This paper illustrates the activities of the projects SMAILE and AILEAP, which are devoted to foster the growth of awareness and readyness to learn artificial intelligence in the general population. The first project was mainly oriented to children and young adults, while the second is more oriented to the personalization of the learning experience also in professionals.
tences play a key role in ensuring the EU’s success in global competition with respect to digitalization and AI.
The new digital society, in which artificial intelligence There is a need for society as a whole to be equipped with (AI) and autonomous systems play an increasingly im- the necessary understanding, knowledge, and skills for portant role, will require all citizens to develop a solid the ‘AI era’, so as to make full use of the overall potential foundation of digital skills. A good level of digital aware- and to keep everyone on board.” Additionally, the recomness and skills will enable citizens to actively participate mendation of the Commissioner for Human Rights on in the new digital world, access public services, find a Artificial Intelligence “Unboxing artificial intelligence: 10 job, and avoid the digital divide, which will be one of the steps to protect human rights”, underlines the importance main reasons for poverty in the future. of investing in the literacy on AI of the general public
In “Digitalization, Al and Equity – How to strengthen through robust awareness raising, training, and educathe EU in the global race of future skills and education, tion eforts. Moreover, AI is already under examination while ensuring social inclusion”, the European Economic as part of the ongoing update to DigComp: the Euroand Social Committee stresses that “skills and compe- pean Digital Competence Framework. Looking at the Digital Economy and Society Index (DESI, https://digitalIntiazle-dIAby20C2I3N:I3,rMdaNyat2i9o–n3a1l,C2o0n2f3e,rPeinscae, IotnalAyrtificial Intelligence, orga- strategy.ec.europa.eu/en/policies/desi), the Italian initial * Corresponding author. situation is quite critical. We urgently need to develop " matteo.baldoni@unito.it (M. Baldoni); cristina.baroglio@unito.it innovative solutions to address the lack of digital skills. (C. Baroglio); alessio.bottrighi@uniupo.it (A. Bottrighi); In order to make more efective the teaching of the (mS.onCiacpae.bcucchcii)a;reellelni@a.ugnanitdoo.iltfi@(Mun.iBtou.citci(aEr.elGlia);nsdaorlafi).;capecchi@unito.it fundamental concepts of AI to the broad public, it is usecristina.gena@unito.it (C. Gena); francesco.iani@unito.it (F. Ianì); ful to adapt the contents, the learning environment, and elisa.marengo@unito.it (E. Marengo); davide.marocco@unina.it the approach to the learners. This process requires iden(D. Marocco); roberto.micalizio@unito.it (R. Micalizio); tifying as accurately as possible how learners difer from luca.piovesan@uniupo.it (L. Piovesan); one another in abilities involved in the specific learning (mAi.chRealpap.p);oanrtoicvoertvao@@ouspneindaa.liet.a(Ml.i.t P(Aon.tRicoovrevtoa));; amon.rapp@unito.it domain and, then, making decisions about how to depaolo.terenziani@uniupo.it (P. Terenziani); fugo@ospedale.al.it sign personalized learning experiences. Specifically, the (F. Ugo) projects SMaILE (http://www.smaile.it/) and AI-LEAP © 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License (http://www.ai-leap.it/) aim at fostering the personalizaCPWrEooUrckReshdoinpgs IhStpN:/c1e6u1r3-w-0s.o7r3g ACttEribUutRion W4.0oInrtekrnsahtioonpal (PCCroBYce4.0e).dings (CEUR-WS.org) tion of the learning experience along three diferent lines. activity was followed by a reflection phase in which
Learning of AI : we claim that the understanding of AI children were asked to summarize the activity they peris improved by the development of basic natural abili- formed, reflect on it in order to understand the principles ties that individuals master to varying degrees. Early of the task they were asked to perform, and, finally, to recognition of each child’s diferent mastery of abilities generalize the principles by finding similar everyday life enables a personalized training aimed at strengthening situations. For instance, in order to strengthen the ability the identified weaknesses. to diferentiate Between Syntax and Semantics, we devel
Embodied AI : AI is often applied to or associated with oped the “Egyptian Room". Inspired by Searle’s Chinese physical devices, like robots. The production of specific room, this activity consists of six rounds. At each round, learning materials that can be flexibly be composed in a one child is selected to sit at the knowledge desk: a desk personalized learning experience, to be tailored to each which will give him or her the power to quickly associate individual’s mastery of specific abilities, is a still open hieroglyphic answers (output) to hieroglyphic questions. ifeld. This is made possible by providing the child at the desk
AI as a tool to personalize learning: AI representations with the right correspondences. The child has to just pick and reasoning allow the construction of powerful tools to the paper string associated with the one amounting to the personalize teaching non-IT professionals. In particular, question and give it to the expert. Afterwards, children in we will study the case of physicians. the class are asked to answer the same hieroglyphic question. To do so, they will have, first, to translate symbols in letters, then letter strings into Italian words. Finally, 2. SMaILE they will select a possible answer from a s et of possible answers. More than one answer is possible, so each of them is collected by the expert at the blackboard.
We involved a group of schools, that are located in the city of Torino, in the experiments. Six classes belong to the fifth grade while six classes belong to the sixth grade.
Half of them (three fifth grade and three sixth grade classes) the training and half were part of the control group. The control group carried out activities that were similar to the training ones in terms of used material and topic, but that do not train the abilities we are interested in. To assess the success of our training, we will perform a test before the training/control session, and the same test was performed at the end of the AI coding lessons (see Figure 2).
Quercetti, in collaboration with the University of Torino, designed some unplugged activities to introduce the AI concepts (symbolic manipulation, non symbolic manipulation and machine learning, and planning)
in December 2023). The authors are involved in the subproject EmpAI (empowering artificial intelligence). We briefly describe the activities and results achieved so far, see [1, 2] for more details.
First, we developed a set of unplugged activities for training basic abilities that learners already have and that we devise as at the basis of acquiring mastery of symbolic AI. We also investigated, in collaboration with Quercetti (http://www.quercetti.it/), the use of educational toys for acquiring skills in AI.
Concerning the first task, the identified abilities that we wish to strengthen are: (i) ability to diferentiate between syntax and semantics; (ii) ability to classify data; (iii) ability to behave based no test-operate-test-exit unit; (iv) ability to plan.
For each such ability, we designed training activities as playful practice tasks for children (see Figure 1); each information in a diferent format (which could be more suitable to bring in class when actually performing the activity) and, on the other hand, it provides a higher level of detail; c) the slides of the lesson to be presented to learners and used as a guide. We decided to supply the pptx format so that lessons could be changed by teachers to tailor them to their needs.
The unplugged activities are grouped in two main categories concerning Artificial Intelligence and Computer Science concepts. The characteristic of this type of activities is that they do not require the use of technological devices but they require the preparation of paper-based material. Accordingly, we provide a teacher guide for Figure 3: Enigma by Quercetti. each activity describing, in the first part, the aim of the activity, the concepts that are addressed, the list of the material needed and the competences and skills that the to children by means of games produced by Quercetti. learners will acquire in performing the activity. After The activities proposed to primary schools tackle topics this introductory part, the activity is described in detail among which are the following: in a way that it can be replicated. Accordingly, both the • trpherpeoebttrlieatmniossndi/unicttetoiroapntrioopgnrorcabomlensmc.e,Tpthhtsea.taicstitvoittyrainnstlraotdeurceeasl TicnuhssWetsrimeuocsnatetitteohurniapastlaotthnsopehpatocreeiwantcththoiaestrppirssrehompovaeuiradleneddttuhtaonesdsmauedrpatdtpaeioktrriieotanlaseraavelnefildrereastp.lhoteyrtpdeiedss-. • problem solving. This activity introduces stu- of interaction between the research group and teachers dents to the notion of “state of afairs” as a set or other interested users accessing the website. This area of properties that hold in a certain moment, the is managed as a forum where few threads (on coding “state of afairs” as a goal to achieve and the no- and on the unplugged activities) are already available tion of solution as a set of steps, organized into an and others can be added. The idea is to leverage on the execution path, that allows to reach the desired interdisciplinary nature of the research team, to support state of afairs in the best way possible. teachers in their activities in class. In this way, both • cryptography, and the concepts of encoding and questions, curiosities and new ideas can be discussed both decoding information. The aim is to underline from the computer science and the educational points of the importance of having strong encryption that view. Moreover, in this area we aim at collecting feedback can withstand a brute force attack. of any kind on the material that we developed. This will Many Quercetti games were used, among which “Pallino be a precious source of ideas and suggestions to improve Coding” and “Tag the Picture". One, Enigma (see Figure 3), the course. Interestingly, forums are visible to every was partially designed and developed from scratch within user registered to the website so that interactions can the project. occur even among users who can exchange ideas and
All the teaching materials produced along the project suggestions on how to perform the course. have been organized in a web site (https://empai.di.unito. Training for teachers was developed it/), that includes a number of online courses in order in collaboration with the crowdfunded to allow the interested teachers to download and use project“Impa.IA.mo l’Intelligenza Artificiale the material in a profitable way. The site is structured Giocando" (https://www.ideaginger.it/progetti/ intto the following parts: Coding, Unplugged Activities, impar-ia-mo-l-intelligenza-artificiale-giocando.html). Discussion and Q&A space, Training for teachers. The We believe that to fully benefit from the course we Coding section collects eight lessons that we developed are proposing, it is important to understand the main to explain the basic concepts of programming (in the first concepts underlying it. To this aim we conceived a four lessons) and of Codey Rocky programming (in the course for teachers consisting of three lessons. In the remaining four). For each lesson one can find a) a video ifrst one we explain the main concepts of Artificial of duration up to 10 minutes meant to serve as a teacher Intelligence (AI), what AI is and what is not and what guide. Here, one of the instructors who performed the is the diference between strong and weak AI. In the activity in class explains the objective of the lesson, the second lesson we introduce and explain the four basic main parts in which it is structured and the main concepts cognitive abilities that are the target of the proposed which are explained; b) a pdf file meant to serve as a unplugged activities and that are involved in the AI teacher guide as well. One the one hand it presents the concepts that we consider in this project. Finally, the third lesson explains the main programming concepts that we address in our coding course and how they are addressed and explained in the course. For each of these three parts we provide a video and a set of slides.
Additionally, we designed a self-evaluation quiz which is meant to be a tool that a user can use to identify the main take-aways of the corresponding lesson and whether she/he grasped the important parts of it.
This project continues the experience of SMaILE but it is more strongly focused on personalization. Personalization is critical to developing methods and techniques that democratize design, innovation, and knowledge creation, and make citizens aware and active members of society. Personalization targets, in the case of AI-LEAP, the learning of AI. This choice is motivated by the desire of supporting the complex process of construction/interpretation of the AI-permeated world, which requires the construction of the right conceptual tools in the people and in the society: “AI is a question of culture” [3] but culture is more easily spread when it is tailored to the individual. Moreover, AI-LEAP investigates also AI as a tool for personalizing the learning experience concerning topics that ar far from AI.
It is organized in three sub-projects, namely: T3-AI (Personalizing Test to Tailor Training of AI), under the responsibility of the principal investigator (UNITO); Teach E-AI 2C (Teaching Embodied Artificial Intelligence to Children), under the responsibility of Progetto Partner Ricerca e Sviluppo 1 (NAC-UNINA, with Treccani Futura and Città della Scienza as "partner territoriali”); PTPCAI (Personalized Training of Professional Competencies with AI), Progetto Partner Ricerca e Sviluppo 2 (AI@UPO, with DAIRI and POP-AI as “partner territoriali”).
T3-AI will investigate how early identification of each child’s diferential mastery of foundational cognitive capabilities involved in machine learning, probabilistic AI, and symbolic AI allows for a personalized training that strengthens the individual weaknesses. For what concerns symbolic AI the project will exploit as a base the results from the project EmpAI@SMAILE. Teach E-AI 2Caims at creating specific learning materials that can be assembled into a personalized learning experience tailored to the individual’s mastery of specific skills. The sub-project is based on instructional design principles and the 4C/ ID model. Adopting an Embodied AI approach leads to: 1) overcoming a strictly algorithm-driven approach, 2) understanding how biological systems work to replicate them in artificial systems, 3) developing principles for intelligent behavior and 4) applying these principles to artificial systems that interact with the real physical world: robots. PTPC-AI is based on the assumption that the personalization that can be achieved through the use of AI representations and reasoning is key to efective and tailored training for professionals who are not IT, such as physicians. 3.1. T3-AI
ties for machine learning and probabilistic AI approaches.
Early identification of each child’s diferential mastery of abilities allows for personalized training aimed at strengthening identified weaknesses. In recent years, we have explored the idea that it is possible to identify “basic abilities” that are already present in humans and that, when trained, favor learning of symbolic AI approaches [1, 2]. In this context we identify the following main goals: Given the importance of dataset construction to machine learning and probabilistic AI approaches, the project will focus on the basic abilities that underlie it.
The child-as-data-analyst metaphor captures the finding that even infants are remarkably sensitive to statistical regularities in the environment and use these regularities in forming categories [4]); the child-as theorist metaphor captures the observation that children’s concepts go beyond obvious or perceptible features to include theoryrich entities such as causes, functions, teleology, and essences in their concepts [5]. The ability to think about the possible/impossible instances of a given category is an example of a basic ability children need in constructing data sets. We will engage primary and secondary school teachers in educational events aimed at increasing the culture of AI and explaining the fundamentals for training the basic abilities of their students.
• Goal 1: Facilitate learning of machine learning and probabilistic AI approaches by training basic abilities that are innate to children. The ability to think about the possible/impossible instances of a given category is an example of a basic ability children need in constructing data sets. • Goal 2: To develop a tool to assess each child’s developmental level of the basic abilities required to learn AI. Such an assessment would allow personalized training sessions to be developed to improve AI learning. • Goal 3: With the support of Treccani Futura and of POP-AI, dissemination of the results of the research, the practices and tools in the school world. 3.2. Teach E-AI 2C Embodied AI aims at creating specific learning materials that can be assembled into a personalized learning experience tailored to an individual’s mastery of specific skills. Specifically, the goal is to understand how biological systems work in order to replicate them in artificial systems, develop principles for intelligent behavior, and apply these principles to artificial systems that interact with the real physical world: robots. In this context we identify the following goals: Developing an integrated platform aimed at children and adolescents, by implementing a robotic farm, an integrated hardware/software system for evolutionary and interactive robotics. This platform will be implemented in such a way that it can be easily used by kids so that they can see E-AI basics and concepts in action and practice them hands-on. In this challenge, implementing a solid and efective system is essential to ensure usability. Engaging a broad audience and allowing kids to use the learning units and perform tasks in a simulated and physical environment to practice with Embodied AI. Learners will have the opportunity to learn about E-AI through both learning materials and hands-on experiences with robotics and evolutionary robotics, working on digital and tangible objects. one learns how to act only by practicing on real patients.
We want to show that such a scenario can be drastically improved by using AI. We will develop advanced AI frameworks (simulation a nd verification) to train physicians to act on virtual patients, based on the best medical practices described in clinical guidelines. Such frameworks will support the verification and personalization of training processes and will also be available to physicians for their autonomous study and self-evaluation. We will select and acquire guidelines for a specific disease as a case study. However, the educational and AI methodologies developed as part of the project are independent of the specific domain and guidelines.
Our challenge is to lay the foundation for an “AI culture” in healthcare. As a pilot project, it can provide physicians with concrete evidence of the benetfis of AI methods. Through a series of lessons, presentations and events, we will promote a culture in which physicians are aware of, trust, and use AI frameworks appropriately.
We are members of the “Laboratorio integrato di intelligenza artificiale e informatica in medicina” in which also the Dipartimento di Attività Integrate di Ricerca e Inno• Goal 1: Introduce E-AI to children and early ado- vazione (DAIRI) of the Azienda Ospedaliera of Alessanlescents. Based on the profiles of the young learn- dria participates. Thanks to this cooperation, we will ers, activities will be tailored to each of them. involve physicians of Alessandria territory. In addition • Goal 2: Create an integrated platform to com- to disseminating key project findings, we will organize plement the Teach E-AI 2C learning units. As presentations and events to show citizens concrete exfar as we know, there are no other tools for E- amples of successful applications of AI in healthcare and AI that specifically target younger learners and train a new generation of citizens who are aware of the are intuitive and easy-to-use in school context. benefits of AI and want to use them.
The project will fill this gap by implementing the Teach E-AI 2C robotic farm, an integrated hardware/software system for evolutionary and interactive robotics. This platform will be implemented in such a way that it can be easily used by children and kids, allowing them to see the fundamentals and concepts of E-AI basics in action and practice them hands-on. In this challenge, implementing a solid and efective system is critical to ensure usability. • Goal 3: dissemination of educational integrated learning path on E-AI, personalized/customized to diferent educational needs to a broad audience. • Goal 1: To apply AI methodologies in order to personalize the education of professionals, considering physicians as a concrete example. Specifically, we will address the context of continuous medical education, and we will focus on advanced AI Clinical Decision Support techniques based on computer-interpretable clinical guidelines. • Goal 2: Design of an innovative AI-based verification framework for evaluation and self-evaluation for the support of personalized training experiences. • Goal 3. Dissemination is a key challenge in order to lay the foundation for an “AI culture” for professionals, in particular, for what concerns “decision support”. 3.3. PTPC-AI
ucation through computer-interpretable clinical guide- 4. Conclusions lines, based on the assumption that the personalization that can be achieved through the use of AI representa- The awareness and competence developed through the tions and reasoning is key to efective and tailored train- activities of the three projects aim not only to educate fuing for professionals who are not IT, such as physicians. ture AI/ICT professionals (e.g., designers, programmers), We address the context of continuous medical education, but also to generate the kind of insight that translates tackling the following goals. Traditional medical educa- into the mindful and ethical use of tools that are built by tion covers a wide range of knowledge. However usually exploiting AI techniques.