Detailed Program of the Conference

Please select a date or location to show only sessions at that day or location. Please select a single session for a detailed view (with speakers and abstracts).

The current Conference time is: 26th May 2022, 08:57:51am CEST

Go to the LIVE PLATFORMGo to the Conference website

Overall view of the program
Parallel session - G.5 Learning AI and AI For Learning
Thursday, 03/June/2021:
2:15pm - 4:30pm

Session Chair: Rita Tegon
Session Chair: Pietro Monari
Session Chair: Annalucia Nardi
Location: Room 8
Session Panels:
G.5. Learning AI and AI for Learning

Show help for 'Increase or decrease the abstract text size'
ID: 148 / THR-PRL-E1-G.5: 1
Submit your paper
Panels: G.5. Learning AI and AI for Learning
Keywords: learning analytics, plug-in, moodle, teaching-learning


Stefano D'Ambrosio, Luca Ferrari

Università degli Studi di Bologna, Italy

Abstract: The contribution focuses on some of the main technical and educational potential of the "plug- ins" that can be installed on the Moodle LMS platform. The paper will describe some of the most frequently downloaded and used learning analytics tools on the Moodle platform. One of the plug-ins presented goes "beyond the click", it is a tool designed to offer more complex processing (going beyond descriptive analysis). Its distinctive open-source nature, combined with machine learning, makes it a particularly interesting tool, opening the way to the possibility of making predictions about a student's success, paving the way for exciting future scenarios.

ID: 350 / THR-PRL-E1-G.5: 4
Submit your abstract
Panels: G.5. Learning AI and AI for Learning
Keywords: school-dropout, performance-prediction, teaching-aid, logbooks


Giulio Angiani1,2, Alberto Ferrari1, Michele Tomaiuolo1, Monica Mordonini1

1Università di Parma, Italy; 2IIS "Blaise Pascal" - Reggio Emilia

In the last few years the amount of electronic data in high schools has grown tremendously, also as a consequence of the introduction of many digital supports as electronic logbooks, web-based structured tests, e-learning platforms. Usually, all the data of these platforms are used only for local evaluations and are rarely integrated with external systems.

Our research project, named ELDM (Electronic Logbook Data Mining),focuses exactly on this issue.

We have collected several data of electronic logbooks to extract information about the real day-by-day learning process of a number of students. These data have been retrieved from different kind of Italian secondary schools between 2015 and 2019. The dataset is made up of about 13000 instances, representing students from various Italian territories.

Our research has focused on all objectives and measurable data present in logbooks, i.e. marks, absences, final-periods-evaluation. Many features have been transformed and created in relation to the specific evaluation method of the Italian school.

To have a similar approach for different study courses, we have grouped all students' data in 6 groups to highlight the main subjects studied in high-school (i.e Italian language, Math, English, History, subjects strictly related to the student course and a last group which contains data of no-categorized subjects).

We have developed a machine-learning-based system which will be published on the web and will be freely usable by all school stakeholders.

So far, (i) the system allows every school to share easily data with the ELDM project, and (ii) it allows to check the various learning levels of the students. On the basis of data collected from involved schools, we have applied data mining techniques to analyze all the students' behaviors and results.

Our findings show that: (i) it is possible to anticipate the outcome prediction in the first school months; (ii) it is also feasible to highlight the best didactical techniques to increase performacences of the best students and to prevent school dropout and (iii) it shows that we can predict the yearly final outcome with an accuracy of about 80% already in the first three months of the courses.

only for local evaluations and are rarely integrated with external systems

ID: 697 / THR-PRL-E1-G.5: 5
Submit your abstract
Panels: G.5. Learning AI and AI for Learning
Keywords: Reinventing Education, Epistemology of Error, Tyranny of Concreteness


Piero Dominici

World Academy of Art and Science and University of Perugia, Italy

What kind of preparation or training can we propose for these children and for our students, what kind of skills or knowledge will they need,and how fast will these become as obsolete as the jobs we hold today?Young people should be encouraged to discover and follow their desires,their passions,and their imagination.Both studies in the humanities and in the sciences (kept separate by the falsest of society’s false dichotomies,1995) can tell us that we must imagine,create, inspire and be inspired; that we must observe, formulate, verify or disprove through experimentation -and/or trial and error- a series of hypotheses which will form the basis of the kind of knowledge which will always be open to modification and even reversals.Regrettably,our educational systems have internalized a pseudo-scientific,zero-sum dogma that everything we learn, do, or learn to do must be useful in some way, must be measured, evaluated and certified as something that will produce concrete results or provide economic returns,must be predictable, controllable.A dogma which I have called “the tyranny of concreteness” (2005).Instead,schools should have the function of awakening and stimulating these passions, of guiding students on pathways capable of merging reason and imagination, of linking thought, action and emotion,factors which are all but neglected in our current school and university curricula.The focus on facts and figures,on calculations and results,should be complemented by the teaching of a culture of error, an epistemology of error (1996), where the three aspects of education I mentioned earlier, error, doubt and unpredictability, should be encouraged and encompassed into the educational format, in the framework of a systemic view of ecosystems and of life itself.How can this be accomplished? To begin with, by ceasing to label,discredit and stigmatize error,from elementary school onwards. Students learn at a very early age to avoid making mistakes and to fear taking any kind of risk or venture that might result in disapproval or low grades,whereas the primary objective should be learning by error,learning through error, rather than being taught that there is only one way to solve a problem, to tell a story, to form an opinion. Questioning “why” rather than “how”. What we are instead training and teaching the new generations to become is nothing more than efficient executors.Executors of functions and rules, incapable of reflection, incapable even of contemplating the nature of these functions and rules, incapable of asking themselves “why”.Both students and teachers have been trapped by these limited perceptions, an inadequacy that leaves them incapable of seeing the connections, the links, the intersecting trajectories and loops characterizing the complexity we inhabit,whereas another primary function of education should be precisely that of teaching how to see and make connections. Ironically, considering that talk about complexity and systems thinking is all the rage today,it is amazing how much unawareness still persists about the strategic significance of thought and thought systems,how blind we are to the ubiquity of complexity in every field and praxis,how little we understand the importance of error and of the crucial opportunity in being free to make errors.

ID: 652 / THR-PRL-E1-G.5: 6
Submit your abstract
Panels: G.5. Learning AI and AI for Learning
Keywords: Artificial Intelligence, Machine Learning, Learner Profile, Customisation


Matthew Montebello

University of Malta, Malta


Matthew Montebello, University of Malta,

Keywords: Artificial Intelligence, Machine Learning, Learner Profile, Customisation

Abstract: The application of artificially intelligent techniques to assist the effectiveness of e-learning within a higher education context has been verified and accepted by both the education and computer science domains. The technologies developed through Artificial Intelligence (AI) whereby electronic systems are able to learn and generate a unique learner profile have made it possible to personalise and improve online learning environments. The digital version of a learner’s distinctive academic portfolio encapsulates specific information regarding that same learner including academic achievements, strengths and weaknesses, together with personal interests, needs and inclinations, as well as, dislikes, limitations and vulnerabilities. Such personal attributes are partially explicitly declared by an online student on enrolling to a virtual learning environment (VLE), and in part generated by the learner while interacting with the same environment. The data that the learner’s activity generates together with all the characteristic information that can contribute to the exclusive academic learner profile is collected by the VLE, processed, and re-employed to customise the next online learning experience. Learning analytics refers to the opportunistic use of such data in an effort to optimise the learning process through the elevated learning environment. A higher frequency of interactions with the smart environment the superior and accurate is the learner profile as it continuously refines itself through an expedient cycle of data collection, information processing and profile re/generation. In this work we present a number of matters that contribute to the above scenario while providing a detailed narration of how AI can contribute to the education domain on a number of counts to contribute to a smart education concept. We converge our attention towards the customisation of a VLE whereby we provide a walk-through of the smart learning environment we have created and employed with our undergraduate students. This same smart VLE has evolved over time as novel machine learning techniques have been integrated including deep learning neural networks and explainable AI. We also present numerous recommendations and best practices through the favourable results achieved and the positive feedback provided by learners and educators. All this provides enormous encouragement to delve further into this interesting and exciting area of research as we seek to continue to improve e-learning effectiveness and enhance the online learning experience. The future of smart education depends entirely on further progress within the AI domain as new and efficient intelligent techniques provide additional value and potentially characterise the future of online education and e-learning.

ID: 151 / THR-PRL-E1-G.5: 7
Submit your abstract
Panels: G.5. Learning AI and AI for Learning
Keywords: Artificial Intelligence, Algorithmic Culture, Education, Prediction, Democracy


Graeme Tiffany

Indepedent, United Kingdom

For very many children and young people, COVID19 has transformed school from a physical to a virtual space. However, in practice, little has changed; they are still recipients of ‘teaching’, albeit through the prism of a screen rather than ‘in person’. There is near unanimity this is far from ideal, hence the clamour to ‘get young people back to school’. What’s implied is the intrinsic value of presence. But when what happens is no different, whether on or off-line, is such a claim reasonable? Put it this way, if we are to value the presence of a teacher, rather than, say, a recording of someone teaching, this value is derived from the capacity and commitment to respond to learners. In turn, this ascribes value to what learners say and do. Logically then, if the latter is absent, we may as well use artificial technologies in teaching.

With this logic in mind, we might consider the value of AI in education; does it, can it, value the learner, or does it degrade relationships, push the learner further away? Arguably, AI is constituted by existing knowledge; it can only work with what data it has. It knows nothing, nor can it ever know, anything about the future, beyond, the presumption it makes that the future will look like the past. This, in essence, is what AI is, a predictive system that looks for patterns in past behaviour and draws the conclusion that those patterns will replicate themselves, thereby making the future knowable, predictable. Note also, AI draws many of these conclusions not from knowledge but proxy variables.

So yes, AI might help us with questions of ‘how?’ and ‘for whom?’, but AI cannot answer the question of what education ‘is for’; and it certainly can’t help up with that fundamental, yet most grievously ignored question: ‘what is education?’ We need philosophers for that.

Might interest in AI in education represent then a failure to conceive of anything outside a techno-economic paradigm, an ideology driven by policy-makers with a vested interest in the neoliberal state for whom AI represents a ‘solution’, to everything, from war, to the climate crisis, and seemingly intractable educational problems?

The abandonment of exams in the UK, at least for now, has put the algorithm back in its cage; and liberated historical research that shows exams (and algorithms) are far from objective, are fraught with bias, and offer not greater inclusion and equity but less.

Step forward not the philosophers but philosophy, and the very idea that education is a kind of conversation. And for this to mean anything, especially in democratic terms, the outcomes of these conversations cannot, nor should not, be predicted. AI then offers us nothing; it is the very antithesis of the creative thinking implied by ‘Reinventing’ education. AI shifts agency to external systems, it works against autonomy and choice; it offers us only the means to squeeze the last vestiges of democracy out of education. But perhaps that is the plan.

Terms and conditions · Contact Address:
Privacy Statement · Conference: Scuola Democratica 2021
Conference Software - ConfTool Pro 2.6.143
© 2001–2022 by Dr. H. Weinreich, Hamburg, Germany