Erasmo Purificato

Mr. Erasmo Purificato

Otto-von-Guericke-Universität Magdeburg
+1
orcid

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 June 2022 in International Journal of Human–Computer Interaction
Verified
Journal Article
Closed access
EPErasmo Purificato
Erasmo Purificato
FLFlavio Lorenzo
Flavio Lorenzo
Francesca FallucchiFrancesca Fallucchi
Francesca Fallucchi
Ernesto William De LucaErnesto William De Luca
Ernesto William De Luca
Despite the existing skepticism about the use of automatic systems in contexts where human knowledge and experience are considered indispensable (e.g., the granting of a mortgage, the prediction of stock prices, or the detection of cancers), our work aims to show how the use of explainability and fairness techniques can lead to the growth of a domain expert’s trust and reliance on an artificial intelligence (AI) system. This article presents a system, applied to the context of loan approval processes, focusing on the two aforementioned ethical principles out of the four defined by the High-Level Expert Group on AI in the document “Ethics Guidelines for Trustworthy AI,” published in April 2019, in which the key requirements that AI systems should meet to be considered trustworthy are identified. The presented case study is realized within a proprietary framework composed of several components for supporting the user throughout the management of the whole life cycle of a machine learning model. The main approaches, consisting of providing an interpretation of the model’s outputs and monitoring the model’s decisions to detect and react to unfair behaviors, are described in more detail to compare our system within state-of-the-art related frameworks. Finally, a novel Trust & Reliance Scale is proposed for evaluating the system, and a usability test is performed to measure the user satisfaction with the effectiveness of the developed user interface; results are obtained, respectively, by the submission of the mentioned novel scale to bank domain experts and the usability questionnaire to a heterogeneous group composed of loan officers, data scientists, and researchers.

3 April 2018 in Multimedia Tools and Applications
Verified
Journal Article
Closed access
PDF file
EPErasmo Purificato
Erasmo Purificato
Antonio RinaldiAntonio Rinaldi
Antonio Rinaldi

21 February 2024 in arXiv
Verified
Preprint
Open access
PDF file
EPErasmo Purificato
Erasmo Purificato
LBLudovico Boratto
Ludovico Boratto
Ernesto William De LucaErnesto William De Luca
Ernesto William De Luca
The integration of artificial intelligence (AI) into daily life, particularly through information retrieval and recommender systems, has necessitated advanced user modeling and profiling techniques to deliver personalized experiences. These techniques aim to construct accurate user representations based on the rich amounts of data generated through interactions with these systems. This paper presents a comprehensive survey of the current state, evolution, and future directions of user modeling and profiling research. We provide a historical overview, tracing the development from early stereotype models to the latest deep learning techniques, and propose a novel taxonomy that encompasses all active topics in this research area, including recent trends. Our survey highlights the paradigm shifts towards more sophisticated user profiling methods, emphasizing implicit data collection, multi-behavior modeling, and the integration of graph data structures. We also address the critical need for privacy-preserving techniques and the push towards explainability and fairness in user modeling approaches. By examining the definitions of core terminology, we aim to clarify ambiguities and foster a clearer understanding of the field by proposing two novel encyclopedic definitions of the main terms. Furthermore, we explore the application of user modeling in various domains, such as fake news detection, cybersecurity, and personalized education. This survey serves as a comprehensive resource for researchers and practitioners, offering insights into the evolution of user modeling and profiling and guiding the development of more personalized, ethical, and effective AI systems.

18 July 2023
Verified
Conference Paper
Open access
MAMohamed Abdelrazek
Mohamed Abdelrazek
EPErasmo Purificato
Erasmo Purificato
LBLudovico Boratto
Ludovico Boratto
Ernesto William De LucaErnesto William De Luca
Ernesto William De Luca
Modern user profiling approaches capture different forms of interactions with the data, from user-item to user-user relationships. Graph Neural Networks (GNNs) have become a natural way to model these behaviours and build efficient and effective user profiles. However, each GNN-based user profiling approach has its own way of processing information, thus creating heterogeneity that does not favour the benchmarking of these techniques. To overcome this issue, we present FairUP, a framework that standardises the input needed to run three state-of-the-art GNN-based models for user profiling tasks. Moreover, given the importance that algorithmic fairness is getting in the evaluation of machine learning systems, FairUP includes two additional components to (1) analyse pre-processing and post-processing fairness and (2) mitigate the potential presence of unfairness in the original datasets through three pre-processing debiasing techniques. The framework, while extensible in multiple directions, in its first version, allows the user to conduct experiments on four real-world datasets. The source code is available at https://link.erasmopurif.com/FairUP-source-code, and the web application is available at https://link.erasmopurif.com/FairUP.

25 August 2021 in Computers
Verified
Journal Article
Open access
PDF file
EPErasmo Purificato
Erasmo Purificato
SWSabine Wehnert
Sabine Wehnert
Ernesto William De LucaErnesto William De Luca
Ernesto William De Luca
In the age of digital information, where the internet and social networks, as well as personalised systems, have become an integral part of everyone’s life, it is often challenging to be aware of the amount of data produced daily and, unfortunately, of the potential risks caused by the indiscriminate sharing of personal data. Recently, attention to privacy has grown thanks to the introduction of specific regulations such as the European GDPR. In some fields, including recommender systems, this has inevitably led to a decrease in the amount of usable data, and, occasionally, to significant degradation in performance mainly due to information no longer being attributable to specific individuals. In this article, we present a dynamic privacy-preserving approach for recommendations in an academic context. We aim to implement a personalised system capable of protecting personal data while at the same time allowing sensible and meaningful use of the available data. The proposed approach introduces several pseudonymisation procedures based on the design goals described by the European Union Agency for Cybersecurity in their guidelines, in order to dynamically transform entities (e.g., persons) and attributes (e.g., authored papers and research interests) in such a way that any user processing the data are not able to identify individuals. We present a case study using data from researchers of the Georg Eckert Institute for International Textbook Research (Brunswick, Germany). Building a knowledge graph and exploiting a Neo4j database for data management, we first generate several pseudoN-graphs, being graphs with different rates of pseudonymised persons. Then, we evaluate our approach by leveraging the graph embedding algorithm node2vec to produce recommendations through node relatedness. The recommendations provided by the graphs in different privacy-preserving scenarios are compared with those provided by the fully non-pseudonymised graph, considered as the baseline of our evaluation. The experimental results show that, despite the structural modifications to the knowledge graph structure due to the de-identification processes, applying the approach proposed in this article allows for preserving significant performance values in terms of precision.

17 July 2024 in Minds and Machines
Verified
Journal Article
Open access
PDF file
EPErasmo Purificato
Erasmo Purificato
LBLudovico Boratto
Ludovico Boratto
Ernesto William De LucaErnesto William De Luca
Ernesto William De Luca
User modeling is a key topic in many applications, mainly social networks and information retrieval systems. To assess the effectiveness of a user modeling approach, its capability to classify personal characteristics (e.g., the gender, age, or consumption grade of the users) is evaluated. Due to the fact that some of the attributes to predict are multiclass (e.g., age usually encompasses multiple ranges), assessing fairness in user modeling becomes a challenge since most of the related metrics work with binary attributes. As a workaround, the original multiclass attributes are usually binarized to meet standard fairness metrics definitions where both the target class and sensitive attribute (such as gender or age) are binary. However, this alters the original conditions, and fairness is evaluated on classes that differ from those used in the classification. In this article, we extend the definitions of four existing fairness metrics (related to disparate impact and disparate mistreatment) from binary to multiclass scenarios, considering different settings where either the target class or the sensitive attribute includes more than two groups. Our work endeavors to bridge the gap between formal definitions and real use cases in bias detection. The results of the experiments, conducted on four real-world datasets by leveraging two state-of-the-art graph neural network-based models for user modeling, show that the proposed generalization of fairness metrics can lead to a more effective and fine-grained comprehension of disadvantaged sensitive groups and, in some cases, to a better analysis of machine learning models originally deemed to be fair. The source code and the preprocessed datasets are available at the following link: https://github.com/erasmopurif/toward-responsible-fairness-analysis.