Search

International audience; In this paper we present a new approach for semantic recognition in the context of robotics. When a robot evolves in its environment, it gets 3D information given either by its sensors or by its own motion through 3D reconstruction. Our approach uses (i) 3D-coherent synthesis of scene observations and (ii) mix them in a multi-view framework for 3D labeling. (iii) This is efficient locally (for 2D semantic segmentation) and globally (for 3D structure labeling). This allows to add semantics to the observed scene that goes beyond simple image classification, as shown on challenging datasets such as SUNRGBD or the 3DRMS Reconstruction Challenge.

We propose a depth map inference system from monocular videos based on a novel dataset for navigation that mimics aerial footage from gimbal stabilized monocular camera in rigid scenes. Unlike most navigation datasets, the lack of rotation implies an easier structure from motion problem which can be leveraged for different kinds of tasks such as depth inference and obstacle avoidance. We also propose an architecture for end-to-end depth inference with a fully convolutional network. Results show that although tied to camera inner parameters, the problem is locally solvable and leads to good quality depth prediction.

International audience; Some studies hypothesize a strong interdependence between speech and tool use development in the first two years of life. To help understand the underlying mechanisms, we present the first robotic model learning both speech and tool use from scratch. It focuses on the role of one important form of body babbling where exploration is directed towards self-generated goals in free play, combined with imitation learning of a contingent caregiver. This model does not assume capabilities for complex sequencing and combinatorial planning which are often considered necessary for tool use. Yet, we show that the mechanisms in this model allow a learner to progressively discover how to grab objects with the hand, how to use objects as tools to reach further objects, how to produce vocal sounds, and how to leverage these vocal sounds to use a caregiver as a social tool to retrieve objects. Also, the discovery that certain sounds can be used as a social tool further guides vocal learning. This model predicts that the grounded exploration of objects in a social interaction scenario should accelerate infant vocal learning of accurate sounds for these objects' names.

International audience; The schooling of children with autism spectrum disorder is a challenge that relies on all stakeholders (families, schools, and medical-social professionals). In many countries, the Individualized Education Plan (IEP) is a monitoring tool of the schooling process, implemented by the aforementioned stakeholders, but they face several barriers to the co-education process. The CoEd web app was designed using participatory methods in France, with interactive services for the children's follow-up and the three-way collaboration. CoEd applicability was assessed in Belgian cultural context (Walloon territory), as they experiences similar issues.The results showed comparable results between the two countries. The Belgian audience appreciated CoEd and would be ready to use it in real conditions, as is currently the case in France. Thus, this participatory design work developed a tool that meets the needs expressed by the French participants and those of stakeholders beyond French borders.; La scolarisation des enfants atteints de troubles du spectre autistique est un défi qui repose sur l'ensemble des acteurs (familles, écoles et professionnels médico-sociaux). Dans de nombreux pays, le plan d'éducation individualisé (PEI) est un outil de suivi du processus de scolarisation, mis en œuvre par les parties prenantes susmentionnées, mais celles-ci se heurtent à plusieurs obstacles dans le processus de coéducation. L'application web CoEd a été conçue en utilisant des méthodes participatives en France, avec des services interactifs pour le suivi des enfants et la collaboration tripartite. L'applicabilité de CoEd a été évaluée dans le contexte culturel belge (territoire wallon), étant donné que les deux pays connaissent des problèmes similaires. Le public belge a apprécié CoEd et serait prêt à l'utiliser dans des conditions réelles, comme c'est actuellement le cas en France. Ainsi, ce travail de conception participative a permis de développer un outil qui répond aux besoins exprimés par les participants français et à ceux des acteurs au-delà des frontières françaises.Traduit avec www.DeepL.com/Translator (version gratuite)

International audience

International audience; In the context of sensory or higher-level cognitive processing, we present a recurrent neural network model, similar to the popular dynamic neural field (DNF) model, for performing approximate probabilistic computations. The model is biologically plausible, avoids impractical schemes such as log-encoding and noise assumptions, and is well-suited for working in stacked hierarchies. By Lyapunov analysis, we make it very plausible that the model computes the maximum a posteriori (MAP) estimate given a certain input that may be corrupted by noise. Key points of the model are its capability to learn the required posterior distributions and represent them in its lateral weights, the interpretation of stable neural activities as MAP estimates, and of latency as the probability associated with those estimates. We demonstrate for in simple experiments that learning of posterior distributions is feasible and results in correct MAP estimates. Furthermore, a pre-activation of field sites can modify attractor states when the data model is ambiguous, effectively providing an approximate implementation of Bayesian inference.

International audience; We propose an incremental, highly parallelizable, and constant-time complexity neural learning architecture for multi-class classification (and regression) problems that remains resource-efficient even when the number of input dimensions is very high (≥ 1000). This so-called projection-prediction (PROPRE) architecture is strongly inspired by biological information processing in that it uses a prototype-based, topologically organized hidden layer trained with the SOM learning rule that updates hidden layer weights whenever an error occurs. The SOM learning adapts only the weights of localized neural sub-populations that are similar to the input, which explicitly avoids the catastrophic forgetting effect of MLPs in case new input statistics are presented. The readout layer applies linear regression to hidden layer activities subjected to a transfer function, making the whole system capable of representing strongly non-linear decision boundaries. The resource-efficiency of the algorithm stems from the fact of approximating similarity in the input space by proximity in the SOM layer due to the topological SOM projection property. This avoids the storage of inter-cluster distances (quadratic in number of hidden layer) or input space covariance matrices (quadratic in input dimensions) as K-means, RBF or LWPR would have to do. Tests on the popular MNIST handwritten digit benchmark show that the algorithm compares favorably to state-of-the-art results, and parallelizability is demonstrated by analyzing the efficiency of a parallel GPU implementation of the architecture.

International audience; PROPRE is a generic and cortically inspired framework that provides online input/output relationship learning. The input data flow is projected on a self-organizing map that provides an internal representation of the current stimulus. From this representation, the system predicts the value of the output target. A predictability measure, based on the monitoring of the prediction quality, modulates the projection learning so that to favor learning of representations that are helpful to predict the output. In this article, we study PROPRE when the input/output relationship is only defined in a small subspace of the input space, that we define as a partially learnable environment. This problem, which is not typical of the machine learning field, is however crucial for the robotic developmental field. Indeed, robots face high dimensional sensory-motor environments where large areas of these sensory-motor spaces are not learnable since a motor action cannot have a consequence on every perception each time. We show that the use of the predictability measure in PROPRE leads to an autonomous gathering of local representations where the input data are related to the output value, thus providing good classification performance as the system will learn the input/output function only where it is defined.

International audience

International audience; Self-organizing map (SOM) is a powerful paradigm that is extensively applied for clustering and visualization purpose. It is also used for regression learning, especially in robotics, thanks to its ability to provide a topological projection of high dimensional non linear data. In this case, data extracted from the SOM are usually restricted to the best matching unit (BMU), which is the usual way to use SOM for classification , where class labels are attached to individual neurons. In this article, we investigate the influence of considering more information from the SOM than just the BMU when performing regression. For this purpose , we quantitatively study several output functions for the SOM, when using these data as input of a linear regression, and find that the use of additional activities to the BMU can strongly improve regression performance. Thus, we propose an unified and generic framework that embraces a large spectrum of models from the traditional way to use SOM, with the best matching unit as output, to models related to the radial basis function network paradigm, when using local receptive field as output.

In this article, we present some preliminary work on integrating an artificial curiosity mechanism in PROPRE, a generic and modular neural architecture, to obtain online, open-ended and active learning of a sensory-motor space, where large areas can be unlearnable. PROPRE consists of the combination of the projection of the input motor flow, using a self-organizing map, with the regression of the sensory output flow from this projection representation, using a linear regression. The main feature of PROPRE is the use of a predictability module that provides an interestingness measure for the current motor stimulus depending on a simple evaluation of the sensory prediction quality. This measure modulates the projection learning so that to favor the representations that predict the output better than a local average. Especially, this leads to the learning of local representations where an input/output relationship is defined [1]. In this article, we propose an artificial curiosity mechanism based on the monitoring of learning progress, as proposed in [2], in the neighborhood of each local representation. Thus, PROPRE simultaneously learns interesting representations of the input flow (depending on their capacities to predict the output) and explores actively this input space where the learning progress is the higher. We illustrate our architecture on the learning of a direct model of an arm whose hand can only be perceived in a restricted visual space. The modulation of the projection learning leads to a better performance and the use of the curiosity mechanism provides quicker learning and even improves the final performance.

International audience; PROPRE is a generic and modular neural learning paradigm that autonomously extracts meaningful concepts of multimodal data flows driven by predictability across modalities in an unsupervised, incremental and online way. For that purpose, PROPRE consists of the combination of projection and prediction. Firstly, each data flow is topologically projected with a self-organizing map, largely inspired from the Kohonen model. Secondly, each projection is predicted by each other map activities, by mean of linear regressions. The main originality of PROPRE is the use of a simple and generic predictability measure that compares predicted and real activities for each modal stream. This measure drives the corresponding projection learning to favor the mapping of predictable stimuli across modalities at the system level (i.e. that their predictability measure overcomes some threshold). This predictability measure acts as a self-evaluation module that tends to bias the representations extracted by the system so that to improve their correlations across modalities. We already showed that this modulation mechanism is able to bootstrap representation extraction from previously learned representations with artificial multimodal data related to basic robotic behaviors and improves performance of the system for classification of visual data within a supervised learning context. In this article, we improve the self-evaluation module of PROPRE, by introducing a sliding threshold, and apply it to the unsupervised classification of gestures caught from two time- of-flight (ToF) cameras. In this context, we illustrate that the modulation mechanism is still useful although less efficient than purely supervised learning.

International audience; PROPRE is a generic and modular unsupervised neural learning paradigm that extracts meaningful concepts of multimodal data flows based on predictability across modalities. It consists on the combination of three modules. First, a topological projection of each data flow on a self-organizing map. Second, a decentralized prediction of each projection activity from each others map activities. Third, a predictability measure that compares predicted and real activities. This measure is used to modulate the projection learning so that to favor the mapping of predictable stimuli across modalities. In this article, we use Kohonen map for the projection module, linear regression for the prediction one and we propose multiple generic predictability measures. We illustrate the properties and performances of PROPRE paradigm on a challenging supervised classification task of visual pedestrian data. The modulation of the projection learning by the predictability measure improves significantly classification performances of the system independently of the measure used. Moreover, PROPRE provides a combination of interesting functional properties, such as a dynamical adaptation to input statistic variations, that is rarely available in other machine learning algorithms.

International audience; PROPRE is a generic and semi-supervised neural learning paradigm that extracts meaningful concepts of multimodal data flows based on predictability across modalities. It consists on the combination of two computational paradigms. First, a topological projection of each data flow on a self-organizing map (SOM) to reduce input dimension. Second, each SOM activity is used to predict activities in all other SOMs. Predictability measure, that compares predicted and real activities, is used to modulate the SOM learning to favor mutually predictable stimuli. In this article, we study PROPRE applied to a classical visual pedestrian data classification task. The SOM learning modulation introduced in PROPRE improves significantly classification performance.

International audience; In recent years, many psycho-educational technologies were studied to address the school-related difficulties encountered by students with autism spectrum disorder (ASD). However, most of them remain individual-centered and do not consider the social environment. To fill this gap, this study reports on the user-centered design of a web-based support tool, which aims to support communication and coordination between parents, school staff, and health professionals of middle and high school students with ASD, in the context of elaborating, implementing, and following an Individualized Education Plan (IEP). The study followed a two-step design process: 1) a need analysis for identifying information domains deemed important by the stakeholders; 2) through a participative iterative design process, a panel of professionals and parents provided ideas and feedback on the design, which was integrated into subsequent prototype versions of the "ToGather" app.; Ces dernières années, de nombreuses technologies psycho-éducatives ont été étudiées pour répondre aux difficultés scolaires rencontrées par les élèves avec trouble du spectre autistique (TSA). Cependant, la plupart d'entre elles restent centrées sur l'individu et ne prennent pas en compte l'environnement social. Pour combler cette lacune, cette étude porte sur la conception centrée-utilisateur d'un outil en ligne, qui vise à soutenir la communication et la coordination entre les parents, le personnel scolaire et les professionnels médico-sociaux des élèves de collège et de lycée atteints de TSA, dans le contexte de l'élaboration, de la mise en œuvre et du suivi d'un projet personnalisé de scolarisation (PPS). L'étude a suivi un processus de conception en deux étapes : 1) une analyse des besoins pour identifier les domaines d'information jugés importants par les parties prenantes ; 2) par le biais d'un processus de conception participatif et itératif, un panel de professionnels et de parents a fourni des idées et des commentaires sur la conception, qui ont été intégrés dans les versions prototypes ultérieures de l'application "ToGather".

International audience; Organizing information is beneficial to episodic memory performance. Among several possible organizational strategies, two consist of organizing the information in semantic clusters (semantic organization) or self-organizing the information based on new associations that do not exist in semantic memory (subjective organization). Here, we investigated in a single study how these two organizational behaviors were underlined by different controlled processes and whether these relations were subjected to age-related differences. We tested 123 younger adults (n = 63) and older adults (n = 60) on two episodic memory tasks, one where the words were organizable and another where the words were not organizable, allowing for semantic and subjective organization, respectively. Additionally, participants were tested on three cognitive control tasks (Wisconsin Card Sorting Test, Stroop Test and Trail Making Test) and three working memory tasks (Backward Digit Span, Alpha Span and N-back test). Results revealed well-established age-related differences in terms of recall performance and organizational strategy implementation. More importantly, we found evidence that the different cognitive tests statistically yielded two different latent factors, a cognitive control factor and a working memory factor. Based on this dissociation, we found that only cognitive control contributed to semantic organization in all age groups whereas only working memory contributed to subjective organization, also in all age groups. These results shed new lights on our understanding of how controlled processes differently contribute to organizational behaviors in episodic memory.

Understanding the impact of robot errors in child-robot-interactions (CRI) is critical, as current technological systems are still limited and may randomly present a variety of mistakes during interactions with children. In this study we manipulate a task-based error of a NAO robot during a semi-autonomous computational thinking task implemented with the Cozmo robot. Data from 72 children aged 7–10 were analysed regarding their attitudes towards NAO (social trust, competency trust, liking, and perceived agency), their behaviour towards the robot (self-disclosure, following recommendations), as well as their task performance. We did not find quantitative effects of the robot’s error on children’s self-reported attitudes, behaviour, or task performance. Age was also not significantly related to either social attitudes or behaviours towards NAO, although there were some age-related differences in task performance. Potential reasons behind the lack of statistical effects and limitations of the study with regards to the manipulation of robot errors are discussed and insights into the design of future CRI studies provided.

Lenia is a family of cellular automata (CA) generalizing Conway's Game of Life to continuous space, time and states. Lenia has attracted a lot of attention because of the wide diversity of self-organizing patterns it can generate. Among those, some spatially localized patterns (SLPs) resemble life-like artificial creatures. However, those creatures are found in only a small subspace of the Lenia parameter space and are not trivial to discover, necessitating advanced search algorithms. We hypothesize that adding a mass conservation constraint could facilitate the emergence of SLPs. We propose here an extension of the Lenia model, called Flow Lenia, which enables mass conservation. We show a few observations demonstrating its effectiveness in generating SLPs with complex behaviors. Furthermore, we show how Flow Lenia enables the integration of the parameters of the CA update rules within the CA dynamics, making them dynamic and localized. This allows for multi-species simulations, with locally coherent update rules that define properties of the emerging creatures, and that can be mixed with neighbouring rules. We argue that this paves the way for the intrinsic evolution of self-organized artificial life forms within continuous CAs.

In order to train children's ability to ask curiosity-driven questions, previous research has explored designing specific exercises relying on providing semantic and linguistic cues to help formulate such questions. But despite showing pedagogical efficiency, this method is still limited as it relies on generating the said cues by hand, which can be a very costly process. In this context, we propose to leverage advances in the natural language processing field (NLP) and investigate the efficiency of using a large language model (LLM) for automating the production of the pedagogical content of a curious question-asking (QA) training. We study generating the said content using the "prompt-based" method that consists of explaining the task to the LLM in natural text. We evaluate the output using human experts annotations and comparisons with hand-generated content. Results suggested indeed the relevance and usefulness of this content. We also conduct a field study in primary school (75 children aged 9-10), where we evaluate children's QA performance when having this training. We compare 3 types of content : 1) hand-generated content that proposes "closed" cues leading to predefined questions; 2) GPT-3-generated content that proposes the same type of cues; 3) GPT-3-generated content that proposes "open" cues leading to several possible questions. We see a similar QA performance between the two "closed" trainings (showing the scalability of the approach using GPT-3), and a better one for participants with the "open" training. These results suggest the efficiency of using LLMs to support children in generating more curious questions, using a natural language prompting approach that affords usability by teachers and other users not specialists of AI techniques. Furthermore, results also show that open-ended content may be more suitable for training curious question-asking skills.

The framework of Language Games studies the emergence of languages in populations of agents. Recent contributions relying on deep learning methods focused on agents communicating via an idealized communication channel, where utterances produced by a speaker are directly perceived by a listener. This comes in contrast with human communication, which instead relies on a sensory-motor channel, where motor commands produced by the speaker (e.g. vocal or gestural articulators) result in sensory effects perceived by the listener (e.g. audio or visual). Here, we investigate if agents can evolve a shared language when they are equipped with a continuous sensory-motor system to produce and perceive signs, e.g. drawings. To this end, we introduce the Graphical Referential Game (GREG) where a speaker must produce a graphical utterance to name a visual referent object consisting of combinations of MNIST digits while a listener has to select the corresponding object among distractor referents, given the produced message. The utterances are drawing images produced using dynamical motor primitives combined with a sketching library. To tackle GREG we present CURVES: a multimodal contrastive deep learning mechanism that represents the energy (alignment) between named referents and utterances generated through gradient ascent on the learned energy landscape. We, then, present a set of experiments and metrics based on a systematic compositional dataset to evaluate the resulting language. We show that our method allows the emergence of a shared, graphical language with compositional properties.

How can a population of reinforcement learning agents autonomously learn a diversity of cooperative tasks in a shared environment? In the single-agent paradigm, goal-conditioned policies have been combined with intrinsic motivation mechanisms to endow agents with the ability to master a wide diversity of autonomously discovered goals. Transferring this idea to cooperative multi-agent systems (MAS) entails a challenge: intrinsically motivated agents that sample goals independently focus on a shared cooperative goal with low probability, impairing their learning performance. In this work, we propose a new learning paradigm for modeling such settings, the Decentralized Intrinsically Motivated Skill Acquisition Problem (Dec-IMSAP), and employ it to solve cooperative navigation tasks. Agents in a Dec-IMSAP are trained in a fully decentralized way, which comes in contrast to previous contributions in multi-goal MAS that consider a centralized goal-selection mechanism. Our empirical analysis indicates that a sufficient condition for efficiently learning a diversity of cooperative tasks is to ensure that a group aligns its goals, i.e., the agents pursue the same cooperative goal and learn to coordinate their actions through specialization. We introduce the Goal-coordination game, a fully-decentralized emergent communication algorithm, where goal alignment emerges from the maximization of individual rewards in multi-goal cooperative environments and show that it is able to reach equal performance to a centralized training baseline that guarantees aligned goals. To our knowledge, this is the first contribution addressing the problem of intrinsically motivated multi-agent goal exploration in a decentralized training paradigm.

Large Language Models (LLMs) have in recent years demonstrated impressive prowess in natural language generation. A common practice to improve generation diversity is to sample multiple outputs from the model. However, there lacks a simple and robust way of selecting the best output from these stochastic samples. As a case study framed in the context of question generation, we propose two prompt-based approaches to selecting high-quality questions from a set of LLM-generated candidates. Our method works under the constraints of 1) a black-box (non-modifiable) question generation model and 2) lack of access to human-annotated references -- both of which are realistic limitations for real-world deployment of LLMs. With automatic as well as human evaluations, we empirically demonstrate that our approach can effectively select questions of higher qualities than greedy generation.

International audience; End-to-end reinforcement learning on images showed significant performance progress in the recent years, especially with regularization to value estimation brought by data augmentation (Yarats et al., 2020). At the same time, domain randomization and representation learning helped push the limits of these algorithms in visually diverse environments, full of distractors and spurious noise, making RL more robust to unrelated visual features. We present DIQL, a method that combines risk invariant regularization and domain randomization to reduce out-of-distribution (OOD) generalization gap for temporal-difference learning. In this work, we draw a link by framing domain randomization as a richer extension of data augmentation to RL and support its generalized use. Our model-free approach improve baselines performances without the need of additional representation learning objectives and with limited additional computational cost. We show that DIQL outperforms existing methods on complex visuo-motor control environment with high visual perturbation. In particular, our approach achieves state-of the-art performance on the Distracting Control Suite benchmark, where we evaluate the robustness to a number of visual perturbators, as well as OOD generalization and extrapolation capabilities.

An intriguing feature of the human species is our ability to continuously invent new problems and to proactively acquiring new skills in order to solve them: what is called open-ended skill acquisition (OESA). Understanding the mechanisms underlying OESA is an important scientific challenge in both cognitive science (e.g. by studying infant cognitive development) and in artificial intelligence (aiming at computational architectures capable of open-ended learning). Both fields, however, mostly focus on cognitive and social mechanisms at the scale of an individual’s life. It is rarely acknowledged that OESA, an ability that is fundamentally related to the characteristics of human intelligence, has been necessarily shaped by ecological, evolutionary and cultural mechanisms interacting at multiple spatiotemporal scales. In this thesis, I present a research program aiming at understanding, modelingand simulating the dynamics of OESA in artificial systems, grounded in theories studying its eco-evolutionary bases in the human species. It relies on a conceptual framework expressing the complex interactions between environmental, adaptive, multi-agent and cultural dynamics. Three main research questions are developed and I present a selection of my contributions for each of them.- What are the ecological conditions favoring the evolution of skill acquisition?- How to bootstrap the formation of a cultural repertoire in populations of adaptive agents?- What is the role of cultural evolution in the open-ended dynamics of human skill acquisition?By developing these topics, we will reveal interesting relationships between theories in human evolution and recent approaches in artificial intelligence. This will lead to the proposition of a humanist perspective on AI: using it as a family of computational tools that can help us to explore and study the mechanisms driving open-ended skill acquisition in both artificial and biological systems, as a way to better understand the dynamics of our own species within its whole ecological context. This document presents an overview of my scientific trajectory since the start of my PhD thesis in 2007, the detail of my current research program, a selection of my contributions as well as perspectives for future work.

International audience

International audience

International audience; Curiosity is an important factor that enables independent learning in children. Research suggests that it is also a competence that can be fostered by training specific metacognitive skills and information-searching behaviors. In this light, we develop a conversational agent that helps children generate curiosity-driven questions, and encourages their use to lead autonomous explorations and gain new knowledge. The study was conducted with 51 primary school students aged between 9 and 10 years old who interacted with either a neutral agent or an “incentive” agent that helped curiosity-driven questioning by offering specific semantic cues. Results showed a significant increase in the number and the quality of the questions generated with the “incentive” agent. This interaction also resulted in longer explorations and stronger learning progress.Together, our results suggest that the more our agent is able to train children’s curiosity-related metacognitive skills, the better they can maintain their information-searching behaviors and the more new knowledge they are likely to acquire.

To solve difficult tasks, humans ask questions to acquire knowledge from external sources. In contrast, classical reinforcement learning agents lack such an ability and often resort to exploratory behavior. This is exacerbated as few present-day environments support querying for knowledge. In order to study how agents can be taught to query external knowledge via language, we first introduce two new environments: the grid-world-based Q-BabyAI and the text-based Q-TextWorld. In addition to physical interactions, an agent can query an external knowledge source specialized for these environments to gather information. Second, we propose the "Asking for Knowledge" (AFK) agent, which learns to generate language commands to query for meaningful knowledge that helps solve the tasks. AFK leverages a non-parametric memory, a pointer mechanism and an episodic exploration bonus to tackle (1) irrelevant information, (2) a large query language space, (3) delayed reward for making meaningful queries. Extensive experiments demonstrate that the AFK agent outperforms recent baselines on the challenging Q-BabyAI and Q-TextWorld environments., Comment: ICML 2022; Project page: https://ioujenliu.github.io/AFK/

Building autonomous machines that can explore open-ended environments, discover possible interactions and build repertoires of skills is a general objective of artificial intelligence. Developmental approaches argue that this can only be achieved by $autotelic$ $agents$: intrinsically motivated learning agents that can learn to represent, generate, select and solve their own problems. In recent years, the convergence of developmental approaches with deep reinforcement learning (RL) methods has been leading to the emergence of a new field: $developmental$ $reinforcement$ $learning$. Developmental RL is concerned with the use of deep RL algorithms to tackle a developmental problem -- the $intrinsically$ $motivated$ $acquisition$ $of$ $open$-$ended$ $repertoires$ $of$ $skills$. The self-generation of goals requires the learning of compact goal encodings as well as their associated goal-achievement functions. This raises new challenges compared to standard RL algorithms originally designed to tackle pre-defined sets of goals using external reward signals. The present paper introduces developmental RL and proposes a computational framework based on goal-conditioned RL to tackle the intrinsically motivated skills acquisition problem. It proceeds to present a typology of the various goal representations used in the literature, before reviewing existing methods to learn to represent and prioritize goals in autonomous systems. We finally close the paper by discussing some open challenges in the quest of intrinsically motivated skills acquisition.

Reinforcement learning (RL) in long horizon and sparse reward tasks is notoriously difficult and requires a lot of training steps. A standard solution to speed up the process is to leverage additional reward signals, shaping it to better guide the learning process. In the context of language-conditioned RL, the abstraction and generalisation properties of the language input provide opportunities for more efficient ways of shaping the reward. In this paper, we leverage this idea and propose an automated reward shaping method where the agent extracts auxiliary objectives from the general language goal. These auxiliary objectives use a question generation (QG) and question answering (QA) system: they consist of questions leading the agent to try to reconstruct partial information about the global goal using its own trajectory. When it succeeds, it receives an intrinsic reward proportional to its confidence in its answer. This incentivizes the agent to generate trajectories which unambiguously explain various aspects of the general language goal. Our experimental study shows that this approach, which does not require engineer intervention to design the auxiliary objectives, improves sample efficiency by effectively directing exploration., 24 pages, 16 figures, 5 tables

The human cultural repertoire relies on innovation: our ability to continuously and hierarchically explore how existing elements can be combined to create new ones. Innovation is not solitary, it relies on collective accumulation and merging of previous solutions. Machine learning approaches commonly assume that fully connected multi-agent networks are best suited for innovation. However, human laboratory and field studies have shown that hierarchical innovation is more robustly achieved by dynamic communication topologies. In dynamic topologies, humans oscillate between innovating individually or in small clusters, and then sharing outcomes with others. To our knowledge, the role of multi-agent topology on innovation has not been systematically studied in machine learning. It remains unclear a) which communication topologies are optimal for which innovation tasks, and b) which properties of experience sharing improve multi-level innovation. Here we use a multi-level hierarchical problem setting (WordCraft), with three different innovation tasks. We systematically design networks of DQNs sharing experiences from their replay buffers in varying topologies (fully connected, small world, dynamic, ring). Comparing the level of innovation achieved by different experience-sharing topologies across different tasks shows that, first, consistent with human findings, experience sharing within a dynamic topology achieves the highest level of innovation across tasks. Second, experience sharing is not as helpful when there is a single clear path to innovation. Third, two metrics we propose, conformity and diversity of shared experience, can explain the success of different topologies on different tasks. These contributions can advance our understanding of optimal AI-AI, human-human, and human-AI collaborative networks, inspiring future tools for fostering collective innovation in large organizations.

International audience; Despite its rise as a prominent solution to the data inefficiency of today's machine learning models, self-supervised learning has yet to be studied from a purely multi-agent perspective. In this work, we propose that aligning internal subjective representations, which naturally arise in a multi-agent setup where agents receive partial observations of the same underlying environmental state, can lead to more data-efficient representations. We propose that multi-agent environments, where agents do not have access to the observations of others but can communicate within a limited range, guarantees a common context that can be leveraged in individual representation learning. The reason is that subjective observations necessarily refer to the same subset of the underlying environmental states and that communication about these states can freely offer a supervised signal. To highlight the importance of communication, we refer to our setting as \textit{socially supervised representation learning}. We present a minimal architecture comprised of a population of autoencoders, where we define loss functions, capturing different aspects of effective communication, and examine their effect on the learned representations. We show that our proposed architecture allows the emergence of aligned representations. The subjectivity introduced by presenting agents with distinct perspectives of the environment state contributes to learning abstract representations that outperform those learned by a single autoencoder and a population of autoencoders, presented with identical perspectives of the environment state. Altogether, our results demonstrate how communication from subjective perspectives can lead to the acquisition of more abstract representations in multi-agent systems, opening promising perspectives for future research at the intersection of representation learning and emergent communication.

International audience; Hierarchical Reinforcement Learning (HRL) offers potential benefits for solving long horizon tasks, generally unhandled by standard Reinforcement Learning (RL) techniques, by decomposing the problem and combining simple policies to achieve the goal. They are however still held back by the curse of dimensionality and the ambiguity of selected subtasks. We explore relevant approaches in HRL while highlighting the key challenges of goal representation, high-level planning and propose a research outline tackling them.

We are interested in interactive agents that learn to coordinate, namely, a $builder$ -- which performs actions but ignores the goal of the task, i.e. has no access to rewards -- and an $architect$ which guides the builder towards the goal of the task. We define and explore a formal setting where artificial agents are equipped with mechanisms that allow them to simultaneously learn a task while at the same time evolving a shared communication protocol. Ideally, such learning should only rely on high-level communication priors and be able to handle a large variety of tasks and meanings while deriving communication protocols that can be reused across tasks. We present the Architect-Builder Problem (ABP): an asymmetrical setting in which an architect must learn to guide a builder towards constructing a specific structure. The architect knows the target structure but cannot act in the environment and can only send arbitrary messages to the builder. The builder on the other hand can act in the environment, but receives no rewards nor has any knowledge about the task, and must learn to solve it relying only on the messages sent by the architect. Crucially, the meaning of messages is initially not defined nor shared between the agents but must be negotiated throughout learning. Under these constraints, we propose Architect-Builder Iterated Guiding (ABIG), a solution to ABP where the architect leverages a learned model of the builder to guide it while the builder uses self-imitation learning to reinforce its guided behavior. We analyze the key learning mechanisms of ABIG and test it in 2D tasks involving grasping cubes, placing them at a given location, or building various shapes. ABIG results in a low-level, high-frequency, guiding communication protocol that not only enables an architect-builder pair to solve the task at hand, but that can also generalize to unseen tasks., Comment: International Conference on Learning Representations (2022)

A long-standing goal of Machine Learning (ML) and AI at large is to design autonomous agents able to efficiently interact with our world. Towards this, taking inspirations from the interactive nature of human and animal learning, several lines of works focused on building decision making agents embodied in real or virtual environments. In less than a decade, Deep Reinforcement Learning (DRL) established itself as one of the most powerful set of techniques to train such autonomous agents. DRL is based on the maximization of expert-defined reward functions that guide an agent’s learning towards a predefined target task or task set. In parallel, the Developmental Robotics field has been working on modelling cognitive development theories and integrating them into real or simulated robots. A core concept developed in this literature is the notion of intrinsic motivation: developmental robots explore and interact with their environment according to self-selected objectives in an open-ended learning fashion. Recently, similar ideas of self-motivation and open-ended learning started to grow within the DRL community, while the Developmental Robotics community started to consider DRL methods into their developmental systems. We propose to refer to this convergence of works as Developmental Machine Learning. Developmental ML regroups works on building embodied autonomous agents equipped with intrinsic-motivation mechanisms shaping open-ended learning trajectories. The present research aims to contribute within this emerging field. More specifically, the present research focuses on proposing and assessing the performance of a core algorithmic block of such developmental machine learners: Automatic Curriculum Learning (ACL) methods. ACL algorithms shape the learning trajectories of agents by challenging them with tasks adapted to their capacities. In recent years, they have been used to improve sample efficiency and asymptotic performance, to organize exploration, to encourage generalization or to solve sparse reward problems, among others. Despite impressive success in traditional supervised learning scenarios (e.g. image classification), large-scale and real-world applications of embodied machine learners are yet to come. The present research aims to contribute towards the creation of such agents by studying how to autonomously and efficiently scaffold them up to proficiency.; Un objectif de longue date du Machine Learning (ML) et de l'IA en général est de concevoir des agents autonomes capables d'interagir efficacement avec notre monde. Dans cette optique, s'inspirant de la nature interactive de l'apprentissage humain et animal, plusieurs axes de travaux se sont concentrés sur la construction d'agents décisionnels incarnés dans des environnements réels ou virtuels. En moins d'une décennie, le Deep Reinforcement Learning (DRL) s'est imposé comme l'un des ensembles de techniques les plus puissants pour former de tels agents autonomes. Le DRL est basé sur la maximisation de fonctions de récompense définies par des experts qui guident l'apprentissage d'un agent vers une tâche ou un ensemble de tâches cible prédéfinies. En parallèle, la Robotique Développementale a travaillé sur la modélisation des théories du développement cognitif et de leur intégration dans des robots réels ou simulés. Un concept central développé dans cette littérature est la notion de motivation intrinsèque : les robots développementaux explorent et interagissent avec leur environnement selon des objectifs auto-sélectionnés dans un mode d'apprentissage ouvert. Récemment, des idées similaires d'auto-motivation et d'apprentissage ouvert ont commencé à se développer au sein de la communauté DRL, tandis que la Robotique Développementale a commencé à considérer les méthodes DRL dans leurs systèmes de développement. Nous proposons de désigner cette convergence de travaux sous le nom de Developmental Machine Learning. Developmental ML regroupe des travaux sur la construction d'agents autonomes incarnés équipés de mécanismes de motivation intrinsèque façonnant des trajectoires d'apprentissage ouvertes. La présente recherche vise à contribuer dans ce domaine émergent. Plus précisément, la présente recherche se concentre sur la proposition et l'évaluation des performances d'un bloc algorithmique de base de tels apprenants: les méthodes d'apprentissage automatique de curriculum (ACL). Les algorithmes ACL façonnent les trajectoires d'apprentissage des agents en les challengeant avec des tâches adaptées à leurs compétences. Ces dernières années, ils ont été utilisés pour améliorer la vitesse d’apprentissage et les performances asymptotiques, pour organiser l'exploration, pour encourager la généralisation ou pour résoudre des problèmes de récompense clairsemée, entre autres. Malgré un succès impressionnant dans les scénarios d'apprentissage supervisé traditionnels (par exemple, la classification d'images), les applications à grande échelle et dans le monde réel des apprenants automatiques incarnés sont encore à venir. La présente recherche vise à contribuer à la création de tels agents en étudiant comment guider leurs apprentissage de manière autonome.

Humans show language-biased image recognition for a word-embedded image, known as picture-word interference. Such interference depends on hierarchical semantic categories and reflects that human language processing highly interacts with visual processing. Similar to humans, recent artificial models jointly trained on texts and images, e.g., OpenAI CLIP, show language-biased image classification. Exploring whether the bias leads to interference similar to those observed in humans can contribute to understanding how much the model acquires hierarchical semantic representations from joint learning of language and vision. The present study introduces methodological tools from the cognitive science literature to assess the biases of artificial models. Specifically, we introduce a benchmark task to test whether words superimposed on images can distort the image classification across different category levels and, if it can, whether the perturbation is due to the shared semantic representation between language and vision. Our dataset is a set of word-embedded images and consists of a mixture of natural image datasets and hierarchical word labels with superordinate/basic category levels. Using this benchmark test, we evaluate the CLIP model. We show that presenting words distorts the image classification by the model across different category levels, but the effect does not depend on the semantic relationship between images and embedded words. This suggests that the semantic word representation in the CLIP visual processing is not shared with the image representation, although the word representation strongly dominates for word-embedded images., Accepted at ICLR 2022

International audience; Few technology-based interventions have addressed mathematical and numeracy skills of individuals with autism spectrum disorder (ASD). However, children and adolescents with ASD may experience difficulties in mathematical learning, and even be behind their peers at school. Intelligent Tutoring System (ITS) is an innovative way for enhancing teaching programs through learning optimization algorithms coping interindividual differences in the educational curriculum progress. They have been used with success with typically-developed students for levering on the linkage learning-motivation, but remains at a theoretical stage for students with ASD and/or intellectual disability (ID). We conducted a pilot study to assess an ITS-based intervention embedded into the KidLearn application for instructing mathematical skills (i.e., adding and subtracting numbers between 0 and 100). Twenty-four secondary school students with ASD and/or intellectual disability, enrolled from special education classrooms, have been recruited: 14 students received the KidLearn intervention, while the 10 others received control intervention. Pre-post assessments showed that calculation and motivational measures were significantly improved in KidLearn condition compared to the control condition. ITS intervention appeared relevant for calculation learning in the setting of special education.

Background The SoBeezy program is an innovative intervention aimed at promoting and fostering healthy aging and aging in place by proposing to older adults concrete solutions to face daily life, tackle loneliness, promote social participation, and reduce the digital divide, thanks to a specific, easy-to-use voice assistant (the BeeVA smart display). Objective This study aims to assess the acceptability of the SoBeezy program and its voice assistant and to identify potential areas of improvement. Methods A 12-month experimentation of the program was deployed in real-life conditions among older adults living in the community in 4 pilot cities of France. Launched during the first lockdown of the COVID-19 crisis, this multisite study aimed to assess acceptability using questionnaires and interviews conducted at baseline and at the end of the experimentation. In addition, a series of meetings were conducted with SoBeezy staff members to obtain direct feedback from the ground. Results In total, 109 older individuals were equipped with BeeVA to use the SoBeezy program; of these, 32 (29.4%) left the experimentation before its end and 69 (63.3%) completed the final questionnaires. In total, 335 interventions were conducted and 27 (39%) of the participants requested services, mainly for supportive calls and visits and assistance with shopping, transportation, and crafting-gardening. Of the whole sample, 52 (75%) considered BeeVA as a reassuring presence, and few persons (15/69, 22%) reported a negative opinion about the program. Among the participants, the voice assistant appeared easy to use (n=57, 82%) and useful (n=53, 77%). They also were positive about the BeeVA smart display and the SoBeezy intervention. Conclusions This multisite study conducted in real-life conditions among more than 100 older adults living in the community provides enlightening results of the reality from the ground of digital tools designed for the aging population. The COVID-19 context appeared both as an opportunity, given the massive needs of the older adults during this crisis, and as limiting due to sanitary constraints. Nevertheless, the experimentation showed overall good acceptability of the voice assistant and a high level of satisfaction of the participants among those who really used the system and could be a way of improving the autonomy and well-being of older adults and their families. However, the findings also highlighted resistance to change and difficulties for the users to ask for help. The experimentation also emphasized levers for next deployments and future research. The next step will be the experimentation of the activity-sharing component that could not be tested due to the COVID-19 context.

International audience; Building autonomous agents able to grow open-ended repertoires of skills across their lives is a fundamental goal of artificial intelligence (AI). A promising developmental approach recommends the design of intrinsically motivated agents that learn new skills by generating and pursuing their own goals - autotelic agents. But despite recent progress, existing algorithms still show serious limitations in terms of goal diversity, exploration, generalisation or skill composition. This perspective calls for the immersion of autotelic agents into rich socio-cultural worlds, an immensely important attribute of our environment that shapes human cognition but is mostly omitted in modern AI. Inspired by the seminal work of Vygotsky, we propose Vygotskian autotelic agents - agents able to internalise their interactions with others and turn them into cognitive tools. We focus on language and show how its structure and informational content may support the development of new cognitive functions in artificial agents as it does in humans. We justify the approach by uncovering several examples of new artificial cognitive functions emerging from interactions between language and embodiment in recent works at the intersection of deep reinforcement learning and natural language processing. Looking forward, we highlight future opportunities and challenges for Vygotskian Autotelic AI research, including the use of language models as cultural models supporting artificial cognitive development.

School inclusion of children with Autism Spectrum Disorder (ASD) represents a challenge forall parties involved (family, school and out-of-school actors such as medico-socialprofessionals). To the best of our knowledge, existing tools focus either on the pupil, or therelationships between two parties, always including parents. Considering a systemic approachand knowing that collaboration among stakeholders of school inclusion allows the latter tosucceed, we decided to create a tool to enhance and facilitate communication and informationsharing among them. For a better usability and to ensure satisfaction of the need expressed bythose stakeholders, we used participatory design to create the web-based app ToGather. Weused (fictitious) realistic cases to create scenarios that contain tasks to do on the webapplication, adapted to each participant role (parent, teacher and external actors). We usedobjective (interaction annotations, website navigation inspection) and subjective(questionnaires) measures to assess the design of ToGather, with regards to usability, userexperience, cognitive load and elicited self-determination. Results indicated that ToGather wasvery usable, with an excellent user experience, and a relatively low cognitive demand. Somedesign mistakes have been identified and corrected for the enhanced final release.

Intrinsic motivation – the desire to do things for their inherent joy and pleasure – has received its first share of scientific attention over 70years ago, ever since we saw monkeys solving puzzles for free. Since then, research on intrinsic motivation has been steadily gaining momentum. We have come to understand, in the context of learning and discovery, that intrinsic motivation (namely, intrinsically motivated information-seeking) is foundational for the biological and technological success of our species. But where does intrinsic motivation to learn and seek information come from? Today, with the thriving synergy between perpetually advancing fields of psychology, neuroscience, and computer science, we are well positioned to investigate this question. The Learning Progress Hypothesis (LPH) proposes that humans are motivated by feelings of and/or beliefs about progress in knowledge (including progress in competence). In artificial learners, progressbased intrinsic motivation enables autonomous exploration of the environment (including the agent’s own body), resulting in better performance, more efficient learning, and richer skill sets. Due to similar computational challenges facing artificial and biological learners, researchers have proposed that progress-based intrinsic motivation might have evolved in humans to help us transition from babies with few skills and little knowledge to knowledgeable grownups capable of performing many sophisticated tasks. The Learning Progress Hypothesis (LPH) is attractive, not only because it is consistent with several studies of human curiosity, but also because it resonates with existing theories on metacognitive self-regulation in learning. However, the LPH has not been extensively studied using behavioral experimentation.This thesis provides an empirical examination of the LPH. We introduce a novel experimental paradigm where participants explore multiple learning activities, some easy, others difficult. The activities involve guessing the binary category of randomly presented stimuli. To let their intrinsic motivation shine, we did not provide any material incentives encouraging specific behaviors or strategies – we simply observed which activities people engaged in and how their knowledge about these activities unfolded over time. We present statistical analyses and a computational model that support the LPH.This thesis also suggests ideas for future investigations into progressbased motivation. These ideas are inspired by a pilot study in which we asked participants to practice a naturalistic sensorimotor skill (a video game) over the course of 3 sessions spanning 5 days. [...]We conclude the thesis with an extended discussion of our findings, where we examine some limitations of our experiments and propose promising future steps. In summary, we believe the behavioral paradigms introduced in this thesis should be reused to not only replicate our results, but also to advance the scientific research of intrinsically motivated information-seeking., La motivation intrinsèque - le désir de faire les choses pour la joie et le plaisir inhérent qu’ils procurent - a suscité l’intérêt des chercheurs pour la première fois il y a plus de 70 ans, en voyant des singes résoudre des énigmes gratuitement. Depuis, la recherche sur la motivation intrinsèque n’a cessé de prendre de l’ampleur. Dans le contexte de l’apprentissage et des comportements de découverte, il est aujourd’hui entendu , que la motivation intrinsèque (c’est-à-dire la recherche intrinsèquement motivée d’informations) est fondamentale au progrès biologique et technologique de notre espèce. Mais d’où vient la motivation intrinsèque pour apprendre et pour rechercher des informations ? Aujourd’hui, avec la synergie florissante entre les domaines en constante évolution de la psychologie, des neurosciences et de l’informatique, nous sommes en capacité d’étudier cette question.L’hypothèse de progrès d’apprentissage (Learning Progress Hypothesis,ou LPH en anglais) propose que les humains sont motivés par leurs sentiments et/ou croyances relatifs à leurs progrès de connaissances (y compris les progrès dans les compétences). Chez les apprenants artificiels, la motivation intrinsèque basée sur le progrès d’apprentissage permet une exploration autonome de l’environnement (y compris le propre corps de l’agent), ce qui se traduit par de meilleures performances, un apprentissage plus efficace et un ensemble de compétences plus riches. En raison des défis informatiques similaires auxquels sont confrontés les apprenants artificiels et biologiques, des chercheurs ont proposé que la motivation intrinsèque basée sur le progrès ait pu évoluer chez l’Homme, pour nous aider à passer de bébés avec peu de compétences et peu de connaissances, à des adultes bien informés capables d’effectuer de nombreuses tâches sophistiquées. La LPH est attrayante, non seulement parce qu’elle est cohérente avec plusieurs études sur la curiosité humaine, mais aussi parce qu’elle résonne avec les théories existantes sur l’autorégulation métacognitive dans l’apprentissage. Cependant, la LPH n’a pas été largement étayée par l’expérimentation comportementale. Cette thèse propose un examen empirique de la LPH. Nous introduisons un nouveau paradigme expérimental où les participants explorent plusieurs activités d’apprentissage, certaines faciles, d’autres difficiles. Les activités consistent à deviner la catégorie binaire de stimuli présentés au hasard. Pour laisser briller leur motivation intrinsèque, nous n’avons fourni aucune incitation matérielle encourageant des comportements ou des stratégies spécifiques - nous avons simplement observé dans quelles activités les personnes s’engageaient et comment leurs connaissances sur ces activités se développaient au fil du temps. Nous présentons des analyses statistiques et un modèle de calcul prenant en charge la LPH.[...]. Nous concluons la thèse par une discussion approfondie de nos résultats, où nous examinons certaines limites de nos expériences et proposons des étapes futures prometteuses. En résumé, nous pensons que les paradigmes comportementaux introduits dans cette thèse devraient être réutilisés non seulement pour reproduire nos résultats,mais aussi pour faire avancer la recherche scientifique sur la recherche d’informations intrinsèquement motivée.

International audience; Intrinsically motivated information-seeking, also called curiositydriven exploration, is widely believed to be a key ingredient for autonomous learning in the real world. Such forms of spontaneous exploration have been studied in multiple independent lines of computational research, producing a diverse range of algorithmic models that capture different aspects of these processes. These algorithms resolve some of the limitations of neurocognitive theories by formally describing computational functions and algorithmic implementations of intrinsically motivated learning. Moreover, they reveal a high diversity of effective forms of intrinsically motivated information-seeking that can be characterized along different mechanistic and functional dimensions. This chapter aims at reviewing different classes of algorithms and highlighting several important dimensions of variation among them. Identifying these dimensions provides means for structuring a comprehensive taxonomy of approaches. We believe this exercise to be useful in working towards a general computational account of information-seeking. Such an account should facilitate the proposition of new hypotheses about informationseeking in humans and complement the existing psychological theory of curiosity.

International audience

International audience; There is an exponential increase in the range of digital products and devices promoting aging in place in particular devices aiming at preventing or detecting falls. However, their deployment is still limited and few studies have been carried out in population-based settings. Such a matter of fact is due to the technological challenges that remain to overcome but also to the barriers that are specific to the users themselves such as the generational digital divide and acceptability factors specific to the elderly population. To date, scarce studies take into account these factors. In order to capitalize technological progress, the further step should be to better take into account these factors and to deploy, in a broader and more ecological way, these technologies designed for home care seniors, in order to assess their effectiveness in real life.

Submitted paper; International audience; Recent advances in Artificial Intelligence (AI) have revived the quest for agents able to acquire an open-ended repertoire of skills. Although this ability is fundamentally related to the characteristics of human intelligence, research in this field rarely considers the processes and ecological conditions that may have guided the emergence of complex cognitive capacities during the evolution of the species. Research in Human Behavioral Ecology (HBE) seeks to understand how the behaviors characterizing human nature can be conceived as adaptive responses to major changes in our ecological niche. In this paper, we propose a framework highlighting the role of environmental complexity in open-ended skill acquisition, grounded in major hypotheses from HBE and recent contributions in Reinforcement learning (RL). We use this framework to highlight fundamental links between the two disciplines, as well as to identify feedback loops that bootstrap ecological complexity and create promising research directions for AI researchers. We also present our first steps towards designing a simulation environment that implements the climate dynamics necessary for studying key HBE hypotheses relating environmental complexity to skill acquisition.

In recent years, many psycho-educational technologies were studied to address the school-related difficulties encountered by students with autism spectrum disorder (ASD). However, most of them remain individual-centered and do not consider the social environment. To fill this gap, this study reports on the user-centered design of a web-based support tool, which aims to support communication and coordination between parents, school staff, and health professionals of middle and high school students with ASD, in the context of elaborating, implementing, and following an Individualized Education Plan (IEP). The study followed a two-step design process: 1) a need analysis for identifying information domains deemed important by the stakeholders; 2) through a participative iterative design process, a panel of professionals and parents provided ideas and feedback on the design, which was integrated into subsequent prototype versions of the "ToGather" app., Ces dernières années, de nombreuses technologies psycho-éducatives ont été étudiées pour répondre aux difficultés scolaires rencontrées par les élèves avec trouble du spectre autistique (TSA). Cependant, la plupart d'entre elles restent centrées sur l'individu et ne prennent pas en compte l'environnement social. Pour combler cette lacune, cette étude porte sur la conception centrée-utilisateur d'un outil en ligne, qui vise à soutenir la communication et la coordination entre les parents, le personnel scolaire et les professionnels médico-sociaux des élèves de collège et de lycée atteints de TSA, dans le contexte de l'élaboration, de la mise en œuvre et du suivi d'un projet personnalisé de scolarisation (PPS). L'étude a suivi un processus de conception en deux étapes : 1) une analyse des besoins pour identifier les domaines d'information jugés importants par les parties prenantes ; 2) par le biais d'un processus de conception participatif et itératif, un panel de professionnels et de parents a fourni des idées et des commentaires sur la conception, qui ont été intégrés dans les versions prototypes ultérieures de l'application "ToGather".

There is an exponential increase in the range of digital products and devices promoting aging in place in particular devices aiming at preventing or detecting falls. However, their deployment is still limited and few studies have been carried out in population-based settings. Such a matter of fact is due to the technological challenges that remain to overcome but also to the barriers that are specific to the users themselves such as the generational digital divide and acceptability factors specific to the elderly population. To date, scarce studies take into account these factors. In order to capitalize technological progress, the further step should be to better take into account these factors and to deploy, in a broader and more ecological way, these technologies designed for home care seniors, in order to assess their effectiveness in real life.

Building embodied autonomous agents capable of participating in social interactions with humans is one of the main challenges in AI. Within the Deep Reinforcement Learning (DRL) field, this objective motivated multiple works on embodied language use. However, current approaches focus on language as a communication tool in very simplified and non-diverse social situations: the "naturalness" of language is reduced to the concept of high vocabulary size and variability. In this paper, we argue that aiming towards human-level AI requires a broader set of key social skills: 1) language use in complex and variable social contexts; 2) beyond language, complex embodied communication in multimodal settings within constantly evolving social worlds. We explain how concepts from cognitive sciences could help AI to draw a roadmap towards human-like intelligence, with a focus on its social dimensions. As a first step, we propose to expand current research to a broader set of core social skills. To do this, we present SocialAI, a benchmark to assess the acquisition of social skills of DRL agents using multiple grid-world environments featuring other (scripted) social agents. We then study the limits of a recent SOTA DRL approach when tested on SocialAI and discuss important next steps towards proficient social agents. Videos and code are available at https://sites.google.com/view/socialai., under review. This paper extends and generalizes work in arXiv:2104.13207

Autonomous discovery and direct instruction are two extreme sources of learning in children, but educational sciences have shown that intermediate approaches such as assisted discovery or guided play resulted in better acquisition of skills. When turning to Artificial Intelligence, the above dichotomy can be translated into the distinction between autonomous agents, which learn in isolation from their own signals, and interactive learning agents which can be taught by social partners but generally lack autonomy. In between should stand teachable autonomous agents: agents that learn from both internal and teaching signals to benefit from the higher efficiency of assisted discovery processes. Designing such agents could result in progress in two ways. First, very concretely, it would offer a way to non-expert users in the real world to drive the learning behavior of agents towards their expectations. Second, more fundamentally, it might be a key step to endow agents with the necessary capabilities to reach general intelligence. The purpose of this paper is to elucidate the key obstacles standing in the way towards the design of such agents. We proceed in four steps. First, we build on a seminal work of Bruner to extract relevant features of the assisted discovery processes happening between a child and a tutor. Second, we highlight how current research on intrinsically motivated agents is paving the way towards teachable and autonomous agents. In particular, we focus on autotelic agents, i.e. agents equipped with forms of intrinsic motivations that enable them to represent, self-generate and pursue their own goals. We argue that such autotelic capabilities from the learner side are key in the discovery process. Third, we adopt a social learning perspective on the interaction between a tutor and a learner to highlight some components that are currently missing to these agents before they can be taught by ordinary people using natural pedagogy. Finally, we provide a list of specific research questions that emerge from the perspective of extending these agents with assisted learning capabilities.

With advances in artificial intelligence, research is increasingly exploring the potential functions that social robots can play in education. As teachers are a critical stakeholder in the use and application of educational technologies, we conducted a study to understand teachers' perspectives on how a social robot could support a variety of learning activities in the classroom. Through interviews, robot puppeteering, and group brainstorming sessions with five elementary and middle school teachers from a local school in Canada, we take a socio-technical perspective to conceptualize possible robot functions and behaviours, and the effects they may have on the current way learning activities are designed, planned, and executed. Overall, the teachers responded positively to the idea of introducing a social robot as a technological tool for learning activities, envisioning differences in usage for teacher-robot and student-robot interactions. Further, Engestr\"om's Activity System Model -- a framework for analyzing human needs, tasks, and outcomes -- illustrated a number of tensions associated with learning activities in the classroom. We discuss the fine-grained robot functions and behaviours conceived by teachers, and how they address the current tensions -- providing suggestions for improving the design of social robots for learning activities., Comment: This is a preprint of an article published in The International Journal of Social Robotics. The final authenticated version is available online at: https://doi.org/10.1007/s12369-021-00820-7