2nd International Conference on Human and Artificial Rationalities

Can we expect AI to solve problems? It is a belief strongly held by Nick Bostrom that we seek to criticize. What do things require to be a problem for us? A difficulty in performing a certain function is not a problem, but a difficulty; a risk or a threat are not as such problems, but risks or threats. A problem is expressed in the form of a question, but not every question is necessarily a problem. There are many different forms of problem facing human rationality, but they all have in common the fact that AI will never have any problems, at least if it is true that a problem is not a calculation.

Can negation be depicted? It has been claimed in various areas, including philosophy, cognitive science, and AI, that depicting negation through visual expressions such images and pictures is challenging. Recent empirical findings have shown that humans can indeed understand certain images as expressing negation, whereas this ability is not exhibited by machine learning models trained on image data. To elucidate the computational ability underlying the understanding of negation in images, this study first focuses on the image captioning task, specifically the performance of models pre-trained on large linguistic and image datasets for generating text from images. Our experiment demonstrates that a state-of-the-art model achieves some success in generating consistent captions from images, particularly in photographs rather than illustrations. However, when it comes to generating captions containing negation from images, the model is not as proficient as humans. To further investigate the performance of machine learning models in a more controlled setting, we conducted an additional analysis using a Visual Question Answering (VQA) task. This task enables us to specify where in the image the model should focus its attention when answering a question. As a result of this setting, the model's performance was improved. These results will shed light on the disparities in the attentional focus between humans and machine learning models.

This paper presents an incremental consistency-based diagnosis (CBD) algorithm that studies and provides explanations for erroneous human decision-making. Our approach relies on minimal correction sets to compute belief states that are consistent with the recorded human actions and observations. We demonstrate that our incremental algorithm is correct and complete wrt classical CBD. Moreover, it is capable of distinguishing between different types of human errors that cannot be captured by classical CBD.

The mammalian brain is a complex system that still holds many secrets regarding how it functions, yet recent research highlights the importance of two types of cells in the recognition of one's location: grid cells and place cells in the enthorinal cortex. Jeff Hawkins' thousand brains theory proposes a model of cortical function which suggests that neurons analogous in function can be found in other parts of the cortex and instead of being activated for specific locations in space would be activated for specific locations in a conceptual, abstract space. In this presentation we explore how this idea can have fundamental implications regarding how we think about human cognition, how we learn concepts and how we explore potentially abstract notions. We also discuss what this means for artificial agents and how this model could be applied to improve artificial agents capabilities regarding abstraction, use of contextual information and learning.

In this research we compare the ability to take into account the intent in moral judgement with the level of Theory of Mind (ToM) in Typically Developing children (4 to 7 y.o.). In the first test we evaluated the moral algebra: combination of intent and consequence. Children had to judge social interactions between two characters in four situations: intentional harm, attempted harm, accidental harm and no harm and for two levels of aggressiveness: low and high. In the second test the ToM of the participants was measured with the scale of Wellman & Liu (2004). They had to answer 7 questions measuring their ability to take the perspective of a character. The results show an effect of ToM on the ability to take into account the intention. Children with the best level of ToM were the most able to judge social interaction according to the intent of the character. In addition, this effect increases with age. The oldest children, who usually have the best scores in ToM take more intent into account and discriminate more accurately levels of aggressiveness.

The use of algorithms in decision-making has increased in various fields, such as medicine, government, and business. Despite their proven accuracy, people often disregard algorithmic advice. When it comes to complex tasks, however, there is some evidence that people are more inclined to follow the advice of algorithms. This evidence is largely based on decision-making in rather artificial contexts, however, and studies in this field tend to rely on rather crude measures of complexity. We therefore investigate the effect of task complexity on trust in model-based advice in a realistic setting, measuring complexity in several standardized ways. We conducted an experiment with 151 participants, each assessing 20 real-life court cases of crimes (in the Dutch legal system). Participants were first asked to estimate the jailtime for each crime. They then received algorithmic advice, and were allowed to adjust their initial estimate. We measured task complexity in several ways. First, by simply counting the number of violations per case. Second, we focused on all the mitigating and aggravating circumstances associated with a specific case. Finally, we narrowed our focus to only the circumstances mentioned in the case text and concentrated on the violated sections of the law. We then used multi-level regression analysis (assessments within participants) on the target variable Weight on Advice to assess the impact of complexity on trust in algorithmic advice. Our findings indicate that participants were more inclined to trust algorithmic advice as the complexity of tasks increased, for two of the three operationalizations of task complexity.

We investigate the possibility that the recently proposed OpenAI’s ChatGPT conversational agent could be examined with classical theory-of-mind paradigms. We used an indirect speech understanding task, the hinting task, a new text version of a False Belief/False Photographs paradigm, and the Strange Stories paradigm. The hinting task is usually used to assess individuals with autism or schizophrenia by requesting them to infer hidden intentions from short conversations involving two characters. In a first experiment, ChatGPT 3.5 exhibits quite limited performances on the Hinting task when either original scoring or revised rating scales are used. We introduced slightly modified versions of the hinting task in which either cues about the presence of a communicative intention were added or a specific question about the character’s intentions were asked. Only the latter demonstrated enhanced performances. No dissociation between the conditions was found. The Strange Stories were associated with correct performances but we could not be sure that the algorithm had no prior knowledge of the test. In the second experiment, the most recent version of ChatGPT (4-0314) exhibited better performances in the Hinting task, although they did not match the average scores of healthy subjects. In addition, the model could solve first and second order False Beliefs tests but failed on items with reference to a physical property like object visibility or more complex inferences. This work offers an illustration of the possible application of psychological constructs and paradigms to a conversational agent of a radically new nature.

This article presents the initial results of the research project titled "Narrative Empowerment for Analysis: Epistemic Challenges in the Technical Construction of Strategic Narratives." To instrument the construction of strategic narratives, it is necessary to characterize their technical features. Gerard Genette's modal narratology, as developed in Figures III, allows us to describe the relationship between the pseudo-time of the narrative -- the succession of sentences and pages -- and the diegetic time -- the time of the story being told -- as well as phenomena of mode and voice. We conducted a micronarrative analysis of DGSE notes related to the Rwandan crisis in the 1990s and are now conducting a macronarrative analysis of the entire set of documents, considering their chronological succession as a single narrative for the recipients. We present the challenges encountered during the decomposition process and provide a description of the analysis output, which will serve as material for reflecting on an appropriate digital medium for the construction of strategic narratives.

People tend to attribute human traits to non-human entities, such as animals, God or robots. This tendency, referred to as anthropomorphism, is seen as a natural, automatic inclination. Within the framework of Dual Process Theories, some researchers have proposed a model of anthropomorphism that distinguishes between implicit and explicit processes. This model states that implicit anthropomorphism is a result of Type 1, intuitive processing, while explicit anthropomorphism is a result of Type 2, deliberative processing. The results obtained so far seem to corroborate this model. Although perfectly well founded, we believe that this model's clear separation between indirect measures - intended to measure implicit anthropomorphism - and direct measures - intended to measure explicit anthropomorphism - is a questionable premise from both a theoretical and a practical point of view. Contrary to these previous studies, we thus recommend testing the automaticity of anthropomorphism toward robots by manipulating the concurrent cognitive load while participants give their explicit attributions. Our hypotheses are that, since we believe anthropomorphism to be an automatic and implicit process, cognitive load will either (i) lead to an increase of the anthropomorphism attributions or (ii) result in an absence of difference between the different condition of load, proving that Type 2 processes are pointless in the anthropomorphizing course. We discuss and propose a paradigm aimed at testing our hypothesis.

This study reviews unresolved identifiable victim effect (IVE) issues. The IVE means that people are more willing to contribute to an identified victim than to an unidentified number of victims such as statistical information. Previous studies demonstrated the robustness of this phenomenon. However, these studies on IVE have been conducted primarily in Western cultures and have not been sufficiently examined whether they are applicable in East Asian cultures. In recent times, some studies have begun to explore cultural differences in IVE and suggest that IVE may be less likely to occur in East Asian cultures with a predominantly interdependent construal of self than in Western cultures with a predominantly independent construal of self. This is the first study to confirm the mechanism of IVE and the methods used to suppress it. Further, studies on meta-analysis and cultural differences in IVE are reviewed and prospects are discussed.

We investigate conjunctions of conditionals, iterated conditionals, and independence within the theory of logical operations on conditionals, where compound conditionals are suitably defined as conditional random quantities. We show how conjunctions of conditionals and conditionals which feature conditionals in the antecedent and the consequent can rationally be interpreted. In particular, we study the behavior of such objects under different logical constraints, by also considering a kind of `independence' property. Differently from alternative approaches, in our framework we avoid counterintuitive consequences, which is necessary for understanding, or improving, human and artificial rationality in general

This study investigates how relationships between speakers influence the interpretation of conditional statements “if p1 then q” and conditional reasoning, particularly among Japanese participants. The standard suppression paradigm of conditional reasoning reveals that adding additional conditional statements “if p2 then q” can suppress MP reasoning, “p1, therefore q”. Using the theory of politeness as a basis, this study examines whether the interpretation of ambiguous additional conditions changes depending on the speakers' relationship and how it impacts conclusions drawn. The results showed that the inclusion of an additional conditional decreased the perceived plausibility of conditional and inverse sentences, but the relationship between speakers had no significant impact on the interpretation of additional conditionals or probability assessments. MP, MT, and DA reasoning were affected by the presence of additional conditionals, while AC reasoning remained unaffected. The study underscores the importance of context and cultural nuances in the interpretation of conditional statements and highlights the need for further research to understand the factors that influence the interpretation of conditional sentences in different cultural settings.

This article highlights the difficulties of "classical" rationality, grounded on classical set-based logic and classical probability calculus, and explains how they can be overcome. Classical rationality may have gone some way towards realizing the age-old project of mechanizing thought, thus making possible the dazzling development of artificial intelligence we are witnessing. However, as shown from experimental data, it suffers from many biases that make it incapable of reliably modeling mental processes. This article shows that mental processes can be more reliably modelled within a generalized probability theory grounded on the vector formalism of quantum theory. The reason is that such a quantum-like approach to cognition is capable of accounting for the contextual, order and interference effects inherent to most of mental processes. An important consequence of this necessary shift, from classical to quantum cognition, is that Bayes’ rule, which plays a fundamental role in categorization tasks, must be replaced by a generalized probabilistic rule capable of accounting for order effects. It is shown that implementing this new probabilistic rule could significantly improve the current deep learning algorithms of artificial emotional intelligence.

The decline in violence and glowing awareness of human rights can be viewed as natural experiments in history answering the question if humans are moral creatures. A dual-process approach, which assumes the intuitive and reflective systems, is adopted to examine how the reflective system controls the intuitive system. which possibly produces our cruelty. Reading a story may enhance our reflective system to suppress our cruelty. However, the power of the reflective system is weak in arousing people’s action against violence. Hence, theory of mind (ToM) accompanied by emotional empathy is necessary for the suppression of cruelty. However, ToM is modular in nature and one of the subsystems of the intuitive system; hence, empathy is narrow in its focus, which possibly causes social fragmentation and political polarisation.

The tramway problem is a moral dilemma studied in psychology. It involves either saving five lives by sacrificing one life through an action (the so-called "utilitarian" response) or refusing to sacrifice one person for the benefit of five (the so-called "deontic" response). Within the Western population, most responses are utilitarian in the so called “lever” condition, in which you have to pull a lever to deviate the trajectory of the train from the five people onto the alone victim. Whereas in the “bridge” condition, the action is to push an overweight person on the tracks to stop the train and there the majority of the responses are deontic. Instead of the refusal to break a moral code, the deontic choice could be motivated by a strong feeling of responsibility toward the victim we have to kill in order to save the others. A disengagement from this responsibility should lead to a utilitarian response. We tested this hypothesis by adding two versions in which before acting, the approval of the sacrificed person is requested. Under these conditions, the responses are utilitarian in both scenarios.

Complex systems that push for decisions complicate the necessary process through unpredictability and the resulting uncertainty. At the same time, numerous methods of data analysis and enormous amounts of decentrally distributed data are available as a basis for decision-making. Social changes also support the desire to spread the challenge of decision-making over several shoulders. Optimal decision-making therefore requires an environment that combines the tactical rationality of artificial decision-making systems with the strategic competences of intuition and creativity of human groups.

The aim of this work is to show how metaphysical conceptions of chance can affect decision-making. For Europeans, chance is merely the measure of our ignorance. For Mohammedans, there is no such thing as chance. Mektoub, destiny is God's will. The endowment effect consists in attributing a higher value to an asset of which one is the owner. The literature shows that when a good is attributed at random, the endowment effect is no longer observed in Western populations. Will the same phenomenon be observed among Mohammedans? 180 adult participants were recruited: 60 French and 120 Arabs (60 Mohammedans and 60 non-practitioners). The experiment was conducted in French or Arabic. After rolling a die, participants received either object A or object B, then performed a task for 10 minutes before deciding whether or not to exchange the object they had received. The results show: 1) that there is no endowment effect for the French, 2) that there is an endowment effect for both Mohammedans (Z = 2.357 p < 001) and non-Mohammedans (Z = 2.182 p < 001). Metaphysical conceptions applied to chance modify decision-making.

Exploring the reasons why a person believes in some statements and not in other is as necessary as it is difficult, for it requires a very precise way to label our claims. Michel and al. (2018) proposed to make a difference between factual and opinion statements. The first being statements open to proof or disproof, the latter being matters of personal preference not amenable to falsification. Even if Brotherton and Son (2021) broadened the scope, adding the key factors of “agreement” and “knowledge”, these labels can and as a matter of fact should be questioned in three different ways. (1) Stating that a claim can be proved (or disproved) should require the knowledge of the proof itself; otherwise, it seems to remain a belief. The “fact” label should then be divided. (2) The Popperian conception of the proof (identified as a possible falsification) doesn’t seem to be common knowledge: is it adequate then to present the “opinion” label recovering such a definition? (3) The tendency to consider a third label –“article of faith” – seems problematic: why should it be treated differently from “opinion”?

This paper proposes a theory of human rationality. What allowed human beings to be animals who invented science, philosophy, technology? We rely on two discoveries in the cognitive science of reasoning. First, our spontaneous tendency for fallacies, especially the fallacies of the affirmation of the consequent and of the negation of the antecedent with conditionals. Second, our tendency to avoid these fallacies in normative conditionals. Analysing the avoidance of conditional fallacies in privilege-price relations, Tooby and Cosmides proposed the activation of an innate cheater detection device for the explanation of this avoidance. Given that not performing these fallacies is logically made possible by taking into account the dual operator of the conditional, we hold that the device is not only a cheater detection device but also an altruist detection device. To avoid the fallacies, we must also recognize the altruist, the one who pays the price and does not take the privilege (~P & Q). We then hold that human beings are spontaneously more logical in normative contexts and are much less in cognitive contexts. In cognitive contexts, we look for causality and conditional fallacies are abductive processes used for finding causes, which do not apply in normative contexts. So, we hold that an innate device has made us relatively competent for social life and. this gave us a foundation, not only for social living, but also for logical reasoning. Our neural plasticity has allowed us to transfer this competence to cognition and so, to invent technologies and to build science.

Custom structures Kanak society. For the European observer, custom consists in the reciprocal transfer of material goods between two groups. For the Kanak, "custom" links these two groups, enabling mutual recognition. Custom" punctuates every moment (large and small) in Kanak life. During "customs", Europeans will notice that the transfer of goods always tends to balance out. The aim of this paper is to study the notion of sharing within the framework of a comparative cultural (Europeans-Kanak) and developmental approach (children aged 4-5 and 9-10), using the dictator game. The literature shows that European children (aged 6-12) give on average between 29% and 35% of their initial endowment. Altruistic sharing begins at age 7. In Oceania, children (aged 3-5) give 47% of their initial endowment. At what age do Kanak children achieve a near-balanced sharing? At 4-5 years? Around 9-10 years? These are the questions our study will answer. 195 participants were recruited: 75 Kanak and 120 European children aged 4-5 and 9-10. The dictator game was played individually, in the vernacular language (Drehu, Nengone, French). The Europeans shared 26% and 46% of the initial endowment respectively, and the Kanak 54% and 46%. There is a clear interaction between age and ethnicity. Oceanic culture modifies the way people share, from an early age.

The endowment effect is one of the most robust cognitive biases. It means that we attribute a higher value to a good that we own than to the same good that we do not own. The exchange paradigm consists of proposing two goods A and B (same value, same level of interest), handing over one of the two goods and then asking the participant whether they would agree to exchange A for B or B for A. Few studies show that the endowment effect is eliminated in the paradigm framework (Apicella & al. 2014; Baratgin & al. 2022). For Baratgin & al. 2022, the absence of endowment effect results from the activation of pragmatic implicatures and the context in which the experience takes place. We propose to examine the context of the experimental situation in the Kanak tribe, with a Kanak experimenter testing the endowment effect in the Kanak language. The analysis will be carried out from an anthropological point of view, based on the notions of relationships, links and interaction.

The increasing use of robots with individuals with Autism Spectrum Disorders (ASD), with the notable aim of supporting the development of social skills, prompts the question of the extent to which children with ASD interact with robots as with social agents. In the literature, resource-sharing tasks (e.g. dictator game) have already been used to study the perception of robots in typically developing children, but few studies have focused on sharing in children with ASD. 38 children aged 9 to 13 (28 typically developing children and 10 children with ASD) played the dictator game with a humanoid robot NAO and a human experimenter, in counterbalanced order. They also answered an adapted version of the Godspeed questionnaire designed to measure anthropomorphism. The results indicate that the quantity of shared resources is better predicted by the Good personality score, which reflects the attribution of positive traits to the agent, than by the type of development: the higher the score, the greater the number of resources given by the participant. Contrary to our initial assumption, children with ASD tend to give fewer resources to the robot than to the human, unlike NT children who give the same amount to both agents. In the human experimenter condition, children with ASD and NT share similarly. Post-hoc analyses, however, indicate an effect of developmental type on Good personality scores, with lower scores for ASD than NT children. The attribution of positive traits to the agent could be the mediating variable between developmental type and sharing behavior.

Mindfulness can be defined relative to mind-wandering, while the former is associated with focused attention and enhanced monitoring of conscious experience, the latter corresponds to a process of distraction. The aim of this human-machine interaction study was to investigate the ocular correlates of the state of mindfulness by comparing it with mind-wandering and resting-state. To this end, experienced meditators and nonmeditators performed a point-fixation task while carrying out three different actions: meditating, reflecting on a philosophical question, and resting. Meditating induced mindfulness, reflecting induced mind-wandering, and resting-induced a resting-state. Eye movement recordings revealed a decrease in microsaccade amplitude and velocity during the meditation task, compared with the other two tasks. Participants also blinked more during the reflecting task than during the other two tasks, especially those in the experienced meditator group. These results suggest that microsaccades are indicators of sustained attention, and blinking of distraction, meaning that it may be possible to detect mind-wandering episodes versus states of mindfulness. Detection of this episodes will be used to develop a biofeedback device to learn mindfulness meditation.

When we make a decision without being influenced by external factors, we feel we are in total control of the subsequent decision. Indeed, recent research shed light on a prospective access to neural precursors of intention, allowing people to veto their impending decisions. Yet, whether the specific content of deliberation preceding free choice can be consciously access remains a debated question. Here we address this question by tracking in real time the neural signals supporting participants’ early deliberation of impending free decisions. Using a decision dissociation task, we designed an experiment where those signals played a crucial role in their ultimate choice. Our findings reveal that the content of early deliberation significantly influences participants' decisions, but that those deliberations remained mostly unconscious as the decision unfolded. Notably, despite the unconscious nature of early deliberations, participants frequently reported being conscious of the content of their forthcoming decision. We showed that those reports correlate with the presence of a neural marker of self-initiated decision. Together, our results indicate that the sense of agency over one’s decision is inflated, especially regarding the content of impending decision. By suggesting that we are only partially aware of our upcoming decisions, our study provides new insights into the metacognitive process supervising free choices.

In everyday life, adaptive behaviour depends on the ability to perceive action possibilities. The field of research on affordances has addressed and continues to explore how individuals perceive action possibilities through the coupling between the perceiver's capacity and environmental features. However, with the increasing digitalization of our environment, people now need to perceive action boundaries in a 2D-cyberspace. Consequently, researchers require new tasks to operationalize action boundary perception in 2D-cyberspace. The current article aims to propose a critical review of two action boundary perception tasks: the Perception-Action Coupling Task and the Action Perception Boundary Online Task, as a tool to assess the virtual action-boundary perception in the digital environment. First, we present the historical background and context of the emergence of these tasks. Then, we review the studies that have used these tasks. Finally, we conclude by discussing the potential of this task, as well as providing a critical description of the limitations that future research will need to address.

Anxiety is a mental disorder that is currently increasing in human daily life. Studies aimed to deepen the understanding of it, to minimize its negative impact on people's lives, have gained significant importance. In this context, the main focus of this work is to study the use of several physiological signals (electrocardiogram, electrodermal activity, and blood volume pulse) to predict the level of anxiety felt by a subject using four different approaches. These involve the use of a different number of features that are selected as more informative, and by training two classification models, with different properties. For the chosen approaches, the obtained results are compared and analysed to understand which performs better, i.e., which has a greater ability in anxiety recognition, being the anxiety level of each participant assessed through the application of the STICSA questionnaire. When selecting 20 features to train a Linear Discriminant Analysis model, an accuracy and precision of over 70% were achieved. This strategy presented the best performance as this model surpasses, for all the used metrics, the results obtained when using the models based on a decision tree. The encouraging obtained results indicate that the simultaneous use of different physiological signals to train the models applied in this study for predicting the level of anxiety is feasible.

The aim of the present study was to evaluate the ability of a robot to conduct cognitive training in elderly people with dementia. Fifty-two institutionalized elderly people were recruited and invited to participate in prospective memory training sessions, with or without a robot. They were divided into two equivalent groups in terms of age, level of education and MMSE score. Before and after the intervention phase, neuropsychological assessment was performed on each participant, including several cognitive and non cognitive (self-esteem) evaluations. Moreover, the sessions were recorded in order to compare the interaction behaviors of the 2 groups, using a validated observation grid. Results showed that : 1) the presence of the robot increases the interaction behaviors of the participants during the sessions (such as smiling, laughing, nodding, reaching out to others, talking, etc); 2) That prospective memory training resulted in a significant increase of prospective memory performance, attentional abilities and executive functioning, but this improvement did not differ between the 2 groups These findings confirm the positive impact of a robot as a mediator of cognitive training, but suggest that further research is necessary to determine the effectiveness of these tools, in comparison with traditional training with humans

Although the development of service robots is increasing, the opportunities and challenges of using robots to support persons with disabilities (PWD) have not been fully discussed. Therefore, a detailed review of previously developed robots for supporting PWD is performed to identify various perspectives and problems to contribute to the effective application of robots to support PWD. In order to understand international research trends, PubMed was used to search. The results identified 41 papers published on the subject of robot rehabilitation for PWD that satisfied the criteria. The disability of the subject that the robot supported was mainly a movement disorder causing a physical disability. Two meta-regression or systematic reviews were found in clinical research, targeting physical disabilities. Twenty-two pieces of engineering development research were conducted, seventeen of which targeted physical disabilities. The robot used in each study changed significantly with the target PWD(s) and research areas. There were few person-type robots in the study of an engineering system. Much research has also been conducted on the robot controller. Outside of engineering system research, humanoid robots have been used in research on developmental diseases. In addition to engineering research to develop robots and medical and rehabilitation research to evaluate their effectiveness, research is needed to support the effective use of robots and accumulate practices. In particular, it will be important to promote clinical research on cognitive system disorders such as developmental disorders. I expect this review will help add a new perspective to future robot development and evaluation research.

Determining the principles for building an intelligent management system aimed at making optimal decisions becomes an urgent need to achieve sustainable development of cooperating enterprises and requires the search for appropriate management models based on data dynamics. The purpose of the work was to indicate the determinants of sustainable development of collaborating enterprises used to determine the management model. The application of systems analysis and multi-agent simulation principles to human-system interactions has indicated a way to construct the data-driven decision making reinforcement learning model. The results obtained indicate that such management eliminates subjectivity in decision-making, leading to the stabilization of the economic development of cooperating enterprises, and hence to sustainable development.

The fourth industrial revolution brought about a constant need to increase processing speed as well as increase throughput. This requires the improvement of existing and the invention of new tools for creating, sharing and consuming data, which can be compared with digital evolution. As a result of this evolution, new creators and consumers of data appear. Data, its creators and consumers, is a network of interconnected elements, part of a data ecosystem that is constantly evolving. Obviously, the ecosystem includes data that needs to be integrated from disparate sources; various types of analysis and skills to obtain information; actions in accordance with the acquired knowledge; as well as tools, applications and infrastructure for storing, processing and distributing data, depending on the purpose. Such multidimensionality makes the training of IT specialists a rather complicated process. There is a need to constantly search for such key skills and knowledge that would make it easy to adapt to the new qualities and properties of the data ecosystem. The purpose of this work is to analyze the properties of the data ecosystem and highlight the key aspects of training IT specialists. The results of the study are illustrated by examples of the use of neural networks in the educational process.

In the context of the e-FRAN program, our research focused on engineering freshmen’s learning processes of 3-D modeling in relation to their spatial skills. One of the approaches developed consisted in identifying the students’ sequences of micro operations using a double observation of their actions by observing their on-screen activity, and of their gestures and postures. The objective consisted in determining the logical sequences of both their bodily attitudes, which are related to their manipulation of the software, and their conceptual approaches, which are linked to the results they obtained by interacting with the software. This preliminary characterization aimed to establish the possible recurring patterns in human behavior in the context studied, to then determine the robustness of the codings with regard to the software functionalities, to finally identify characteristic sequences of operations. The later phase consisted in feeding a parser with the encoded traces left by the subjects’ interactions with the software interface, using a heuristic library of actions. The objective was to identify sequences indicative of efficient, or non-efficient, conceptual processes in the task performance. The encoded operations logs revealed recurring sequences, which were then processed statistically to identify prototypes of rationality of intentions. The encoded sequences were submitted to a further manual encoding to identify strategies that arose at the students’ first encounter with the modeling interface.

The aim of this research is to take account of the relationship between the members in the education system and the rationality of innovation. On the one hand, there is the institution, which communicates and encourages innovation while at the same time communicating about existing actions that it has selected. On the other hand, the local members of the educational community - headteachers, teachers, pupils, associations, researchers and parents - who inform their institution of actions in progress and their evaluations. The aim is to identify what comes under the heading of 'creative rationality' (Forest, 2018), the relationship between support, innovation and public policy (Cros, 1996, 1997, 2009, 2020) and creative potential (T.M. Amabile, 1996; H. Gardner, G. Delacôte and N. Weinwurzel, 1996; T. Lubart, C. Mouchiroud, S. Tordjman, and F. Zenasni, 2003). For this work, a database was set up with more than 5,650 records on the "portal for educational innovation and experimentation" of the French Ministry of Education and Youth (https://vu.fr/yHDcg). They represent what the institution has prioritised based on feedback from the field between 2000 and 2020 in terms of educational innovation and experimentation. The aim of the multi-factorial analysis carried out is to identify the existence of clusters that are meaningful in terms of the members and their own representations of what innovation means within the education system, in terms of a rational common ground that can be communicated and potentially transposed to other contexts, while taking account of educational otherness.

Integrating educational robotics in early childhood educational programs presumes an understanding of how children perceive and represent them. However, the literature review provides little data about children's representations and ideas for robots and their differentiation in programming them, specifically in preschool children (4-6 years). Greff (2002) studied the representations of eight pre-schoolers (5 years) about the programmable toy Roamer showing how they depict the programming procedure as was taught. Research by Bhamjee et al. (2010) and Berana et al. (2011) provide data on children's (5-12 years) perceptions and interpretations of robots as well as their behaviour. Monaco et al. (2018), exploring preschool children's cognitive representations of robots showed that only some children mentioned some programming-related concepts. Mioduser & Kuperman (2020) argue that children have more difficulty explaining the behaviour of a robot than programming such behaviour. Representations of early childhood children and their development of computational thinking skills while tangibly programming a robot are presented in this study. Data was taken from multiple case studies in preschool settings. Research protocols recorded children's representations in individual basis semi-structured interviews in pre and post-sessions of an almost monthly educational intervention delivered by teachers in their classrooms. Some examples of depictions on children's drawings will also be presented, showing children's learning development in their computational thinking skills. The results show that while children attribute an animate identity to a robot, at the same time, they state and depict data for its properties and basic operational features.

Understanding and dealing with Climate Change issues can be of enormous difficulty due to the cognitive biases and barriers that may hinder the development of Environmental Literacy. Real and truthful communication is necessary, but not enough to obtain effects in promoting understanding and pro-environmental behaviors in the presence of biases. Therefore, the development of Critical thinking on climate change appears to be an important tool in the development of Environmental Literacy. Critical thinking is a discipline which aim is to render people more careful in approving/rejecting arguments, more capable of constructively intervening in discussions and debates, and more adept at uncovering fallacies and rhetorical tricks in public discussion. Fostering the formation of critical thinking equips with the necessary tools to think, formulate correct opinions, accept the plurality of viewpoints, and develop deliberative skills. This discipline is therefore being adopted within the UE EduS4EL Project whose objective is the development of educational activities - among secondary school students - aimed at promoting behavioural changes in individual and collective choices and in consumption habits through innovative strategies and methodologies applied to environment goals. The AI could be included in this process, as a tool such as ChatGPT to which the key materials that are the subject of the project could be given, so that the students can dialogue with it, learning by asking. A kind of return to Socratic maieutics, to dialogic learning. Active learning, since ChatGpt is dialogue, it asks for our intervention and we have to know how to ask.

The aim of the EduS4EL project is the development of educational activities applied to environment and climate goals, to promote, among secondary school children, behavioural changes in individual and collective choices and in consumption habits. The process of improving Environmental Literacy cannot consist in merely implementing an educational intervention: it is preliminarily crucial to investigate the concern, feelings, awareness, and knowledge about climate change to address a relevant intervention and communication. We, therefore, conducted a pilot study in which secondary school students were asked (1) their concerns, knowledge, habits, and feelings, (2) a definition of eight frequently used terms linked to climate change, and (3) how difficult they considered each term, (4) their involvement on the topic and (5) their feedback on the lectures and activities carried out during the pilot intervention. Preliminary results illustrate considerable concern about climate change as, when asked how they feel about climate change, some more frequently used terms were «worried» and «powerless». These terms also suggest that participants perceive that they do not have the necessary tools to cope with this issue. For what concerns the definitions of climate change terms, most of them were considered as easily understandable; however, most were erroneous or not linked to the context of climate and indirectly revealed some biases. These results highlight that it seems to be essential for educational interventions to represent and communicate the complex concepts related to climate change and environmental issues in a more accessible way, always considering biases and reasoning fallacies.

The Sustainable Development Goals established by the UN in the 2030 Agenda involve developing training, educational and social actions that raise awareness of sustainability and environmental care. In this sense, schools and universities play a fundamental role in establishing lines of work that enable a human, committed and socially responsible dimension with the planet. In line with these needs, we present the key lines developed by the ReMida creative recycling centre, as an educational, cultural and ethical project that is sensitive to training, research and the reuse of waste materials. It is a local initiative but developed as a global phenomenon, which invites social awareness, the training of education professionals and the promotion of actions for sustainable human development. It is an example of globalisation, with an international projection, which is committed to implementing sustainable initiatives in other educational contexts. This is the case of the ReCrea project, an educational proposal that is sensitive to reduction, recycling and educational, creative and artistic recreation using different discarded materials derived from any resource, including electronic and technological devices.

We review some of the literature on Bayesian Reasoning with special emphasis on typical fallacies reported in during the last decades. We reduce these fallacies to the basic difficulty encountered in proportional thinking. We cite and illustrate results from neuroscientific research confirming the thesis that working with proportions is much more difficult than working with natural numbers. We describe studies thet show how geometric representations of proportions can foster reasoning about probabilities and risks.