This paper introduces a systematic framework for the identification of memory statements in human-AI interactions. We present a novel benchmark dataset of 4000 labeled statements spanning five types of memory (episodic, semantic, spatial, emotional, associative) and a non-memory category, generated through a controlled GPT-4o and Faker-based pipeline with thematic randomization. The dataset was balanced across memory categories to ensure equitable class representation, while domain contexts (e.g., medical, educational, personal) were grouped by span identifiers to maintain contextual consistency without affecting label stratification. A representative, categorybalanced subset was reviewed and validated by board-certified psychiatrists to confirm semantic and cognitive plausibility, enhancing dataset reliability. On this corpus, we fine-tuned a RoBERTa-based transformer backbone, resulting in a specialized model we term MemoryBERT, which achieved perfect accuracy across all classes. Unlike prior studies that presuppose reliable memory categorization, our contribution establishes the first dedicated backbone for systematic memory identification in AI systems. As this work represents a feasibility study and preliminary stage of a broader research effort, future work will extend expert validation, expand the dataset with humanauthored and adversarial samples, and benchmark additional transformer architectures on the same classification task.

The increasing integration of renewable sources introduces volatility and uncertainty into energy systems, particularly under extremely low production and/or extremely high consumption events. To address this, we present BASTION.ev (Bayesian Simulation & Forecasting for Extreme Events), a modular and scalable framework for intelligent energy management, which combines Extreme Value Analysis, Machine Learningbased forecasting, and simulation-driven optimization. We opt for grid search instead of Bayesian Optimization to simulate and optimize control strategies due to its lower implementation complexity and transparent evaluation across parameter combinations. Specifically, our proposed framework ingests time-series energy data to engineer temporal and contextual features to support predictive modeling. Extreme Value Analysis is used to define critical thresholds using generalized Pareto distributions, allowing the classification of operational states into extreme low, normal, and extreme high imbalance scenarios. These states are forecasted using tree-based multi-class classifiers across multiple time horizons (e.g., 1 to 72 hours ahead). The simulation of battery behavior is then evaluated under forecasted imbalance conditions, with control strategies (charge/discharge rates) optimized through a Bayesian Optimization to finetune parameters in order to maximize energy efficiency and, subsequently, economic return. Although based on mock-up data, simulations reveal that high feed-in profits are achieved with aggressive charging and minimal discharging. As a use case in this work, we focused our simulation on a 72-hour ahead forecasting window, where, despite class imbalance, the model achieved a macro F1-score of 0.63 and an overall accuracy of 91 %. The predicted imbalance states drive a simulation engine that evaluates energy system responses across control strategies. To optimize performance, we employ grid search as a baseline, over manually predefined charge/discharge configurations. Using mock-up data, the top configuration (charge$=15 \text{kWh}$, discharge$=1 \text{kWh}$) yielded a simulated profit of € 420.3, with only 66 kWh imported and 4335 kWh fed into the grid. Then, we use Bayesian Optimization for parameter finetuning and results improve greatly, Yielding € 584.00 simulated profit with 0 kWh imported energy for a fully self-sufficient PV production and 5840 kWh surplus fed into the grid with a charge rate of 20 kWh and a discharge rate of 0.1 kWh.

This paper introduces Energy-VERITIES, a framework for modeling epistemic risk and trust towards the integration of generative AI and large language models (LLMs) within the energy digital spine ecosystem. It defines epistemic risk zones using a set-theoretic Venn model over AI outputs ($\boldsymbol{A}$), expert ($E$) and non-expert ($N$) beliefs, internet sources ($D \_I$), and verified domain facts ($D$), with a critical risk zone formalized as (D_I$\cap A \cap E \cap N$) - D, representing hallucinated outputs that appear credible due to overlapping AI, expert, and stakeholder beliefs but diverge from verified domain facts. To assess energy stakeholderAI interaction, we propose a set of Trust Performance Indicators (TPIs) that quantify epistemic outcomes from the user's perspective, extending the confusion matrix and the previously devised VIRTSI model, which characterizes dynamic human trust states through a finite state automaton informed by user validation behavior. These include, for example, MJO-HRAR (Misconceptually Justified Overtrust), where users accept false outputs after flawed validation, and BOV-HRAR (Blind Overtrust), where no validation occurs. An empirical study of ChatGPT's energy advice in Greece illustrates the framework's diagnostic value, revealing that 39 % of validated responses were still false, over 30 % of hallucinations were accepted without validation, and correct answers were often dismissed. These findings indicate the need for improvements in interface-level explainability, validation support, and targeted AI literacy for energy stakeholders. The Energy-VERITIES framework offers a foundation for identifying epistemic risks and guiding the trustworthy integration of generative AI and LLMs in high-stakes domains, like energy.

We present COGNET-MD-X-REQ: a COGnitive NETwork Evaluation Toolkit that identifies precise fine-tuning needs for Large Language Models (LLMs) in multimodal medical diagnosis. COGNET-MD-X-REQ is a novel, two-step evaluation framework, inspired by Objective Structured Clinical Examinations (OSCEs), that enhances LLM applicability and precision. Our framework integrates IoT-driven data retrieval with structured interaction evaluation and domain-specific analysis. Leveraging Image-Metadata Analysis, Named Entity Recognition, and Knowledge Graphs, COGNET-MD-X-REQ identifies weak performance areas by analyzing multimodal model outputs across medical subdomains. We selected GPT-4V to use in the evaluation of our proposed framework on a set of publicly available image-based MCQs in General Pathology. The model achieved 84% accuracy, with notable weaknesses in cardiovascular conditions like atherosclerosis. The framework pinpointed domain-specific deficiencies, enabling targeted fine-tuning. The evaluation leads to the conclusion that our proposed COGNET-MD-X-REQ framework introduces a precision-focused, iterative approach to fine-tuning LLMs by dynamically identifying under-performing areas or knowledge domains that can be related to organs or/and specific conditions and health states. This method reduces reliance on broad retraining, making it suitable for resource-sensitive and safety-critical domains such as medical diagnostics.

In an era of information overload, distinguishing fact from misinformation is a critical challenge. We introduce AlethaNet, a multilayered linguistic framework for automated fact-checking that integrates syntactic, semantic, and pragmatic analysis with contextual metadata. Our approach combines feature extraction, machine learning (XGBoost), and SHAP-based interpretability with a dynamic credibility scoring system. This score is computed using a weighted formula incorporating historical truthfulness, speaker metadata (e.g., job title, party, platform), and AI-inferred reasoning via DeepSeek R1—a generative LLM shown to approximate expert evaluations on the LIAR dataset. Unlike static, binary classification models, AlethaNet frames the problem as truthful vs. misleading, offering a more nuanced detection of misinformation. Experiments demonstrate improved accuracy (75.65%) when incorporating credibility scores. DeepSeek R1 extends the framework’s ability to assess previously unseen sources, while SHAP ensures transparency by highlighting the most influential features. AlethaNet thus presents a scalable, explainable, and autonomous method for adaptive credibility assessment, advancing AI-driven misinformation detection.

Artificial Intelligence (AI) has been growing significantly recently, based on machine learning, deep learning and the latest Generative AI and Large Language Models (LLMs), like ChatGPT. All AI tools are promising to improve decision-making in many domains, including complex ones, such as energy. As such, there have been energy-specific AI systems that incorporate domain expertise, in addition to general-purpose LLMs that integrate knowledge on a plethora of domains, including energy, based on their training from the internet. However, AI systems can produce errors due to their probabilistic nature, and many users, like energy stakeholders, lack the training to use them effectively. Which AI do we mean and how well is it trusted. It is worth exploring differences in AI systems.By employing the VIRTSI model, this paper compares trust states of energy stakeholders on domain specific AI systems versus general-purpose generative AI. For the purposes of the comparison, we use a domain specific AI-based Energy system entitled AI-ENERGIA-SYS that has been previously developed versus ChatGPT 4.0 by OpenAI. VIRTSI is a rigorous computational model for the dynamics of human trust states, spanning from overtrust to distrust, through user modelling and quantifies the efficiency of the interaction in VIRTSI-adapted confusion matrices. The findings reveal that ChatGPT is persuasive and user-friendly to stakeholders, but its lack of domain awareness and explainability leads to overtrust, risking decision quality. In contrast, AI-ENERGIA-SYS, though more complex, supports better trust calibration, meaning that, when trusted, it is usually correct, but it is not trusted so frequently as ChatGPT. The results suggest that future combinations of energy-specific AI systems with general-purpose generative AI could provide improved trust dynamics and more effective decision support.

Objective: Large language models constitute a breakthrough state-of-the-art Artificial Intelligence technology which is rapidly evolving and promises to aid in medical diagnosis. In this study, we propose a novel evaluation framework for extraction of fine-tuning requirements based on the Objective Structured Clinical Examinations (OSCE) that can increase LLM potential and applicability. Methods: We developed an OSCE based evaluation meta-framework leveraging IoT-based data retrieval with a two-step approach designed to analyze and guide improvement of LLMs in multimodal medical diagnosis: (1) structured interaction evaluation and (2) domain-specific analysis of extracted data. Using Image-Metadata Analysis (IMA), Named Entity Recognition (NER), and Knowledge Graphs (KG), this framework identifies image domains, extracts relevant entities, and assesses connections in KGs. These methods collectively reveal areas for improvement, guiding fine-tuning to enhance diagnostic accuracy and contextual understanding in medical applications. Results: Using this paradigm, (1) we evaluate the correctness and accuracy of generated medical diagnosis with publicly available multimodal-multiple-choice-questions in the vast domain of General Pathology and (2) proceed to the domain-specific analysis. We identify and visualize the model performance across specific organs, diseases, and pathological themes, detecting areas of lower accuracy, such as in cardiovascular conditions like atherosclerosis. This targeted approach enables precision-focused fine-tuning, applying additional data to specific weaknesses rather than a broad, generalized tuning across all pathology. Contributions: Our framework’s primary contribution is its OSCE-inspired ability to dynamically identify and target under-performing areas within a broad domain, enhancing fine-tuning efficiency and diagnostic accuracy in a resource-effective and iterative manner removing the dependency on bulk adjustments, making it particularly suitable for sensitive applications where precision and resource efficiency are essential, such as in medical diagnostics.

In this paper, we explore memory and Schema through the lens of mathematical representation. By defining themes related to cognitive processes, we propose a compression algorithm that measures and retains important information of previous (historic) user-Generative AI interaction. Time and memory are interconnected via a decay mechanism, where context gets more abstract as time passes. At the same time, memories with thematic resemblance can be structured into Schemas using set theory, based on rules influenced by a maturity threshold. This threshold is determined by factors such as criticality, emotional-user interaction, and mass, which is the measure of the summation of related memories. Our approach reconstructs memories into reusable parameters and formulates Schemas, offering a foundation for more personalized and cost-effective human-GAI interactions.

In Large Language Models (LLMs), such as GPT, BERT, Mistral or others, “reducing the domain space” for text generation involves limiting the range of content that can be utilized for generating responses. This approach aims at enhancing the relevance and precision of the text produced. While various strategies exist to achieve this, this work explores Sequential Language Model Integration (SLMI), which mirrors the organization and distribution of knowledge across different fields of expertise. More specifically, SLMI is the technique of linking multiple LLMs (LLM-Chains) in a systematic manner. In this paper, we refer to a process of choosing, linking and connecting LLMs with other services (often to complete a generative task, invoke external functions and machine learning services, or tackle problems) as “Large Language Models as a Service”. We outline the development and evaluation process of an SLMI methodology to refine response accuracy. Focusing on the medical field, we also establish a framework for knowledge reduction based on “knowledge paths”, analogous to the distinct specializations within medicine. We apply this framework to a dermatology case study and utilize our evaluation pipeline to assess the results. Reducing the knowledge domain from medicine in general down to dermatology, we tested our methodology and found gains regarding accuracy and diagnostic improvement, as well as a reduction in costs regarding total tokens generated.

This paper introduces Morpheas, an AI-empowered sleep evaluation and calibration system that leverages state-of-the-art technologies, like Large Language Models (LLMs) and Named Entity Recognition (NER). Morpheas integrates Sequential Language Model Integration (SLMI) workflows to simulate the initial steps of sleep disorder diagnosis, utilizing the GPT-4 engine enhanced with Rules of Conduct. Using SLMI and medical and psychological diagnostic tools, we propose a novel multi-step personalisation methodology for creating a gradation system for the improvement of patient-AI interactions. To test and showcase this personalisation approach, we simulate keeping a sleep diary for diagnosing insomnia using a trait-based personalised LLM aimed at addressing sleep concerns. For this purpose, we apply the Big Five Personality Traits (BFPT) where LLMs are again used to extract the responses and facilitate the patient throughout the process, providing guidance and explanation. We then extract the entities from these interactions with NER, in order to identify patterns, provide an explainability basis for patients and effectively customize Cognitive Behavioral Therapy for Insomnia (CBTi), whether through one-on-one sessions or via digital platforms.

The rapid integration of intelligent processes and methods into information systems in the Artificial Intelligence (AI) era has led to a substantial shift towards autonomous software decision-making. This evolution necessitates robust human oversight, especially in critical domains like Healthcare, Education, and Energy. Human trust in AI plays a vital role in influencing decision-making processes of users interacting with AI. This paper presents VIRTSI (Variability and Impact of Reciprocal Trust States towards Intelligent systems), a novel rigorous computational model for human-AI Interaction. VIRTSI simulates human trust states, spanning from overtrust to distrust, through user modelling. It comprises: 1. A trust dynamics representational model based on Deterministic Finite State Automata (DFAs), illustrating transitions among cognitive trust states in response to AI-generated replies. 2. A trust evaluation model based on Confusion Matrices, originating from machine learning and Accuracy Metrics, providing a quantitative framework for analysing human trust dynamics. As a result, this is the first time that trust dynamics have been thoroughly traced in a representational model and a method has been developed to assess the impact of possibly harmful states like overtrust and distrust. An empirical study on the recently launched Large Language Model of generative AI, ChatGPT (version 3.5), provides a radical underexplored AI-generated platform for evaluating the human-AI interaction through VIRTSI. The study involved 1200 interactions of real users as well as AI experts together with experts in two very different domains of evaluation, namely software engineering and poetry. This study traces trust dynamics and the emerging human-AI interaction, in concrete examples of real user synergies with generative AI. The research reveals the vital role of maintaining normal trust states for optimal human-AI interaction and that both AI and human users need further steps towards this goal. The real-world implications of this research can guide the creation and evaluation of user interfaces with AI and the incorporation of functionalities in the development of generative AI chatbots in terms of trust by providing a new rigorous DFA representational method of trust dynamics and a corresponding new perspective of confusion matrix evaluation method of the dynamics’ impact in the efficiency of human-AI dialogues.

In this paper, we evaluate the validity, accuracy, usefulness, and specificity of medical diagnoses related to lung cancer and its staging provided by ChatGPT based on symptoms described by humans. The evaluation is grounded on three main pillars: the validity and accuracy of answers in relation to context and associated references. The specificity and usefulness of the information for both doctors and patients. The economic value added to the healthcare system, determined by several weighted factors derived from the provided answers. The system’s responses are expected to return proposed diagnoses and diagnostic steps, ranked by probability and importance. A specialist conducts the review process.

In this paper, we explore multiple theoretical frameworks to understand and predict user behavior concerning the adoption of innovative, AI-empowered technologies in healthcare. Specifically, our research centers on evaluating the potential adoption rate of AI-empowered medical applications among physicians. To provide empirical support for our investigation, we carried out a comprehensive study employing questionnaires that were disseminated to a practicing medical doctors and medical students. Our methodological framework incorporates two key theories: the Technology Acceptance Model (TAM) and the Diffusion of Innovation Theory (DOI). Utilizing these theories allows us to examine critical factors that influence physicians' willingness to adopt new technologies, such as perceived ease of use and perceived usefulness. Through a nuanced understanding of doctors' perceptions and attitudes toward AI, our research aims to craft targeted strategies that could enhance the rate of adoption for these cutting-edge medical technologies. The overarching goal is to accelerate the integration of AI applications into clinical practice, thereby improving healthcare outcomes and operational efficiencies.

In this paper, we investigate the enhancement of Artificial Intelligence (AI) technologies in healthcare and the better understanding of medical literature with the use of Large Language Models (LLMs) and Natural Language Processing (NLP). Specifically, we introduce a rule-augmented AI-empowered system which incorporates a rule-based decision system, the ChatGPT application programming interface (API), and other external machine learning and analytical APIs to offer diagnostic suggestions to patients. The complexities of patient healthcare experiences, including doctor-patient interactions, understanding levels, treatment procedures, and preventive care, are considered. We illustrate how a diagnostic process typically integrates various strategies depending on various factors. To digitize the greatest portion of the process, we propose and illustrate the use of LLMs for humanizing the communication process and investigating ways to reduce burdens and costs in primary healthcare. We also outline a theoretical decision model for evaluating the use of technological components from external sources versus building them from scratch. The paper is structured into sections detailing background theories and context, our proposed and implemented rule-augmented AI-empowered system, as well as a system test in a corresponding use case. Finally, the paper key findings are presented, which contribute valuable insights for future work in this field.

The great momentum of Artificial Intelligence-empowered applications makes the requirement for detailed and tailored explainability frameworks more crucial. This is particularly evident in the medical domain, where validation of methodologies and outcomes is very important to the adoption of such systems. The depth and the level of understanding of Artificial Intelligence-related concepts is a significant design parameter and necessitates a systemic approach to ensure that a proper level of transparency is incorporated in an Artificial Intelligence-empowered application. In this paper, we propose a novel and generalised approach for the analysis of user requirements and abilities in relation to Artificial Intelligence-empowered applications. Specifically, we use the Technology Acceptance Model (TAM), as a technical methodology to measure the user perception of usefulness and usability of a technology and, subsequently, identify the corresponding depths of explainablity requirements. As a result, we design a layered and personalised explainability framework that may increase adoption rates of domain-specific Artificial Intelligence-empowered technologies.

Abstract An educational game aims to employ the pleasant and fascinating environment of a game for educational purpose. However, when the game’s educational content or playing environment does not match with the player’s learning needs or game-playing skills, the player can find the game either too easy or too difficult and decide to drop it. Therefore, a successful educational game has to offer Dynamic Difficulty Adjustment (DDA) of both the game’s educational content and the game’s playing environment. Considering the above, in this paper a fuzzy-based DDA mechanism of a 3D-game, that teaches the programming language ‘HTML’, is presented. The game adapts dynamically the difficulty of battles and mazes navigation to each individual player’s skills. The novelty of the presented game lies on the application of DDA in the two dimensions (educational content and playing environment) of an educational game and on the use of fuzzy logic for succeeding balance between the game’s difficulty and the user’s playing skills. The adaptation is realized in real-time, during the playing and not each time the player changes game level/track. Furthermore, fuzzy logic allows to apply DDA without the need of many learning data, like other Artificial Intelligence techniques for DDA (i.e. machine learning techniques). The presented game was evaluated through a comparative evaluation, which included questionnaires, an experiment in real conditions and t-test statistical analysis method. The evaluation results showed that the incorporated adaptation mechanism increases the user’s motivation and succeeds to keep players’ interest for the game undiminished.

The Rational Unified Process (R.U.P.) is an iterative Software Engineering Process, that ensures alignment between engineers and stakeholders through optimized and detailed partinionalised steps within predefined constraints [1]. The independently deployable services that are components of distributed systems are called microservices. They are part of applications and are easier to manage and scale. Each microservice serve different purpose and has a unique responsibility making it easier to understand manage and collaborate on. Medical applications are created to assist in treatment, disease prevention and health optimization. However patients' needs and abilities vary and patients' requirements and definitions on the usability aspect of an application are different and should be acknowledged for the application to be successful and for the patients/users to benefit from it. In this study the development of an A.I.-infused medical application, is outlined through the R.U.P. methodology and built using microservices, where the patients' needs and requirements are at the center of continuous development process focused on improvements.

In this paper, we present a full evaluation of the learning outcomes of an Intelligent Tutoring Systems, which teaches the logic of computer programming and the programming language ‘C’. The adaptation of the evaluated ITS is based on a fuzzy mechanism, which is responsible to identify the learner’s current knowledge level and misconceptions. The system takes into consideration the knowledge dependencies that exist among the domain concept of the learning material and applying fuzzy rules decides about the learning material that has to be delivered to the learner and the lesson sequence. The system was used in real conditions by 70 students of a postgraduate program in Informatics of the University of Piraeus in Greece. After the period of system’s experimental usage, the following four metrics- characteristics were calculated: (1) the learners’ performance, (2) the number of the learners that dropped out from the usage of the system, (3) the number of learners that succeeded to reach the target knowledge by completing, successfully, the lessons of the system, and (4) the mean number of interactions with the system that are needed until the learner reaches the target knowledge. The calculated numbers were compared with the corresponding results after the usage of a similar Intelligent Tutoring System from which the fuzzy mechanism was absent. For ensuring the validity of the evaluation results, t-tests were conducted. The evaluation results show that the presented fuzzy-based Intelligent Tutoring system increases the learners’ knowledge level and offers a personalized learning experience that promotes the active participation of students in the learning process and decreases the number of dropouts.

The rapid progress in the deep learning field and high-performance computing, has contributed significantly to the growth of object detection in images techniques. This paper presents the development of an innovative smartphone application, which recognizes and counts objects in images that users offer. As we all know, object detection, is a complicated algorithm that consumes high-performance hardware to be executed in real time. Therefore, because of the limited computational power of smartphones, it is necessary to develop a methodology, so the process of recognizing and counting items, will be achieved effectively in terms of time and resources.

The digitization of intangible cultural heritage has given rise to new needs but also great opportunities for the preservation, analysis, education and propagation of cultural items that often constitute vast data bases of sophisticated works of art. This is particular the case of songs that comprise a combination of lyrics and music, each constituting an important domain of art. The preservation and propagation of this kind of art around the world becomes yet harder by taking into account the different languages that this art was originally created in, e.g. English, French, Portuguese Greek, Ancient Greek etc. The aim of this specific research is to develop a software tool that will support computer-aided functions of musical and lyrical analysis. The provision of these particular functions is realized through the integration in the proposed software tool of advanced natural language processing algorithms, text mining and thematic modeling. Therefore, it is an application that proposes the utilization of innovative computational methods of the broader object of Machine Learning and Pattern Recognition for the automated processing, clustering and semantic analysis of textual information. As a case of observation, the automated analysis of a part of the poetic activity of a Greek lyrics-writer is presented with the ultimate goal of promoting the cultural heritage.

Sentiment analysis is a rather intriguing subject that modern ICT tools enable us to explore and analyze. In this work we perform, to the best of our knowledge, the most wide analysis of sentiment mapping to geographic locations and time through smartphones, in an attempt to both visualize them and also reveal possible correlations and patterns. Our vast dataset consisting of more than 56.000 samples, from 100 individuals, for a time period of nine months, revealed patterns, both in space and time, that are directly linked to geographic locations of users and provide an aggregated real-time insight on how people feel, allowing for a wide range of applications.

The explosive growth in the amount of available digital information has increased the demand for recommender systems. Recommender systems are information filtering systems that deal with the problem of information overload by filtering vital information fragment out of large amount of dynamically generated information according to user's preferences or interests. Recommender systems have the ability to predict whether a particular user would prefer an item or not based on his/her personal profile. To this direction, this paper presents multi-algorithmic techniques, such as content-based filtering and collaborative filtering, which increase the efficiency of recommender systems. Moreover, a hybrid model for recommendation, employing content-based and collaborative filtering, is introduced. The presented recommender system takes as input information about users from their profile in Facebook, one of the most well-known social networking services. Examples of operation are given and they hold promising results for the described techniques. Finally, the paper attests that the aforementioned techniques can be used for different kind of software, such as e-learning, e-commerce, etc.

Engagement of learners in computer-based tutoring constitutes an important feature that is sought by learning applications to maximise the educational effectiveness. Moreover, the recognition of human emotions in interactive computer-based learning applications has also been considered important although it had been overlooked for many years in the past. In view of these, dynamic personalisation and a variety of artificial intelligence techniques offer new perspectives, insights and realistic results in rendering computer-based learning more human-like and engaging than it has ever been before. This talk will present and discuss research challenges and effective approaches towards a new era of personalisation and human-like behaviour of computer-based learning software. I will review the research advancements on this topic that have been achieved in the software engineering lab of our department. In particular, the talk will focus on the automatic analysis of computer observations in computer-based learning that are collected, taking input from at least three modalities of interaction, namely the keyboard, camera and microphone in conjunction with contextual information and are then used to draw inferences about the users' cognitive status, reasoning, social classmate behaviour, preferences and emotions. In return, the tutoring content and the user interface is automatically adapted accordingly to address the individual user's needs by making appropriate recommendations and presenting adaptive guidance. Contextual information differs depending on the kind of computer- based learning applications. As such, a diversity of paradigms of fully developed and evaluated computer-based learning systems, in our lab, will be presented and discussed to exemplify the above research in the context of virtual reality educational games, social network learning, mobile learning and multi-modal stand-alone learning, based on the combination of cognitive theories about human reasoning and emotions, machine learning algorithms, decision making theories and fuzzy logic. The talk will conclude by highlighting open research areas for further research.

Towards the last decade, digital education has become a burning issue in the related scientific literature and involves the production of Intelligent Tutoring Systems (ITSs). ITSs are adaptive educational applications that enrich the tutoring and learning processes with "intelligence" by divulging the abilities and weaknesses of each student, in order to provide him/her with a personalized learning experience. A crucial factor of adaptive learning systems is testing, and indeed adaptive testing. It is a challenge to create an adaptive test that includes the most suitable exercise/question/activity of a large pool of test items for a particular learner taking into consideration her/ his particular learning characteristics, needs and ability. In this paper, a framework for creating automated adaptive tests using multiple-criteria decision analysis and the weighted sum model is presented. The presented framework takes into consideration multiple students' criteria along with the types of exercises and the desirable learning objective. The aforementioned assessment framework was incorporated in two adaptive e-learning systems and was fully evaluated. The evaluation results are very encouraging.

Social Networking Services can serve as a great platform for learning. As such, the use of Facebook in learning contexts can be proved beneficial. Following this direction, this paper presents a prototype Facebook application for learning which is supported by Latent Dirichlet allocation (LDA). LDA is a generative model that allows sets of observations to be explained by unobserved groups that clarify why some parts of the data are similar. Hence, making automatic predictions about the interests of students can be made by collecting their preferences and characteristics. Hence, by tracking user interests, accurate recommendations can be made. The experimental results, presented in this paper, reinforce the view that automatic predictions using LDA to students in social networks can be a powerful idea in personalizing instruction.

Conversational interfaces are used for a variety of applications. They are constructed to offer useful services and to interact with the users in order to assist them. Towards this direction, the current paper presents the incorporation of the interactive chatterbot ALICE in a mobile-assisted English language learning application. This chatterbot is further enriched with mechanisms in order to support students while learning vocabulary in English. As such, apart from making conversation with this conversational agent, they can practice their vocabulary or even be evaluated by the chatterbot. Hence, this study offers a fertile ground to enhance pedagogical results such as fostering motivation and engagement, incrementing crucial language learning and assisting in the acquisition of cognitive skills. Finally, the students can chat with the pedagogic conversational agent either orally or by writing.

In this paper, a novel approach to web-based education that performs individualized instruction on the domain of programming languages is presented. This approach is fully implemented and evaluated in an educational application module, called fuzzy knowledge state definer (FuzKSD). In particular, FuzKSD performs user modeling by dynamically identifying and updating a student's knowledge level of all the concepts of the domain knowledge. The operation of FuzKSD is based on fuzzy cognitive maps (FCMs) that are used to represent the dependences among the domain concepts. FuzKSD uses fuzzy sets to represent a student's knowledge level as a subset of the domain knowledge. Thus, it combines fuzzy theory with the overlay model. Moreover, it employs a novel inference mechanism that dynamically updates user stereotypes using fuzzy sets. It should be noted that the overlay model and stereotypes constitute two widely used methods for user modeling. However, they have not been combined with fuzzy sets thus far in the literature. The gain from this novel combination is significant as a student level of knowledge is represented in a more realistic way by automatically modeling the learning or forgetting process of a student with respect to the FCMs and thus the system can provide individualized adaptive advice. The application of this approach is not limited to adaptive instruction. It can also be used in other systems with changeable user states, such as e-shops, where consumers' preferences change over time and affect one another. Therefore, the particular module constitutes a novel generic fuzzy tool, which offers dynamic adaptation to users' needs and preferences of adaptive systems.

This research investigates the novel combination of multicriteria decision-making theories with a cognitive theory called human plausible reasoning (HPR) to provide personalized assistance via graphical user interfaces (GUIs). A GUI called intelligent file manipulator (IFM) helps with organizing computer file storage. The system reasons about user actions, goals, plans, and possible errors and offers automatic assistance in case of a problematic situation. Three multicriteria decision-making theories [simple additive weighting, multiattribute utility theory, and data envelopment analysis] were adapted, implemented, and combined with HPR, in turn. This process resulted in three different versions of IFM that were evaluated. The protocols of 30 users of different levels of expertise were provided to the system with each of the versions and to ten human experts. The responses of the experts and the different versions of the system were compared. The evaluation revealed that decision-making theories can increase control in an intelligent user interface. They can be combined with a cognitive theory like HPR. There was 66-81% compatibility of the different systems' responses to the proposals of the majority of experts. All decision-making theories were better in simulating the human experts' reasoning when there was unanimity of the human experts' opinion. In such cases, there was 68-84% compatibility of the different systems' responses to the experts' proposals. This research supports that HPR could successfully be completed by a multicriteria decision-making theory, and such combination can be effectively used for providing personalized help in GUIs.

Towards the creation of a multiple language learning environment which supports and enhances collaboration among its students we propose an approach that uses user modeling and machine learning. The well known theory of user modeling is used to collect user characteristics and as second step a classical machine learning approach is incorporated in order to intelligently use these characteristics to create student groups. The resulting student groups promote win-win collaboration, thus support the learning process and provide additional educational benefits for the learners.

Towards building a multimodal affect recognition system, we have built a facial expression recognition system and a audio-lingual affect recognition system. In this paper, we present and discuss the development and evaluation process of the two subsystems, concerning the recognition of emotions from audio-lingual and visual-facial modalities. Many researchers agree that these modalities are complementary to each other and that the combination of the two can improve the accuracy in affective user models. Therefore in this paper we present a combination of two modes using multi-criteria decision making theories. The resulted system takes advantage of the strengths of each mode and is more accurate in emotion recognition.

In this paper, we investigate the possibility of improving the accuracy of visual-facial emotion recognition through use of additional (complementary) information. The investigation is based on three empirical studies that we have conducted involving human subjects and human observers. The studies were concerned with the recognition of emotions from a visual-facial modality, audio-lingual and keyboard-stroke information, respectively. They were inspired by the relative shortage of such previous research in empirical work concerning the strengths and weaknesses of each modality so that the extent can be determined to which the keyboard-stroke and audio-lingual information complements and improves the emotion recognition accuracy of the visual-facial modality. Specifically, our research focused on the recognition of six basic emotion states, namely happiness, sadness, surprise, anger and disgust as well as the emotionless state which we refer to as neutral. We have found that the visual-facial modality may allow the recognition of certain states, such as neutral and surprise, with sufficient accuracy. However, its accuracy in recognizing anger and disgust can be improved significantly if assisted by keyboard-stroke information.

In this paper we describe an intelligent and adaptive TV-shopping application that has been designed to recommend mobile phones to perspective buyers. The application is called iTVMobi and its adaptivity is achieved through a user model. The user model monitors users' behaviour and adapts the interaction to them depending on their inferred interests. The underlying reasoning is based on grouping users with similar tastes by the use of k-means clustering. After the clustering, representative vectors of users' preferences are calculated and compared through association rules with the products database in order to achieve product recommendations.

This paper presents a user model server based on Web Services. User model servers are very important because they allow reusability of user modeling reasoning mechanisms which are typically very complex and difficult to construct from scratch. In this paper we show how the potential of interoperability, reusability and component sharing offered by the technology of Web Services have been exploited in the design of a user model server that performs decision making. The reasoning of the user modeling is based on a multi-criteria decision making theory and has been implemented as a Web Service to provide intelligent assistance to users over the Web. Reusability has been shown through the successful application of the user model server into two different applications: an e-mail and a file manager application.

This paper presents an affective educational application that bases its inferences about students' emotions on user evidence provided by the keyboard and the microphone. The actual combination of evidence from these two modes of interaction has been performed based on our innovative inference mechanism for emotions that uses stereotypes of users' affective reactions and a multi-criteria decision making theory. The user stereotypes have been resulted from an empirical study that resulted in recording common user reactions of the target group of users of the educational application

In this paper we describe a Web-based educational application for individualised instruction on the domain of programming and algorithms. The application adapts to each user dynamically. This is achieved via the system's user model, which relies on stereotypes. The determination of the stereotypes is based on the knowledge level of the learner. The transition of a learner from one stereotype to another one reveals her/his learning state. In particular, learners are modelled for all possible cognitive states with respect to their progress or non-progress, while they interact with the educational application

Adaptivity in training software systems is of great importance as it can adjust one single training environment to the needs of each individual trainee. ASSET is such a UML training software system as it provides adaptive support to both the trainer and the trainees. Specifically, it provides an environment that can be parameterised by the trainer to suit the training needs of his/her company or educational institution. The adaptive help to the trainer concerns mainly advice on the best schemes of collaboration between the trainees and the most efficient and productive forming of group as well. The adaptive help destined for the trainee refers to knowledge, user type and personality attributes. The special characteristic of ASSET is that it does not base its inferences on specific tests, but constantly monitors every action of the trainees during their interaction with the system

Decision-making theories aiming at solving decision problems that involve multiple criteria have often been incorporated in knowledge-based systems for the improvement of these systems' reasoning process. However, multicriteria analysis has not been used adequately in intelligent user interfaces, even though user-computer interaction is, by nature, multicriteria-based. The actual process of incorporating multicriteria analysis into an intelligent user interface is neither clearly defined nor adequately described in the literature. It involves many experimental studies throughout the software life-cycle. Moreover, each multicriteria decision-making theory requires different kinds of experiments for the criteria to be determined and then for the proper respective weight of each criterion to be specified. In our research, we address the complex issue of developing intelligent user interfaces that are based on multicriteria decision-making theories. In particular, we present and discuss a software life-cycle framework that is appropriate for the development of such user interfaces. The life-cycle framework is called MBIUI. Given the fact, that very little has been reported in the literature about the required experimental studies, their participants and the appropriate life-cycle phase during which the experimental studies should take place, MBIUI provides useful insight for future developments of intelligent user interfaces that incorporate multicriteria theories. One significant advantage of MBIUI is that it provides a unifying life-cycle framework that may be used for the application of many different multicriteria decision-making theories. In the paper, we discuss the incorporation features of four distinct multicriteria theories: TOPSIS, SAW, MAUT, and DEA. Furthermore, we give detailed specifications of the experiments that should take place and reveal their similarities and differences with respect to the theories

This paper examines the utility of a multiple attribute decision making method, the Simple Additive Weighting (SAW), for the purposes of an Intelligent Learning Environment (ILE) that provides adults with personalised e-learning. The ILE is called Web Intelligent Trainer and is meant to help novice users learn how to manipulate the file store of their personal computer. The generation of advice makes use of adaptive hypermedia techniques and is adapted to each individual learner's needs, depending on their knowledge level, age, habits and difficulties. SAW has been applied in the ILE and has been evaluated with respect to the performance of the ILE. As a result, SAW seems particularly appropriate for the ILE.

This paper describes the adaptation of a cognitive theory, called Human Plausible Reasoning (HPR), for the purposes of an intelligent graphical user interface (GUI). The GUI is called intelligent file manipulator (IFM) and manages files and folders in a similar way as the Windows 98/NT Explorer. However, IFM also incorporates intelligence, which aims at rendering the interaction more human-like than in a standard explorer in terms of assistance to users' errors. IFM constantly reasons about users' actions, goals, plans, and possible errors, and offers automatic assistance in case of a problematic situation. HPR is used in IFM to simulate the reasoning of users in its user modeling component and the reasoning of human expert helpers when they try to provide assistance to users. The adaptation of HPR in IFM has focused on the domain representation, statement transforms, and certainty parameters. The certainty parameters of HPR have been combined in a novel way with user stereotypes and the simple additive weighting theory. IFM has been evaluated and the evaluation results showed that IFM could generate plausible hypotheses about users' errors and helpful advice to a satisfactory extent; hence, HPR seemed to have fulfilled the purpose for which it was incorporated in IFM.

We describe how a multiagent, intelligent learning environment can provide adaptive tutoring based on learner modelling through Web services. The system is called Web F-SMILE and is meant to help novice users learn how to manipulate their file store. The system has assigned an agent, called learner modelling (LM) agent, to constantly observe the user and collect information about him/her. This information is used by other agents in order to provide individualised help and tutoring. The LM Agent interacts with a learner modelling server through Web services. In this way, each learner model is available to any client application that requests it. The main advantage of this approach is that the learner model is not reinitialised every time the user changes the PC s/he uses. Moreover, Web services allow information to pass through most security systems and thus the reliability of the learner model information is ensured.

The issue of initializing the model of a new student is of great importance for educational applications that aim at offering individualized support to students. We introduce a general framework for the initialization of the student model in Web-based educational applications. According to the framework, the student modeler makes initial estimations of the new student's knowledge level and error proneness based on the models of other similar students. This is done by applying a machine learning technique. The similarity between students is calculated taking into account different student characteristics for different teaching domains. We have implemented the proposed methodology in two different tutoring domains, namely language learning and mathematics. An evaluation study conducted in the case of the Web-based language learning system, showed that the use of the framework can initialize student models in a sufficiently accurate way, given the little information about individuals available.

We have developed a virtual reality (VR) game for tutoring English as a foreign language over the Web. The educational game is called VIRGE (virtual reality game for English) and intends to motivate and engage students through the use of a game. We have shown how an ILTS may be combined with the technology of virtual reality games and used over the Web for teaching English as a second language. The benefits of the combination of the above technologies are particularly important for the domain of a foreign language teaching because they allow the creation of engaging and motivating educational applications that are cross-cultural.

We describe an agent that has been incorporated as an evaluation component into a knowledge-based authoring tool for educational games. The agent is meant to simulate the behaviour and memory of a student and aims at helping authors to evaluate the courses that they have authored before these are delivered to real students. Authors may try their course using the simulated student-player and they can see what a real student is likely to learn and remember after each lesson. In case authors are not completely satisfied with the result they may redesign the whole course or some parts of it. Such components are very useful for authoring tools because they encourage and facilitate multiple iterations of the design process and thus ensure better quality of the resulting software applications.

This paper presents the way that an Index-Maximizing Linear Programming (LP) Model has been applied for dynamic reasoning in an intelligent Graphical User Interface. The graphical user interface is called I-Mailer and provides intelligent help to users during their interaction with an e-mailing system. For this purpose, I-Mailer constantly reasons about every user action in terms of the user's goals and plans. In case the user is involved in a problematic situation, it transforms the user's problematic action and generates alternative actions to be suggested to the user. The actions suggested should be better than the problematic one in the context of the user's hypothesized intentions. However, this process usually results in the production of too many alternative actions that could be suggested to the user instead of the one issued. To solve this problem, we have applied an Index-Maximizing LP model in order to select the best alternative actions to be suggested to a user.

This paper describes the conversion of an existing standalone, multi-agent, intelligent tutoring system (ITS) to one that can operate trough the Internet. The particular ITS works as a virtual reality game and is meant to help students learn by providing them a more motivating environment and tutoring help. The system has three agents that have explicit roles. The companion that gives friendly advice, the advisor that provides help and the guard that asks the questions. In order to achieve the goal of the conversion we used a new technology, that of the Web services which has proved very satisfactory for the purposes of our research. We describe the conversion of our standalone multi-agent application to a web-based one, and we discuss the benefits of web services for the purposes of educational software.

Intelligent tutoring systems are meant to provide individualised tutoring to students by adapting the teaching material to their specific needs and abilities. However, their development is a hard task and the end-result is difficult to evaluate. In this paper we present a novel approach for the evaluation of these systems, which relies on an agent that may be used as a simulated student-user. The evaluation agent incorporates modelling techniques of real users that are based on both cognitive and temperamental data. The cognitive model is based on cognitive psychology and simulates the memorisation and retention capabilities of a student. The temperamental data creates an image of the student concerning the way she/he behaves and the kind of personality she/he has. Developers may evaluate the tutoring systems using the agent rather that real students. Thus better quality of the end result may be achieved at no cost of the educational process.

Web-based education is particularly good for remote teaching and learning at any time and place, away from classrooms and without necessarily the presence of a human teacher. However, this independence from real teachers and classrooms may cause emotional problems to students who may feel deprived of the benefits of human-human interaction. In this paper we address this problem by rendering Web-based human-computer interaction more human-like. This is achieved by an authoring tool that human teachers may use to create their own educational software for their remote students. Through this authoring tool, human teachers may also create their own tutoring characters for the user interface of the educational application. These characters may be parameterised in many aspects, the way they speak, the pitch, speed and volume of their voice, their body-language and the content of their messages. In this way they may reflect the human teachers' vision of teaching behaviour.

Authoring tools for intelligent tutoring system (ITSs) are meant to provide environments where instructors may author their own ITSs in varying domains. In this way, painful constructions of ITSs, which are not reusable, my be avoided. However, the construction of an authoring tool is associated with many problems, such as the generality of the techniques incorporated, domain-independence, effectiveness for the prospective authors (instructors), and effectiveness for the students who will use the resulting ITSs. We report on an empirical study that we conducted in order to design and develop WEAR, an ITS authoring tool for algebra-related domains operating over the Web. We investigated several aspects concerning the attitude and behaviour of both students and instructors. We also considered aspects relating to the fact that WEAR would be operating over the Web. The study revealed important issues and was then used for the specification of the design of WEAR. A brief description of the developed system is also included in the paper so that the way that the design specification were put late practice may be shown.

This article is about a graphical user interface (GUI) that provides intelligent help to users. The GUI is called IFM (Intelligent File Manipulator). IFM monitors users while they work; if a user has made a mistake with respect to his or her hypothesized intentions, then IFM intervenes automatically and offers advice. IFM has two underlying reasoning mechanisms: One is based on an adaptation of a cognitive theory called human plausible reasoning and the other one performs goal recognition based on the effects of users' commands. The requirement analysis of the system has been based on an empirical study that was conducted involving real users of a standard file manipulation program like the Windows Explorer; this analysis revealed a need for intelligent help. Finally, IFM has been evaluated in comparison with a standard file manipulation GUI and in comparison with human experts acting as consultants. The results of the evaluation showed that IFM can produce successfully advice that is helpful to users.

The authors describe the adaptive courseware authoring capabilities of WEAR, an ITS authoring tool for algebra-related domains. The system allows the authoring of electronic textbooks and delivers them over the WWW. Then, learners are offered navigation support adapted to their individual needs and knowledge. The domain knowledge and the information kept in learner models are used in WEAR to provide adaptive navigation to students and also to support the instructional designers in the authoring process. An instructor model incorporated in WEAR's architecture contains additional information that is exploited by the system to achieve the goal of authoring support.

This paper describes a formative evaluation of the advice generator of an intelligent learning environment for novice users of a GUI. 20 human tutors participated in this procedure. The human tutors had to answer theoretical questions about their way of generating advice. In addition, they were asked what they would advise learners in real-life examples which were acquired by a previous empirical study. Their answers were compared to the responses generated by the learning environment. As a consequence, improvements were made to the system based on the evaluations' conclusions.

We describe Web Passive Voice Tutor (Web PVT), an adaptive teaching Web-based intelligent computer assisted language learning (ICALL) program that is aimed at teaching non-native speakers the passive voice of the English language. The design of the system has been largely based on the results of an empirical study that was conducted at schools with the collaboration of human teachers. Web PVT incorporates techniques from intelligent tutoring systems (ITS) and adaptive hypermedia (AH) technologies to provide students with individualised instruction and feedback. The system uses a combination of stereotypes and the overlay technique for the initialisation of the student model, which is then refined by observing the student while working with the system. The resulting student model is used for the annotation of the links to topics presented to the student. In addition, it is also used in the process of error diagnosis and the adaptation of feedback and advice provided to the student.

Human errors occur frequently in the interaction of a user with an operating system. However, current user interfaces of operating systems lack some reasoning ability about user's intentions and beliefs. Intelligent Help Systems (IHS) can provide additional reasoning and help. This paper presents a discussion of the features of IHSs and a review of a few IHSs for users of operating systems. Then it describes the research and results of employing a cognitive theory of Human Plausible Reasoning Theory in error diagnosis for users interacting with an operating system. This theory has formalized the reasoning based on similarities, generalizations and specializations that people use to make plausible guesses about questions. Here we exploit the fact that plausible guesses can be incorrect and thus turned into human errors. The error diagnosis is performed by the user modelling component of an IHS, called RESCUER.