The cognitive load or effort required over time in order to finish a task in a complex system might be perceived as mental workload. Assessing mental workload boundaries could help improve designs and determine if user efficiency is influenced by the design. Mental workload has recently been recognized as a significant measure for the design and assessment of web interfaces. In evaluating the usability of websites, web logs that contain information related to the interaction of users have proven helpful to use information tracked of hundreds of users. A design experiment and predefined formulas were developed to evaluate user activity for a web-based assignment to evaluate whether mental workload can be measured through javascript events. In order to evaluate the position of Mental Workload in evaluating applications using interactive infographics. this experiment consisted of multiple approaches during the development of the task. Three different tasks was passed to 106 members to investigate the user interaction records could be utilized in the study of mental workload in web design. MWL techniques taken into consideration RTLX (Raw-TLX) and NASA-TLX were used to analyse the correlation between the NASA-TLX and RTLX. Even though very little has been possible, only weak correlations were found between predefined formula and the dimensions of the mental workload

Computer technology is strongly accessible for everyone, anywhere, and anytime. Consequently, users of computer technologies have diversified and grown. As more and more individuals promote websites in their daily routines, creating an interaction that is as simple as possible is essential. One of the necessary requirements of the web-design is to keep a balance between the text and picture ratio. Images seem to offer both intimacy and immediacy, whereas text seems to offer only immediacy. To enhance the user experience, it is therefore preferable to maintain the equilibrium between images and text material. Sometimes due to lack of understanding ,the major hindrance to the user within the design structure can be a substantial problem. The need for usability evaluation methods crafted for the Web that support the usability design process has become critical. The developer can take the cognitive load measurement outcomes and work on the issue regions being redesigned. Therefore effective methods for monitoring and evaluating mental workload are highly helpful and should be taken into account when developing web system interfaces. An online experiment was carried to evaluate Mental workload imposed by web-based tasks designed around images and graphics with its text counterparts. The results of this research suggests that the idea of the differences between the MWL imposed by tasks designed around images and texts showed significant changes for most of the Image categories covered in W3C.

Computational trust is an ever-more present issue with the surge in autonomous agent development. Represented as a defeasible phenomenon, problems associated with computational trust may be solved by the appropriate reasoning methods. This paper compares two types of such methods, Defeasible Argumentation and Non-Monotonic Fuzzy Logic to assess which is more effective at solving a computational trust problem centred around Wikipedia editors. Through the application of these methods with realdata and a set of knowledge-bases, it was found that the Fuzzy Logic approach was statistically significantly better than the Argumentation approach in its inferential capacity.

It is illegal in Ireland to discriminate in the provision of education on the basis of multiple characteristics including gender, race and religion. While the increased use of machine learning models can open multiple avenues to identify early intervention strategies in education, caution must be exercised to ensure that any intervention does not discriminate with respect to a protected class. Poor literacy in childhood can have long term effects as the child ages, including on employment and mental health outcomes. Early intervention is key in mitigating this. In this dissertation, a model was created that predicted the outcome of a literacy test at age 9 based on information about the individual child at ages 9 months, 3 years and 5 years, including their development, parental education levels and literacy, early exposure to books and reading, and early educational abilities. Each of these areas had been suggested in current literature to contribute to or be a risk factor for childhood literacy. As is particularly common in survey data, there is missing data. This was dealt with through deductive imputation, exclusion, and automatic imputation. The best performing model as measured by a minimal mean squared error was produced when data was deductively imputed and, where that was not possible, excluded. It was then investigated whether the resultant best performing model discriminated based on gender, race or religion. To achieve this, synthetic sets of ‘twins’ were created who were identical in every feature apart from the protected characteristic. These populations were created from the original data that was used to create the model. The best performing model, which minimised the mean squared error, was shown to explain 33.1% of the variance in literacy scores between children. It was also shown to discriminate based on religion and ethnicity with a weak effect. A model using deductive imputation followed by automatic imputation performed less well and was shown to discriminate based on religion with a weak effect and discriminate based on ethnicity with a weak to medium effect. The overall experiment showed that it was possible to create a model to partially explain the variance in a measure of literacy in 9-year-old children using features from earlier in their childhood. However, this model displays some discrimination based on ethnicity. Although the effect of the discrimination observed is weak, caution should be exercised in the implementation of any real-world interventions based on similar models.

Epilepsy is the second most common neurological disorder and it affects approximately 50 million people worldwide. One of the main characteristics of this disorder is the presence of recurrent seizures which tend to be controlled through medication. Nonetheless, 20% of the patients with this disorder are resistant to drug treatment meaning that they need to go through alternative procedures. One common option is the surgical resection of the epileptogenic tissue in the region of the brain that is responsible for the seizures. This study presents a supervised machine learning approach to identify pathological oscillations in the seizure onset zone based on the presence of high frequency oscillations. The method implements a SVM algorithm using features extracted from iEEG data. The model was trained, validated and tested with 25 patients suffering from refractory epilepsy and then evaluated in a dataset containing additional 8 patients. The algorithm was capable of detecting 77% of the positive cases in the test dataset and an average of 61% in the dataset of additional patients. Also, it presented an AUC of 77% and 85% accuracy in the test dataset. In addition, this work also discusses which are the features extracted from iEEG data that are more relevant to identify the pathological HFO and are able to satisfactorily represent the characteristics of the signals. In this respect, frequency domain features proved to be among the best model predictors. The results indicate that when these attributes and time-frequency domain attributes were added the model performance increased significantly.

This dissertation proposes LightGWAS, a novel machine learning procedure for genome-wide association study (GWAS) based on LightGBM and k-fold cross-validation. The conducted literature review identified that the currently available GWAS implementations rely on massive manual quality control steps to address statistical issues, such as controlling for false-positive inflation and power reduction. It also showed they demand a specific GWAS method for each type of genomic dataset morphology, which consequently increases the human dependency and open margins for misleadings. LightGWAS is a potential single, resilient, autonomous and scalable solution to address such concerns. Through this research, LightGWAS was contrasted against the current state-of-the-art for GWAS throughout secondary research method. It has been compared with a GWAS implementation based on general linear model (GLM) with support to Firth regularisation. Quantitative empirical tests and deductive reasoning have been employed to reach and evaluate the results. The models were submitted to balanced (case:control=1:1), imbalanced (case:control=1:10), and high-imbalanced (case:control=1:100) genomic datasets of binary phenotypes. The results from statistical tests denoted that LightGWAS performs equivalently to the compared GLM method for balanced dataset scenarios, and outperforms for imbalanced and high-imbalanced datasets. The assessed metrics were weighted average of the precision and recall (F1), recall, average precision score (APS), receiver operating characteristic (ROC)/area under the curve (AUC), accuracy, and precision.

Android smart phones have become so prevalent that most of the personal activities and financial activities are conducted through them. With increasing popularity comes security risks. Hackers are trying to exploit the consumers by getting malicious applications installed, which in turn are used to steal personal and financial information. Android maintains security through sophisticated permission mechanism. However, some of the applications gain invasive permissions and attack systems by deceiving users and security mechanism. Hence it is pivotal to consider the actual API calls in the source code to build malware detection models. There are research works that use permission mechanism to detect malicious applications and other research works that use runtime information to classify malware. This research aims at using Malgenome dataset that has both the permission features and API call features in the source code and building models to detect the malware applications. The dataset contains 1260 malicious and 2539 non-malicious applications and has around 180 features. Two datasets containing only permission features and only API calls are created from the primary dataset. SMOTE over sampling was applied to address the data imbalance and ANOVA feature selection was applied to consider the top 20 features for building models to enhance the performance and efficiency. This research uses three machine learning algorithms: Support Vector Machines , Neural Networks, Random Forest to build around 9 models in total on the different datasets containing Permission Only (or) API calls Only (or) Both permission and API calls. When both the features are considered(ensembled), Random Forest produced the best malware detection accuracy 98.42%. The best accuracy for Only permission feature based models is 96.3% and only API call feature based models is 97.57%. This research found there is improvement in malware detection accuracy when both permission features and API call features are considered and reinforces the understanding that ensembling both the feature sets may produce robust malware detection models and can reduce the risk of attacks caused by deceiving the permission mechanism. This approach further addresses the problem of over provisioning and reduces false positives.

Ethereum is a rapidly growing blockchain network with a range of applications including cryptocurrency, identity management and smart grid energy management. These applications operate through the use of smart contracts. However, as the adoption of Ethereum increases, so does the demand for greater scalability in the network. Existing literature points to several areas in which bottlenecks exist, limiting the number of transactions per second. Ethereum developers have questioned whether the current execution engine, the Ethereum Virtual Machine, could be improved with a WebAssembly implementation. This dissertation aims to address the gaps presented in existing literature, to evaluate the performance of smart contracts on clients implementing both an EVM and eWasm execution engine. The designed experiment carries out empirical work to evaluate the performance of Ethereum execution engines. The paper finds that in certain scenarios, an eWasm engine performs better than an EVM engine.

This dissertation investigates the optimization of constructing Argumentative Decisive Graphs (ADGs) within the realm of machine learning, focusing on enhancing their explainability and accuracy. It systematically reviews existing literature on argumentation frameworks and their integration with machine learning techniques, particularly ensemble learning and boosting algorithms. Employing a empirical methodology, the study implements and test a novel building algorithm named Boosted Building Process (BBP) on diverse datasets. The evaluation of models built with the proposed algorithm is made in terms of both accuracy and model size. The results demonstrate the potential of BBP to provide more transparent and efficient models of inference for classification problems, paving the way for broader application. The research concludes with recommendations for further development of ADGs.

Multiple systems have been proposed to perform computational argumentation activities, but there is a lack of options for dealing with quantitative inferences. This multi-layer, web, argument-based framework has been proposed as a tool to perform automated reasoning with numerical data. It is able to use boolean logic for the creation of if-then rules and attacking rules. In turn, these rules/arguments can be activated or not by some input data, have their attacks solved (following some Dung or rank-based semantics), and finally aggregated in different fashions in order to produce a prediction (a number). The framework is implemented in PHP for the back-end. A JavaScript interface is provided for creating arguments, attacks among arguments, and performing case-by-case analyses.

Explainable Artificial Intelligence (XAI) is a research area that clarifies AI decision-making processes to build user trust and promote responsible AI. Hence, a key scientific challenge in XAI is the development of methods that generate transparent and interpretable explanations while maintaining scalability and effectiveness in complex scenarios. Rule-based methods in XAI generate rules that can potentially explain AI inferences, yet they can also become convoluted in large scenarios, hindering their readability and scalability. Moreover, they often lack contrastive explanations, leaving users uncertain why specific predictions are preferred. To address this scientific problem, we explore the integration of computational argumentation—a sub-field of AI that models reasoning processes through defeasibility—into rule-based XAI systems. Computational argumentation enables arguments modelled from rules to be retracted based on new evidence. This makes it a promising approach to enhancing rule-based methods for creating more explainable AI systems. Nonetheless, research on their integration remains limited despite the appealing properties of rule-based systems and computational argumentation. Therefore, this study also addresses the applied challenge of implementing such an integration within practical AI tools. The study employs the Logic Learning Machine (LLM), a specific rule-extraction technique, and presents a modular design that integrates input rules into a structured argumentation framework using state-of-the-art computational argumentation methods. Experiments conducted on binary classification problems using various datasets from the UCI Machine Learning Repository demonstrate the effectiveness of this integration. The LLM technique excelled in producing a manageable number of if-then rules with a small number of premises while maintaining high inferential capacity for all datasets. In turn, argument-based models achieved comparable results to those derived directly from if-then rules, leveraging a concise set of rules and excelling in explainability. In summary, this paper introduces a novel approach for efficiently and automatically generating arguments and their interactions from data, addressing both scientific and applied challenges in advancing the application and deployment of argumentation systems in XAI.

This paper presents a novel framework for structured argumentation, named extended argumentative decision graph (xADG). It is an extension of argumentative decision graphs [10] built upon Dung’s abstract argumentation graphs. The xADG framework allows for arguments to use boolean logic operators and multiple premises (supports) within their internal structure, resulting in more concise argumentation graphs that may be easier for users to understand. The study presents a methodology for construction of xADGs from an input decision tree and evaluates their size and predictive capacity for classification tasks of varying magnitudes. Resulting xADGs achieved strong (balanced) accuracy, kept from the input decision tree, while also reducing the average number of supports needed to reach a conclusion. The results further indicated that it is possible to construct plausibly understandable xADGs that outperform other techniques for building ADGs in terms of predictive capacity and overall size. In summary, the study suggests that xADG represents a promising framework to developing more concise argumentative models that can be used for classification tasks and knowledge discovery, acquisition, and refinement.

Dealing with uncertain, contradicting, and ambiguous information is still a central issue in Artificial Intelligence (AI). As a result, many formalisms have been proposed or adapted so as to consider non-monotonicity. A non-monotonic formalism is one that allows the retraction of previous conclusions or claims, from premises, in light of new evidence, offering some desirable flexibility when dealing with uncertainty. Among possible options, knowledge-base, non-monotonic reasoning approaches have seen their use being increased in practice. Nonetheless, only a limited number of works and researchers have performed any sort of comparison among them. This research article focuses on evaluating the inferential capacity of defeasible argumentation, a formalism particularly envisioned for modelling non-monotonic reasoning. In addition to this, fuzzy reasoning and expert systems, extended for handling non-monotonicity of reasoning, are selected and employed as baselines, due to their vast and accepted use within the AI community. Computational trust was selected as the domain of application of such models. Trust is an ill-defined construct, hence, reasoning applied to the inference of trust can be seen as non-monotonic. Inference models were designed to assign trust scalars to editors of the Wikipedia project. Scalars assigned to recognised trustworthy editors provided the basis for the analysis of the models’ inferential capacity according to evaluation metrics from the domain of computational trust. In particular, argument-based models demonstrated more robustness than those built upon the baselines despite the knowledge bases or datasets employed. This study contributes to the body of knowledge through the exploitation of defeasible argumentation and its comparison to similar approaches. It provides publicly implementations for the designed models of inference, which might be a useful aid to scholars interested in performing non-monotonic reasoning activities. It adds to previous works, empirically enhancing the generalisability of defeasible argumentation as a compelling approach to reason with quantitative data and uncertain knowledge.

Several non-monotonic formalisms exist in the field of Artificial Intelligence for reasoning under uncertainty. Many of these are deductive and knowledge-driven, and also employ procedural and semi-declarative techniques for inferential purposes. Nonetheless, limited work exist for the comparison across distinct techniques and in particular the examination of their inferential capacity. Thus, this paper focuses on a comparison of three knowledge-driven approaches employed for non-monotonic reasoning, namely expert systems, fuzzy reasoning and defeasible argumentation. A knowledge-representation and reasoning problem has been selected: modelling and assessing mental workload. This is an ill-defined construct, and its formalisation can be seen as a reasoning activity under uncertainty. An experimental work was performed by exploiting three deductive knowledge bases produced with the aid of experts in the field. These were coded into models by employing the selected techniques and were subsequently elicited with data gathered from humans. The inferences produced by these models were in turn analysed according to common metrics of evaluation in the field of mental workload, in specific validity and sensitivity. Findings suggest that the variance of the inferences of expert systems and fuzzy reasoning models was higher, highlighting poor stability. Contrarily, that of argument-based models was lower, showing a superior stability of its inferences across knowledge bases and under different system configurations. The originality of this research lies in the quantification of the impact of defeasible argumentation. It contributes to the field of logic and non-monotonic reasoning by situating defeasible argumentation among similar approaches of non-monotonic reasoning under uncertainty through a novel empirical comparison

This paper proposes a novel machine learning procedure for genome-wide association study (GWAS), named LightGWAS. It is based on the LightGBM framework, in addition to being a single, resilient, autonomous and scalable solution to address common limitations of GWAS implementations found in the literature. These include reliance on massive manual quality control steps and specific GWAS methods for each type of dataset morphology and size. Through this research, LightGWAS has been contrasted against PLINK2, one of the current state-of-the-art for GWAS implementations based on general linear model with support to firth regularisation. The mean differences measured upon standard classification metrics, extracted via quantitative empirical tests through k-fold cross-validation technique, indicated that LightGWAS outperforms PLINK2 for balanced, imbalanced, and high-imbalanced genomic datasets. Paired difference tests denoted statistical significance in the results extracted from the experiments with imbalanced datasets. This article contributes to the body of knowledge by presenting a potentially more efficient GWAS procedure based on nonparametric approaches. LightGWAS ensures adaptability with higher precision in the discovery of causal single-nucleotide polymorphisms, thanks to the leaf-wise tree growth algorithm offered by the state-of-the-art for gradient boosting decision trees. Control for false-positives and statistical power are automatically addressed by the model’s training process, which significative reduces human dependency during the study design.

Argumentation has recently shown appealing properties for inference under uncertainty and conflicting knowledge. However, there is a lack of studies focused on the examination of its capacity of exploiting real-world knowledge bases for performing quantitative, case-by-case inferences. This study performs an analysis of the inferential capacity of a set of argument-based models, designed by a human reasoner, for the problem of trust assessment. Precisely, these models are exploited using data from Wikipedia, and are aimed at inferring the trustworthiness of its editors. A comparison against non-deductive approaches revealed that these models were superior according to values inferred to recognised trustworthy editors. This research contributes to the field of argumentation by employing a replicable modular design which is suitable for modelling reasoning under uncertainty applied to distinct real-world domains.

The ultimate goal of Explainable Artificial Intelligence is to build models that possess both high accuracy and degree of explainability. Understanding the inferences of such models can be seen as a process that discloses the relationships between their input and output. These relationships can be represented as a set of inference rules which are usually not explicit within a model. Scholars have proposed several methods for extracting rules from data-driven machine-learned models. However, limited work exist on their comparison. This study proposes a novel comparative approach to evaluate and compare the rulesets produced by four post-hoc rule extractors by employing six quantitative metrics. Findings demonstrate that these metrics can actually help identify superior methods over the others thus are capable of successfully modelling distinctively aspects of explainability.

Knowledge-representation and reasoning methods have been extensively researched within Artificial Intelligence. Among these, argumentation has emerged as an ideal paradigm for inference under uncertainty with conflicting knowledge. Its value has been predominantly demonstrated via analyses of the topological structure of graphs of arguments and its formal properties. However, limited research exists on the examination and comparison of its inferential capacity in real-world modelling tasks and against other knowledge-representation and non-monotonic reasoning methods. This study is focused on a novel comparison between defeasible argumentation and non-monotonic fuzzy reasoning when applied to the representation of the ill-defined construct of human mental workload and its assessment. Different argument-based and non-monotonic fuzzy reasoning models have been designed considering knowledge-bases of incremental complexity containing uncertain and conflicting information provided by a human reasoner. Findings showed how their inferences have a moderate convergent and face validity when compared respectively to those of an existing baseline instrument for mental workload assessment, and to a perception of mental workload self-reported by human participants. This confirmed how these models also reasonably represent the construct under consideration. Furthermore, argument-based models had on average a lower mean squared error against the self-reported perception of mental workload when compared to fuzzy-reasoning models and the baseline instrument. The contribution of this research is to provide scholars, interested in formalisms on knowledge-representation and non-monotonic reasoning, with a novel approach for empirically comparing their inferential capacity.

@inproceedings{DBLP:conf/aics/AmbrozioLR20, author = {Bruno Ambrozio and Luca Longo and Lucas Rizzo}, editor = {Luca Longo and Lucas Rizzo and Elizabeth Hunter and Arjun Pakrashi}, title = {LightGWAS: {A} Novel Machine Learning Procedure for Genome-Wide Association Study}, booktitle = {Proceedings of The 28th Irish Conference on Artificial Intelligence and Cognitive Science, Dublin, Republic of Ireland, December 7-8, 2020}, series = {{CEUR} Workshop Proceedings}, volume = {2771}, pages = {25--36}, publisher = {CEUR-WS.org}, year = {2020}, url = {http://ceur-ws.org/Vol-2771/AICS2020\_paper\_12.pdf} }

@article{RIZZO2023100547, title = {ArgFrame: A multi-layer, web, argument-based framework for quantitative reasoning}, journal = {Software Impacts}, pages = {100547}, year = {2023}, issn = {2665-9638}, doi = {https://doi.org/10.1016/j.simpa.2023.100547}, url = {https://www.sciencedirect.com/science/article/pii/S2665963823000842}, author = {Lucas Rizzo}, keywords = {Defeasible argumentation, Automated reasoning, Knowledge-based systems, Dung semantics, Data analysis} }

@article{RIZZO2023537, title = {Comparing and extending the use of defeasible argumentation with quantitative data in real-world contexts}, journal = {Information Fusion}, volume = {89}, pages = {537-566}, year = {2022}, issn = {1566-2535}, doi = {https://doi.org/10.1016/j.inffus.2022.08.025}, url = {https://www.sciencedirect.com/science/article/pii/S1566253522001245}, author = {Lucas Rizzo and Luca Longo}, keywords = {Defeasible argumentation, Knowledge-based systems, Non-monotonic reasoning, Argumentation theory, Fuzzy logic, Expert systems, Computational trust} }

@mastersthesis{Raja2019, author = {N.N. Raja}, title = {Measuring Subjective Mental Workload Through User Interactions on Web-Based Tasks}, year = {2019}, school = {Technological University Dublin} }

@mastersthesis{Das2019, author = {S. Das}, title = {An Investigation to Observe How Mental Workload Imposed by Web-Based Tasks Designed Around Images and Graphics Change When Compared to the Mental Workload Imposed by Their Text-Only Counterparts}, year = {2019}, school = {Technological University Dublin} }

@mastersthesis{Kirwan2019, author = {R. Kirwan}, title = {Comparing Defeasible Argumentation and Non-Monotonic Fuzzy Reasoning Methods for a Computational Trust Problem with Wikipedia}, year = {2019}, school = {Technological University Dublin} }

@mastersthesis{Byrne2020, author = {K. Byrne}, title = {A Discrimination Aware Model to Predict Childhood Literacy Levels}, year = {2020}, school = {Technological University Dublin}, doi = {10.21427/cmtz-1m13} }

@mastersthesis{SicardiRosell2020, author = {I.L. Sicardi Rosell}, title = {Detection of Pathological HFO Using Supervised Machine Learning and iEEG Data}, year = {2020}, school = {Technological University Dublin}, doi = {10.21427/ppgz-e055} }

@mastersthesis{Ambrozio2020, author = {B. Ambrozio}, title = {LightGWAS: A Novel Genome-Wide Association Study Procedure}, year = {2020}, school = {Technological University Dublin}, doi = {10.21427/ngzh-xw62} }

@mastersthesis{Kondaveeti2021, author = {H.C. Kondaveeti}, title = {Static Analysis: Ensembling Model for Android Malware Detection Using Both API Calls and Permission Features}, year = {2021}, school = {Technological University Dublin} }

@mastersthesis{OBrien2021, author = {D. O'Brien}, title = {Evaluating the Performance of Smart Contracts on Ethereum Clients}, year = {2021}, school = {Technological University Dublin}, }

@mastersthesis{Cai2024, author = {B. Cai}, title = {Boosted Building Processes for Learning Argumentation Graphs from Data for Enhanced Interpretability and Accuracy}, year = {2024}, school = {Technological University Dublin} }

@Article{make6030101, AUTHOR = {Rizzo, Lucas and Verda, Damiano and Berretta, Serena and Longo, Luca}, TITLE = {A Novel Integration of Data-Driven Rule Generation and Computational Argumentation for Enhanced Explainable AI}, JOURNAL = {Machine Learning and Knowledge Extraction}, VOLUME = {6}, YEAR = {2024}, NUMBER = {3}, PAGES = {2049--2073}, URL = {https://www.mdpi.com/2504-4990/6/3/101}, ISSN = {2504-4990}, DOI = {10.3390/make6030101} }

@InProceedings{10.1007/978-3-031-44070-0_20, author="Rizzo, Lucas", editor="Longo, Luca", title="A Novel Structured Argumentation Framework for Improved Explainability of Classification Tasks", booktitle="Explainable Artificial Intelligence", year="2023", publisher="Springer Nature Switzerland", address="Cham", pages="399--414", isbn="978-3-031-44070-0" }

@inproceedings{DBLP:conf/aics/RizzoDL20, author = {Lucas Rizzo and Pierpaolo Dondio and Luca Longo}, editor = {Luca Longo and Lucas Rizzo and Elizabeth Hunter and Arjun Pakrashi}, title = {Exploring the potential of defeasible argumentation for quantitative inferences in real-world contexts: An assessment of computational trust}, booktitle = {Proceedings of The 28th Irish Conference on Artificial Intelligence and Cognitive Science, Dublin, Republic of Ireland, December 7-8, 2020}, series = {{CEUR} Workshop Proceedings}, volume = {2771}, pages = {37--48}, publisher = {CEUR-WS.org}, year = {2020}, url = {http://ceur-ws.org/Vol-2771/AICS2020\_paper\_13.pdf} }

@inproceedings{DBLP:conf/aics/ViloneRL20, author = {Giulia Vilone and Lucas Rizzo and Luca Longo}, editor = {Luca Longo and Lucas Rizzo and Elizabeth Hunter and Arjun Pakrashi}, title = {A comparative analysis of rule-based, model-agnostic methods for explainable artificial intelligence}, booktitle = {Proceedings of The 28th Irish Conference on Artificial Intelligence and Cognitive Science, Dublin, Republic of Ireland, December 7-8, 2020}, series = {{CEUR} Workshop Proceedings}, volume = {2771}, pages = {85--96}, publisher = {CEUR-WS.org}, year = {2020}, url = {http://ceur-ws.org/Vol-2771/AICS2020\_paper\_33.pdf} }

@article{rizzo2020empirical, title = "An empirical evaluation of the inferential capacity of defeasible argumentation, non-monotonic fuzzy reasoning and expert systems", journal = "Expert Systems with Applications", pages = "(in press)", year = "2020", issn = "0957-4174", doi = "https://doi.org/10.1016/j.eswa.2020.113220", url = "http://www.sciencedirect.com/science/article/pii/S0957417420300464", author = "Lucas Rizzo and Luca Longo", keywords = "Defeasible Argumentation, Argumentation Theory, Explainable Artificial Intelligence, Non-monotonic Reasoning, Fuzzy Logic, Expert Systems, Mental Workload", }

@article{LONGO2021106514, title = {Examining the modelling capabilities of defeasible argumentation and non-monotonic fuzzy reasoning}, journal = {Knowledge-Based Systems}, volume = {211}, pages = {106514}, year = {2021}, issn = {0950-7051}, doi = {https://doi.org/10.1016/j.knosys.2020.106514}, author = {Luca Longo and Lucas Rizzo and Pierpaolo Dondio}, keywords = {Defeasible reasoning, Non-monotonic reasoning, Fuzzy logic, Argumentation, Empirical research, Knowledge-representation, Mental workload}, }

Inferences through knowledge driven approaches have been researched extensively in the field of Artificial Intelligence. Among such approaches argumentation theory has recently shown appealing properties for inference under uncertainty and conflicting evidence. Nonetheless, there is a lack of studies which examine its inferential capacity over other quantitative theories of reasoning under uncertainty with real-world knowledge-bases. This study is focused on a comparison between argumentation theory and non-monotonic fuzzy reasoning when applied to modeling the construct of human mental workload (MWL). Different argument-based and non-monotonic fuzzy reasoning models, aimed at inferring the MWL imposed by a selection of learning tasks, in a third-level context, have been designed. These models are built upon knowledge-bases that contain uncertain and conflicting evidence provided by human experts. An analysis of the convergent and face validity of such models has been performed. Results suggest a superior inferential capacity of argument-based models over fuzzy reasoning-based models.

@inproceedings{rizzo2019, title={Inferential models of mental Workload with defeasible argumentation and non-monotonic fuzzy reasoning: a comparative study}, author={Rizzo, Lucas and Longo, Luca}, year = {2019}, pages = {11--26}, booktitle= {2nd Workshop on Advances In Argumentation In Artificial Intelligence}, }

Defeasible argumentation has advanced as a solid theoretical research discipline for inference under uncertainty. Scholars have predominantly focused on the construction of argument-based models for demonstrating non-monotonic reasoning adopting the notions of arguments and conflicts. However, they have marginally attempted to examine the degree of explainability that this approach can offer to explain inferences to humans in real-world applications. Model explanations are extremely important in areas such as medical diagnosis because they can increase human trustworthiness towards automatic inferences. In this research, the inferential processes of defeasible argumentation and non-monotonic fuzzy reasoning are meticulously described, exploited and qualitatively compared. A number of properties have been selected for such a comparison including understandability, simulatability, algorithmic transparency, post-hoc interpretability, computational complexity and extensibility. Findings show how defeasible argumentation can lead to the construction of inferential non-monotonic models with a higher degree of explainability compared to those built with fuzzy reasoning.

@inproceedings{rizzo2018qualitative, title={A Qualitative Investigation of the Degree of Explainability of Defeasible Argumentation and Non-monotonic Fuzzy Reasoning}, author={Rizzo, Lucas and Longo, Luca}, year = {2018}, pages = {138--149}, booktitle= {26th AIAI Irish Conference on Artificial Intelligence and Cognitive Science}, }

Inferences through knowledge driven approaches have been researched extensively in the field of Artificial Intelligence. Among such approaches Argumentation Theory has recently achieved promising results as a solid theoretical research discipline. Nonetheless, there is a lack of studies which examine its impact over other quantitative theories of reasoning under uncertainty. In this study, computational argumentation theory and non-monotonic fuzzy reasoning are selected for comparison and evaluated on the domain of biological markers, or biomarkers. This application has a potential capacity to identify prognostic and diagnostic indicators in health-care. However, the relationship between biomarkers and relevant clinical endpoints, such as survival, is not a clear one. Different argument-based models and non-monotonic fuzzy reasoning models, aimed at inferring survival using biomakers information, have been designed using an extensive knowledge base from an expert in the field. An analysis of the true positive and false positive rate of such models has been performed. Results indicate a superior inferential capacity of survival by argument-based models.

@InProceedings{RizzoML18, author="Rizzo, Lucas and Majnaric, Ljiljana and Longo, Luca", editor="Ghidini, Chiara and Magnini, Bernardo and Passerini, Andrea and Traverso, Paolo", title="A Comparative Study of Defeasible Argumentation and Non-monotonic Fuzzy Reasoning for Elderly Survival Prediction Using Biomarkers", booktitle="AI*IA 2018 -- Advances in Artificial Intelligence", year="2018", publisher="Springer International Publishing", address="Cham", pages="197--209", }

Research on the discovery, classification and validation of biological markers, or biomarkers, have grown extensively in the last decades. Newfound and correctly validated biomarkers have great potential as prognostic and diagnostic indicators, but present a complex relationship with pertinent endpoints such as survival or other diseases manifestations. This research proposes the use of computational argumentation theory as a starting point for the resolution of this problem for cases in which a large amount of data is unavailable. A knowledge-base containing 51 different biomarkers and their association with mortality risks in elderly was provided by a clinician. It was applied for the construction of several argument-based models capable of inferring survival or not. The prediction accuracy and sensitivity of these models were investigated, showing how these are in line with inductive classification using decision trees with limited data.

@InProceedings{rizzo2018investigation, author="Rizzo, Lucas and Majnaric, Ljiljana and Dondio, Pierpaolo and Longo, Luca", editor="Iliadis, Lazaros and Maglogiannis, Ilias and Plagianakos, Vassilis", title="An Investigation of Argumentation Theory for the Prediction of Survival in Elderly Using Biomarkers", booktitle="Artificial Intelligence Applications and Innovations", year="2018", publisher="Springer International Publishing", address="Cham", pages="385--397", }

In the last few decades several fields have made use of the construct of human mental workload (MWL) for system and task design as well as for assessing human performance. Despite this interest, MWL remains a nebulous concept with multiple definitions and measurement techniques. State-of-the-art models of MWL are usually ad-hoc, considering different pools of pieces of evidence aggregated with different inference strategies. In this paper the aim is to deploy a rule-based expert system as a more structured approach to model and infer MWL. This expert system is built upon a knowledge-base of an expert and translates into computable rules. Different heuristics for aggregating these rules are proposed and they are elicited using inputs gathered in an user study involving humans performing web-based tasks. The inferential capacity of the expert system, using the proposed heuristics, is compared against the one of two ad-hoc models, commonly used in psychology: the NASA-Task Load Index and the Workload Profile assessment technique. In detail, the inferential capacity is assessed by a quantification of two properties commonly used in psychological measurement: sensitivity and validity. Results show how some of the designed heuristics can over perform the baseline instruments suggesting that MWL modelling using expert system is a promising avenue worthy of further investigation.

@inproceedings{Rizzo2016, author={Rizzo, Lucas and Dondio, Pierpaolo and Delany, Sarah Jane and Longo, Luca}, editor={Iliadis, Lazaros and Maglogiannis, Ilias}, title={Modeling Mental Workload Via Rule-Based Expert System: A Comparison with NASA-TLX and Workload Profile}, bookTitle={Artificial Intelligence Applications and Innovations: 12th IFIP WG 12.5 International Conference and Workshops, AIAI 2016, Thessaloniki, Greece, September 16-18, 2016, Proceedings}, year={2016}, publisher={Springer International Publishing}, address={Cham}, pages={215--229} }

This work proposes a heuristic approach for the role assignment problem in wireless sensor networks based on the classical problem of vertex coloring in graph theory. Functions or roles define activities to be performed by sensor nodes. In the literature, there are several algorithms for the role assignment in sensor networks and in this work we present a different strategy based on the vertex coloring problem. There is a strong relationship between them, allowing us to use the vertex coloring as the start point in the solution of the role assignment problem. Such approach leads to good solutions when we consider various simultaneous events, differently from other proposals that do not consider this scenario. Even in the case of a high number of events, our algorithm remains efficient, maintaining the network connectivity, keeping the rate of non-sensed events and preserving the low energy consumption. The proposed solution tries to save the most important sensors in the network, with respect to their roles, increasing their lifetime.

@INPROCEEDINGS{rizzorole2013, author={L. M. {Rizzo} and S. {Urrutia} and A. A. F. {Loureiro}}, booktitle={2013 IEEE Seventh International Symposium on Service-Oriented System Engineering}, title={Role Assignment in Wireless Sensor Networks Based on Vertex Coloring}, year={2013}, pages={537-545},} }

This paper describes a GRASP heuristic to the recently introduced Extended Car Sequencing Problem. A constructive heuristic is developed based on an heuristic for the classical Car Sequencing Problem. The local search procedure uses a very simple neighborhood easy to be evaluated. Computational results on instances from the CSPLib's library verify the efficiency of the method in comparison with the literature.

@INPROCEEDINGS{rizzo2011, author = {L. M. Rizzo and S. Urrutia}, title = {A GRASP Heuristic to the Extended Car Sequencing Problem}, booktitle = {In proceedings of the 5th Multidisciplinary International Conference on Scheduling : Theory and Applications (MISTA 2011), 9-11 August 2011, Phoenix, Arizona, USA}, year = {2011}, editor = {J. Fowler and G. Kendall and B. McCollum}, pages = {366--373}}

@INPROCEEDINGS{rizzorole2013, author={L. M. {Rizzo} and S. {Urrutia} and A. A. F. {Loureiro}}, booktitle={2013 IEEE Seventh International Symposium on Service-Oriented System Engineering}, title={Role Assignment in Wireless Sensor Networks Based on Vertex Coloring}, year={2013}, pages={537-545},} }

Limited work exists for the comparison across distinct knowledge-based approaches in Artificial Intelligence (AI) for non-monotonic reasoning, and in particular for the examination of their inferential and explanatory capacity. Non-monotonicity, or defeasibility, allows the retraction of a conclusion in the light of new information. It is a similar pattern to human reasoning, which draws conclusions in the absence of information, but allows them to be corrected once new pieces of evidence arise. Thus, this thesis focuses on a comparison of three approaches in AI for implementation of non-monotonic reasoning models of inference, namely: expert systems, fuzzy reasoning and defeasible argumentation. Three applications from the fields of decision-making in healthcare and knowledge representation and reasoning were selected from real-world contexts for evaluation: human mental workload modelling, computational trust modelling, and mortality occurrence modelling with biomarkers. The link between these applications comes from their presumptively non-monotonic nature. They present incomplete, ambiguous and retractable pieces of evidence. Hence, reasoning applied to them is likely suitable for being modelled by non-monotonic reasoning systems.

An experiment was performed by exploiting six deductive knowledge bases produced with the aid of domain experts. These were coded into models built upon the selected reasoning approaches and were subsequently elicited with real-world data. The numerical inferences produced by these models were analysed according to common metrics of evaluation for each field of application. For the examination of explanatory capacity, properties such as understandability, extensibility, and post-hoc interpretability were meticulously described and qualitatively compared. Findings suggest that the variance of the inferences produced by expert systems and fuzzy reasoning models was higher, highlighting poor stability. In contrast, the variance of argument-based models was lower, showing a superior stability of its inferences across different system configurations. In addition, when compared in a context with large amounts of conflicting information, defeasible argumentation exhibited a stronger potential for conflict resolution, while presenting robust inferences. An in-depth discussion of the explanatory capacity showed how defeasible argumentation can lead to the construction of non-monotonic models with appealing properties of explainability, compared to those built with expert systems and fuzzy reasoning. The originality of this research lies in the quantification of the impact of defeasible argumentation. It illustrates the construction of an extensive number of non-monotonic reasoning models through a modular design. In addition, it exemplifies how these models can be exploited for performing non-monotonic reasoning and producing quantitative inferences in real-world applications. It contributes to the field of non-monotonic reasoning by situating defeasible argumentation among similar approaches through a novel empirical comparison.

@phdthesis{rizzo2020thesis, title={Evaluating the Impact of Defeasible Argumentation as a Modelling Technique for Reasoning under Uncertainty}, author={Rizzo, Lucas}, year={2020}, school = {Technological University Dublin} }

Devido à tecnologia disponível atualmente viabilizou-se a utilização de amplas e densas Redes de Sensores sem Fio (RSSFs). Um dos problemas inerentes deste tipo de rede é a otimização na distribuição das funções empregadas à cada sensor. Basicamente qualquer sensor em uma RSSF requer algum tipo de auto-configuração, onde são atribuídos papéis a cada um deles sem que haja alguma intervenção manual. Este trabalho propõe e implementa uma abordagem heurística, com versões centralizadas e distribuídas, para o problema de atribuição de papéis. Tal abordagem é realizada através de uma relação com o problema de coloração de vértices e implementada utilizando-se o simulador de redes Sinalgo. São considerados a geração de eventos no campo de sensoriamento, e têm-se como objetivo a economia de energia na detecção e notificação destes eventos. Para economizar energia a rede deve ser capaz de utilizar o menor número de sensores possíveis na detecção de eventos além de encaminhar as informações obtidas pela melhor rota possível com o nó sorvedouro. São apresentados resultados computacionais que comprovam a eficiência do método abordado através de métricas como tempo de execução, porcentagem de eventos não sensoreados e número de mensagens enviadas por evento.

@mastersthesis{rizzomaster, author = {Rizzo, Lucas}, title = {Atribuição de papéis em redes de sensores sem fio baseada em coloração de vértices}, school = {Universidade Federal de Minas Gerais}, year = {2013} }

Este artigo descreve uma heurística GRASP para o recentemente introduzido Problema Estendido de Sequenciamento de Carros. Uma heurística construtiva é desenvolvida com base em uma heurística para o Problema Clássico de Sequenciamento de Carros. O procedimento de busca local utiliza uma vizinhança simples e de fácil avaliação. Resultados computacionais sobre as instâncias da biblioteca CSPLIB verificam a eficiência do método em comparação com a literatura.

@INPROCEEDINGS{rizzograsp2013, author={L. M. {Rizzo} and S. {Urrutia}, booktitle={XLIII Simpósio Brasileiro de Pesquisa Operational}, title={Uma heurística grasp para o problema estendido de sequenciamento de carros}, year={2011}, pages={1745--1752}}

Mental workload (MWL) is an imprecise construct, with distinct definitions and no predominant measurement technique. It can be intuitively seen as the amount of mental activity devoted to a certain task over time. Several approaches have been proposed in the literature for the modelling and assessment of MWL. In this paper, data related to two sets of tasks performed by participants under different conditions is reported. This data was gathered from different sets of questionnaires answered by these participants. These questionnaires were aimed at assessing the features believed by domain experts to influence overall mental workload. In total, 872 records are reported, each representing the answers given by a user after performing a task. On the one hand, collected data might support machine learning researchers interested in using predictive analytics for the assessment of mental workload. On the other hand, data, if exploited by a set of rules/arguments (as in [3]), may serve as knowledge-bases for researchers in the field of knowledge-based systems and automated reasoning. Lastly, data might serve as a source of information for mental workload designers interested in investigating the features reported here for mental workload modelling. This article was co-submitted from a research journal "An empirical evaluation of the inferential capacity of defeasible argumentation, non-monotonic fuzzy reasoning and expert systems" [3]. The reader is referred to it for the interpretation of the data.

@article{RIZZO2020data, title = "Self-reported data for mental workload modelling in human-computer interaction and third-level education.", journal = "Data in Brief", pages = "(in press)", year = "2020", issn = "2352-3409", doi = "https://doi.org/10.1016/j.dib.2020.105433", author = "Lucas Rizzo and Luca Longo", keywords = "Knowledge-based systems, Fuzzy reasoning, Expert systems, Mental workload, Automated reasoning, Argumentation theory", }