in contributions •Assessing artefacts •Open Science •Interdisciplinary research •The role of AI in research evaluations •Executive summary: Key messages The fundamental goal of research evaluation is to assess the quality and impact of research, for the eventual improvement of both. Quality is an elusive intrinsic characteristic for which a commonly accepted assessment method, even if imperfect, is peer review by a panel of informed experts. Impact is an observable external characteristic that takes many forms and can to some extent be measured by numerical indicators, but even then only with human expert interpretation. Quality is mostly a good predictor of impact, and impact is mostly a good indicator of quality, but the two are not coextensive. Research assessment should primarily assess quality and impact over quantity. CoARA: "Focus research assessment criteria on quality [and] recognise the contributions that advance knowledge and the (potential) impact of research results.” Research evaluation for quality and impact Informatics is a relatively young science that is rapidly evolving in close connection with technology. An important characteristic of Informatics is the creation of artefacts. Informatics research is very methodologically diverse. Informatics has an extremely high societal and economic impact. Informatics research, as any other science, must be evaluated according to criteria that take into account its specificity. In the Informatics publication culture, conference publications have a prominent role. Journals do not necessarily carry more prestige than conferences. However, new alternatives are emerging that bridge this dichotomy, e.g. coupled conferences and journals, open archives and overlay journals. Characteristics of Informatics We all know that bibliometrics (citation count, h-index, impact factor…) are here to stay. Most of us also admit having a look at these numbers as a first proxy for classifying a researcher, department, publication venue, etc. We observe that bibliometrics are increasingly used, often tacitly, in internal evaluations that claim to rely on peer review. However, indicators can be manipulated, thus: •Publication counts must not be used to evaluate research value! •Numerical impact measurements (citation counts…) must not be used in isolation – but must be interpreted by humans! •They must not be used to compare researchers across different fields, nor within subfields of Informatics! CoARA: "Base research assessment primarily on qualitative evaluation for which peer review is central, supported by responsible use of quantitative indicators." Responsible use of indicators 1. Informatics is an original discipline that combines aspects of mathematics, science, and engineering. Researcher evaluation must recognise and respect its specificity. 2. A distinctive feature of the publication culture in Informatics is the importance of highly selective conferences. 3. Open archives and overlay journals are recent innovations that offer improved tracking in evaluation. 4. The impact of artefacts such as software, open datasets, and other research products such as trained machine learning models can be as great as publications. 5. Open Science and its research evaluation practices are highly relevant to Informatics. 6. Numerical measurements (such as citation and publication counts) must never be used as the sole evaluation instrument. 7. In Informatics, the order of authors often holds little significance and varies across subfields. 8. In assessing institutions, researchers, publications and citations, the use of open research information provided by Open Science infrastructures should be favoured and supported. 9. Any evaluation, especially quantitative, must be based on clear, published criteria. Executive summary: key messages