Recent Large Language Models (LLMs) demonstrate problem-solving capabilities suitable for educational use. This paper investigates using LLMs to create virtual agents that mimic student behavior and interact with learning platforms. Testing a modest-sized LLM on an Intelligent Tutoring System for word problem-solving revealed that the LLM could fully solve 92% of single-step problems, although their performance decreased to 14% when attempting more complex problems.

Recent advances in generative AI epitomized by Large Language Models (LLMs) demonstrated remarkable capabilities in generating human-like text and understanding contextual nuances. In parallel, advances in domestic applications of AI continue to break new grounds bringing to the forefront the need to address the impact of such systems especially on users' well being through socially aware interactive intelligent systems. This paper takes one step forward in this endeavour to examine the potential of LLMs based systems to generate empathy in its responses. In doing so, we identified through initial tests the key factors to designing empathy triggering responses. We then designed a set of experiments to test systematically the impact each of these factors would have on triggering empathy. We analysed the results against the degree of expressed empathy and its perceived quality.

Abstract Motivation to engage in learning is essential for learning performance. Learners’ motivation is traditionally assessed using self-reported data, which is time-consuming, subjective, and interruptive to their learning process. To address this issue, this paper proposes a novel framework for multimodal assessment of learners’ motivation in e-learning environments with the ultimate purpose of supporting intelligent e-learning systems to facilitate dynamic, context-aware, and personalized services or interventions, thus sustaining learners’ motivation for learning engagement. We investigated the performance of the machine learning classifier and the most and least accurately predicted motivational factors. We also assessed the contribution of different electroencephalogram (EEG) and eye gaze features to motivation assessment. The applicability of the framework was evaluated in an empirical study in which we combined eye tracking and EEG sensors to produce a multimodal dataset. The dataset was then processed and used to develop a machine learning classifier for motivation assessment by predicting the levels of a range of motivational factors, which represented the multiple dimensions of motivation. We also proposed a novel approach to feature selection combining data-driven and knowledge-driven methods to train the machine learning classifier for motivation assessment, which has been proved effective in our empirical study at selecting predictors from a large number of extracted features from EEG and eye tracking data. Our study has revealed valuable insights for the role played by brain activities and eye movements on predicting the levels of different motivational factors. Initial results using logistic regression classifier have achieved significant predictive power for all the motivational factors studied, with accuracy of between 68.1% and 92.8%. The present work has demonstrated the applicability of the proposed framework for multimodal motivation assessment which will inspire future research towards motivationally intelligent e-learning systems.

In the past 50 years, natural disasters worldwide have accounted for 2.06 million deaths and US$3640 billion in economic losses. These natural disasters are heavily influenced by the composite earth system processes and human interactions. In this paper, we focus our investigation to assess the impact of flooding in rivers and coastal regions and its impact on low-income communities. For this purpose, a transdisciplinary perspective from Artificial Intelligence (AI), Climate science, Socio-economics discipline is leveraged to map and identify their inter-relationships and challenges using Soft System Methodology (SSM). A transdisciplinary framework, named ADRELO1 Disaster Support System (ADSS), is therefore proposed to (1) identify the key parameters that can influence climate change, (2) stitch together a reusable multilayered transdisciplinary knowledge model, and (3) apply the observed multivariant data to AI-based algorithm to forecast climate change, analyze the impact of climate change on socio-economic outcomes and suggest potential disaster risk reduction actions. Research-based outcomes, from the given framework, will be used for policy prescription towards making flood-affected local communities self-resilient. ADSS will be applied first in a flood-prone region, such as Nyando in Kenya and Mozambique. It will then be extrapolated in other coastal regions of Florida and North-eastern Brazil to examine the applicability of the framework.

Artificial intelligence (AI) enabled products and services are becoming a staple of everyday life. While governments and businesses are eager to enjoy the benefits of AI innovations, the mixed impact of these autonomous and intelligent systems on human well-being has become a pressing issue. The purpose of this article is to review one of the first international standards focused on the social and ethical implications of AI: The Institute of Electrical and Electronics Engineering's (IEEE) Standard (Std) 7010-2020 Recommended Practice for Assessing the Impact of Autonomous and Intelligent Systems on Human Well-being. Incorporating well-being factors throughout the lifecycle of AI is both challenging and urgent and IEEE 7010 aims to provide guidance for those who design, deploy, and procure these technologies. Before introducing IEEE 7010, we consider possible benefits of an approach for AI centered around well-being and the measurement of well-being data. Next, we critically examine how the standard relates to approaches and perspectives in place in the AI community. Finally, we indicate where future efforts are needed for IEEE 7010 to better achieve its ambitions.

This research aims to identify a feasible model to predict a learner’s stress in an online learning platform. It is desirable to produce a cost-effective, unobtrusive and objective method to measure a learner’s emotions. The few signals produced by mouse and keyboard could enable such solution to measure real-world individual’s affective states. It is also important to ensure that the measurement can be applied regardless of the type of task carried out by the user. This preliminary research proposes a stress classification method using mouse and keystroke dynamics to classify the stress levels of 190 university students when performing three different e-learning activities. The results show that the stress measurement based on mouse and keystroke dynamics is consistent with the stress measurement according to the changes in duration spent between two consecutive questions. The feedforward back-propagation neural network achieves the best performance in the classification.

Most previous work in emotion recognition has fixed the available classes in advance, and attempted to classify samples into one of these classes using a supervised learning approach. In this paper, we present preliminary work on combining supervised and unsupervised learning to discover potential latent classes which were not initially considered. To illustrate the potential of this hybrid approach, we have used a Self-Organizing Map (SOM) to organize a large number of Electroencephalogram (EEG) signals from subjects watching videos, according to their internal structure. Results suggest that a more useful labelling scheme could be produced by analysing the resulting topology in relation to user reported valence levels (i.e., pleasantness) for each signal, refining the original set of target classes.

In location-based services (LBSs), the service is provided based on the users' locations through location determination and mobility realization. Several location prediction models have been proposed to enhance and increase the relevance of the information retrieved by users of mobile information systems, but none of them studied the relationship between accuracy rate of prediction and the performance of the model in terms of consuming resources and constraints of mobile devices. Most of the current location prediction research is focused on generalized location models, where the geographic extent is divided into regular-shape cells. These models are not suitable for certain LBSs where the objectives are to compute and present on-road services. One such technique is the Prediction Location Model (PLM), which deals with inner cell structure. The PLM technique suffers from memory usage and poor accuracy. The main goal of this paper is to propose a new path prediction technique for Location-Based Services. The new approach is competitive and more efficient compared to PLM regarding measurements such as accuracy rate of location prediction and memory usage.

Arabic is widely spoken language but very few text mining tools have been developed for Arabic language. This paper presents an algorithm for automatically building dictionaries. The target domain is crime profiling. The corpus is mined for crime type, location and nationality. This work is then validated through three experiments, the results of which show that the algorithm developed here is promising.

This paper presents the theoretical account for designing and developing an algorithm for network security inspired by the behavior of Beavers. The algorithm uses the beaver behavioral patterns in constructing dams and water tunnels to create analogous secure tunnels and information lakes. The approach is a user-centric and the paper demonstrates the use of the algorithm in security and route optimization with the assumption that the beaver agent is deployed on a mobile device (e.g. smart mobile phone). An algorithmic approach to design the beaver agent and swarm is followed here. The set of algorithms presented is complemented by critical review outlining the further work needed.

Accurate estimation of the software effort and schedule affects the budget computation. Bidding for contracts depends mainly on the estimated cost. Inaccurate estimates will lead to failure of making a profit, increased probability of project incompletion and delay of the project delivery date. In this paper, we explore the use of Soft Computing Techniques to build a suitable model structure to utilize improved estimations of software effort for NASA software projects. In doing so, we plan to use Particle Swarm Optimization (PSO) to tune the parameters of the famous COnstructive COst MOdel (COCOMO). We plan also to explore the advantages of Fuzzy Logic to build a set of linear models over the domain of possible software Line Of Code (LOC). The performance of the developed model was evaluated using NASA software projects data set [1]. A comparison between COCOMO tuned-PSO, Fuzzy Logic (FL), Halstead, Walston-Felix, Bailey-Basili and Doty models were provided.

This paper discusses the application of artificial neural networks (ANNs) in the area of identification and control of nonlinear dynamical systems. Since chemical processes are getting more complex and complicated, the need of schemes that can improve process operations is highly demanded. ANNs are capable of learning from examples, perform non-linear mappings, and have a special capacity to approximate the dynamics of nonlinear systems in many applications. This paper describe the application of neural network for modeling reactor level, reactor pressure, reactor cooling water temperature, and reactor temperature problems in the Tennessee Eastman (TE) chemical process reactor. The potential of neural network technology in the process industries is great. Its ability to model process dynamics makes it powerful tool for modeling and control processes. A comparison between the applications of ANNs to model the TE plant is compared with other soft computing techniques like fuzzy logic (FL) and adaptive neuro-fuzzy inference systems (ANFIS).

This paper provides preliminary results of fuzzy individual model (FIM) being developed for realistic modelling of crowd behaviour under normal and abnormal conditions. It explores the impact of personality, perception and emotions on the behaviour. Fuzzy logic is used to deal with the vagueness aspects of perceptions and the uncertainty of emotions. The fuzzy rules link these parts together feeding into behavioural rules. The implied behaviours are observed in simulation and evaluated.

Sound is perhaps the most elementary and yet common communication vehicle used by humans and animals alike. Similarly, reactive robots that are based on animal intelligence require real-time simple communication mechanisms to imitate their animal examples. However, emotions are not often considered in developing reactive robots interaction rules and communication. Therefore, these robots, unlike their natural examples, are not capable of expressing basic emotions such as happiness and excitement, fear and stress, anger or contentment as the case with many animals