Leonardo Bezerra
My thesis on fundamental artificial intelligence (AI) resulted on three journal and nine top-tier conference papers, being seminal to my Lectureships in Brazil, where I worked between 2016 and 2023 on data science (DS), AI, and their impact on socially relevant problems.
-
I proposed and collaborated with applied data science projects with both the public and private sectors. Partners included the Brazilian Judicial Branch, Ministry of Education, and Public Safety Secretariats, as well as retail and energy companies, e.g. Neoenergia, the Brazilian subsidiary of the Spanish utility company Iberdrola.
-
I supervised graduate students on theses involving deep and automated machine learning, as well as the intersection of multi-objective optimization with other CI domains, such as multi-dimensional visualization and dynamic optimization. The quality of those theses is evidenced by the 10 conference and two journal [papers](#publications) published as a result, including an EMO best paper award nomination.
-
I developed a deep concern for socially relevant problems, having for instance assisted in the fight against the COVID-19 pandemic through science publication and communication. Importantly, my Lectureships led to contributions to policymaking in public health, law (enforcement), and higher education, e.g. the revised and updated Brazilian Standard Classification of Education for census, evaluation, and regulation.
I relocated to the University of Stirling, UK in 2023, deepening my research focus on assessing and addressing the social impacts of AI systems.
-
I have joined a research consortium led by the University of Glasgow to propose methodologies and workbenches that enable stakeholders with diverse backgrounds to participate in AI harm auditing. The consortium was awarded £3.5 million by the Responsible AI Keystone Projects call funded by UKRI. The project started in May 2024 and will last for 48 months. Importantly, the project currently has over 20 external partners who will provide case studies and help devise and evaluate our AI harm auditing workbenches and methodologies.
-
I am applying for funding to promote a society where AI systems are fair, accountable, inclusive, responsible, transparent, ethical, carbon-neutral, and human-centred (FAIRTECH) by design. To achieve this, I will:
- structure an Observatory of the Potentially Incurred Social Damage from Artificial Intelligence (PISD.ai), which will survey current (and future) relevant real-world examples where the lack of appropriate AI regulation (potentially) incurs significant social damage. A prototypical example is the role of social media recommendation algorithms in disinformation and its impact on democracy, human rights, public health, and climate change, and how this impact can be further worsened by generative AI.
- survey the existing AI regulation proposals from major Western and non-Western democratic players, as well as the legal, economical, and technical limitations for appropriate regulation. Major player examples include the European Union, the United Kingdom, the United States of America, Brazil, Chile, and Japan.
- propose validation frameworks to assist AI regulators and policymakers in their proposals and/or revisions. Examples include the appropriate definition of artificial intelligence systems to reduce the risk of overregulation.
- formulate an educational framework to promote a FAIRTECH by design culture, targeting key players in government, academia, industry, and civil society more broadly. Relevant partners that have already demonstrated interest include the Brazilian Computing Society, Brazilian Ministry of Human Rights and Citizenship, and Toyota Motor Europe.
Contents
- News
- Places & People
- History
- Projects and funding
- Supervision
- Key and relevant publications
- Science communication, networking, and community
News
Spring 2024
-
Our project “PHAWM: Participatory Harm Auditing Workbenches and Methodologies” was awarded £3.5mi by Responsible AI UK to propose approaches that enable society to take part in AI auditing over the next four years, starting May 2024. Our consortium comprises 25 investigators from 7 UK universities, namely Glasgow, Strathclyde, Stirling, Edinburgh, Sheffield, York, and King’s College London, as well as 23 external partners.
-
I was invited by Responsible AI UK to attend the “RAi UK Integrator Sandpit”, which will take place from June 11th to 13th, 2024 at the De Vere Latimer Estate, Chesham, United Kingdom.
-
I was nominated Outstanding Early Career Researcher by the Research Culture Awards of the University of Stirling, Stirling, UK.
-
I was invited by Dr. Annalisa Ricardi to give a talk at the University of Strathclyde, Glasgow, UK. The talk will be hosted by the Aerospace Centre of Excellence on May 22nd, entitled “FAIRTECH by design: assessing and addressing the social impacts of artificial intelligence systems”.
-
I was invited by Prof. Holger Hoos to give a talk at RWTH Aachen University, Aachen, DE. The talk was hosted by the Chair for Artificial Intelligence Methodology (AIM) on April 17th, entitled “FAIRTECH by design: assessing and addressing the social impacts of artificial intelligence systems”.
-
I attended the “BeMCO: Benchmarking in Multi-Criteria Optimization” workshop, organized by the Benchmarking Network, which took place from April 15th to 19th, 2024 at the Lorentz Center@Oort, Leiden, The Netherlands.
-
I was invited by Dr. Gianpiero Francesca to give a talk at Toyota Motor Europe, Brussels, BE. The talk was hosted by the Advanced Technology R&D Division on April 15th, entitled “FAIRTECH by design: assessing and addressing the social impacts of artificial intelligence systems”.
-
I was invited by Prof. Thomas Stützle, Prof. Mauro Birattari, and Prof. Marco Dorigo to give a talk at Université Libre de Bruxelles, Brussels, BE. The talk was hosted by the IRIDIA Artificial Intelligence Laboratory on April 11th, entitled “FAIRTECH by design: assessing and addressing the social impacts of artificial intelligence systems”.
Winter 2024
-
I was invited by Dr. Anthony O’Hare to give a public lecture at the University of Stirling, Stirling, UK. The talk was hosted by the Computing Science and Mathematics Division (CSM) on March 18th, entitled “Promoting and sustaining accountability in artificial intelligence applications”.
-
I was invited by Prof. Kenneth Benoit to give a talk at the London School of Economical and Political Sciences, London, UK. The talk was hosted by the Data Science Institute (DSI) on February 26th, entitled “Promoting and sustaining accountability in artificial intelligence applications”.
-
I was invited by Prof. Simone Stumpf to give a talk at the University of Glasgow, Glasgow, UK. The talk was hosted by the Information, Data, and Analysis Section (IDA) on February 1st, entitled “Promoting and sustaining accountability in artificial intelligence applications”.
-
I was invited by Prof. Federica Sarro to give a talk at University College London, London, UK. The talk was hosted by the Software Systems Engineering Group (SSE) on January 24th, entitled “Promoting and sustaining accountability in artificial intelligence applications”.
Autumn 2023
-
I was invited by the Benchmarking Network to attend the “BeMCO: Benchmarking in Multi-Criteria Optimization” workshop, which will take place from April 15th to 19th, 2024 at the Lorentz Center@Oort, Leiden, The Netherlands.
-
I was invited by Dr. Wen-Shin Lee to give a talk at the University of Stirling, Stirling, UK. The talk was hosted by the Computing Science and Mathematics Division (CSM) on November 24th, entitled “Promoting and sustaining accountability in artificial intelligence applications”.
-
I was invited by Dr. Elizabeth Wanner to give a talk at Aston University, Birmingham, UK. The talk was hosted by the Aston Centre for Artificial Intelligence Research and Applications (ACAIRA) on October 16th, entitled “Promoting and sustaining accountability in artificial intelligence applications”.
Summer 2023
- I was offered and accepted a Lecturer in Artificial Intelligence/Data Science position at the University of Stirling, Stirling, UK.
- I was offered a Senior Lecturer in Data Science position at the University of Westminster, London, UK.
- I was offered a Senior Teaching Fellow in Data Science position at the Coventry University, Coventry, UK.
- I was awarded an Alain Bensoussan fellowship from the European Research Consortium for Informatics and Mathematics (ERCIM) to work at Universidad de Malaga (UMA), Malaga, Spain.
Places & People
History
Appointments (3)
Lecturer @ University of Stirling
Assistant professor @ Federal University of Rio Grande do Norte (UFRN)
Assistant professor @ Federal University of Paraíba (UFPB)
Awards (6)
Responsible AI UK Keystone Project @ Responsible AI UK
Outstanding Early Career Researcher nomination @ Research Culture Awards, University of Stirling
Alain Bensoussan fellowship @ European Research Consortium for Informatics and Mathematics (ERCIM)
Best paper award nomination @ Evolutionary Multi-Criterion Optimization (EMO)
Ph.D. degree in Engineering and Technology @ Université Libre de Bruxelles (ULB)
F.R.I.A doctoral fellowship @ Fonds de la Recherche Scientifique (FNRS)
Best paper award @ Brazilian Symposium on Augmented and Virtual Reality (SVR)
Languages (5)
Projects and funding
Financial details about the projects are provided in my CV.
Proponent (5)
PHAWM: Participatory harm auditing workbenches and methodologies @ Responsible AI UK
Technological innovation cell @ Iberdrola Neoenergia COSERN
Applied research and human resource education in hardware technologies for artificial intelligence @ Huawei Telecommunications in Brazil
Revision and update of the Brazilian Standard Classification of Education (CINE Brasil 2018) @ UNESCO & Brazilian Ministry of Education – INEP
Generalization of metaheuristics for optimization problems with three or more objectives @ Fonds de la Recherche Scientifique (FNRS)
Collaborator (5)
Information technology postgraduate apprenticeship (class of 2024) @ 5th Region Federal Regional Court (TRF5)
Information technology postgraduate apprenticeship (class of 2023) @ 5th Region Federal Regional Court (TRF5)
SmartMetropolis @ Multiple local and national government branches
Combinatorial optimization: metaheuristics and exact methods (COMEX) @ Belgian Federal Science Policy Office (BELSPO)
Algorithms for mobile robot path planning considering multiple objectives @ National Council for Scientific and Technological Development (CNPq)
Supervision
Design configuration for the MMAS algorithm applied to the travelling salesman problem with dynamic demands @ Federal Center of Technological Education of Minas Gerais (CEFET-MG)
A case study on customer segmentation of a supermarket chain @ Federal University of Rio Grande do Norte (UFRN)
Sales forecasting for a supermarket chain in Natal, Brazil: an empirical assessment @ Federal University of Rio Grande do Norte (UFRN)
Assessing irace for automated machine and deep learning in computer vision @ Federal University of Rio Grande do Norte (UFRN)
Predspot: predicting crime hotspots with machine learning @ Federal University of Rio Grande do Norte (UFRN)
A metaheuristic approach to the high school timetabling problem at IFRN @ Federal University of Rio Grande do Norte (UFRN)
Key and relevant publications
An exhaustive publication list with full author description is provided in my CV.
Journals (5)
A computational study on ant colony optimization for the traveling salesman problem with dynamic demands
This paper was the main contribution from the first Ph.D. thesis I co-supervised, and demonstrates how multi-objective and dynamic optimization intersect. The relevance of this paper is evidenced by its best paper award nomination at the EMO 2019 conference, where a preliminary version of the journal paper was first published. In addition, this paper is a concrete example of how I bridge different research topics into multi-disciplinary work.
Comparing community mobility reduction between first and second COVID-19 waves
This paper was the main contribution of my efforts in science publication and communication to assist in the fight against the COVID-19 pandemic. Indeed, the first author of this paper is one of the undergraduate students that I helped mobilize in those initiatives. The relevance of this paper is evidenced by the number of different continents and COVID-19 waves included in the assessment. In addition, this paper is a concrete example of how I use computational intelligence in the context of socially relevant problems.
Automatically designing state-of-the-art multi-and many-objective evolutionary algorithms
A large-scale experimental evaluation of high-performing multi-and many-objective evolutionary algorithms
Automatic component-wise design of multiobjective evolutionary algorithms
These papers comprise the contributions of my Ph.D. thesis, having been accepted for publication prior to my defense or shortly after. Their relevance is evidenced by their ongoing impact on the evolutionary computation community, one of the most important in the context of CI, and by the rigorous journals where they were published. More importantly, these papers demonstrate how I am able to plan and deliver on a research project. In detail, each paper meets an specific objective of my thesis proposal, incrementally achieving the general objective of the project.
Conference papers (11)
High school timetabling at a federal educational institute in Brazil
Retail sales forecasting for a Brazilian supermarket chain: an empirical assessment
Supermarket customer segmentation: a case study in a large Brazilian retail chain
Time-series features for predictive policing
Towards a crime hotspot detection framework for patrol planning
These papers comprise the contributions of data science M.Sc. theses I (co-)supervised in partnership with public and private institutions. The relevance of these papers is evidenced by the socially relevant scenarios they address. In detail, the first paper focuses on the Brazilian Federal Network of Vocational, Scientific and Technological Education, which provides education to over two million students, with over half of the students that declared income, gender, and ethnicity coming from low income families, being women, and self-declaring as non-white. The remainder 2022 papers use AI techniques to model different business processes in the 3rd largest retail supermarket chain in the Northeast of Brazil, and is instrumental to assess the impact of the COVID-19 pandemic in the industry. Finally, the 2018 papers address predictive policing to assist the local government in the forecasting of criminal occurrences.
Evaluating anytime performance on NAS-Bench-101
iSklearn: automated machine learning with irace
Comparing contextual embeddings for semantic textual similarity in Portuguese
These papers are the contributions of M.Sc. theses I supervised in deep and automated machine learning. The relevance of these papers is evidenced by the state-of-the-art techniques that were employed. In addition, the application domains considered are among the most relevant that use unstructured data, namely computer vision, natural language processing, and time series forecasting. Importantly, these papers demonstrate that I understand the technological complexity of current state-of-the-art AI models, their potential impact on society, and therefore their need for accountability.
Revisiting Pareto-optimal multi-and many-objective reference fronts for continuous optimization
Archiver effects on the performance of state-of-the-art multi-and many-objective evolutionary algorithms
An empirical assessment of the properties of inverted generational distance on multi-and many-objective optimization
These papers comprise follow-up works on my Ph.D. thesis. The relevance of these papers is evidenced by their ongoing impact on the evolutionary computation community, as well as the conferences where they were published, which are among the top-tier venues in their field. More importantly, these papers are a concrete example that the work I conducted in my Ph.D. was seminal to relevant future work. In addition, they demonstrate that I understand that seeking autonomy as an independent researcher does not mean discontinuing previous research.
Science communication, networking, and community
COVID-19
During the climax of the first wave of the COVID-19 pandemic in 2020, I proposed and collaborated with several science communication efforts to help disseminate the guidelines from the World Health Organization (WHO) and counter the disinformation campaigns that were strongly affecting the social distance adherence in Brazil.
Communication and networking
FAIRTECH by design: assessing and addressing the social impacts of artificial intelligence
@ Chair for Artificial Intelligence Methodologies (AIM) - RWTH Aachen University
@ Advanced Technology R&D Division - Toyota Motor Europe
@ IRIDIA Artificial Intelligence Lab - Université Libre de Bruxelles (ULB)
Promoting and sustaining accountability in artificial intelligence applications
@ Computing Science and Mathematics (CSM) Public Lectures - University of Stirling
@ Data Science Institute (DSI) - London School of Economics and Political Sciences
@ Information, Data, and Analysis (IDA) Section - University of Glasgow
@ Software Systems Engineering (SSE) Group - University College London
@ Computing Science and Mathematics (CSM) Seminars - University of Stirling
@ Aston Centre for Artificial Intelligence Research and Applications (ACAIRA) - Aston University, Birmingham, UK
Brazilian Standard Classification of Education (CINE Brasil 2018)
@ Brazilian Ministry of Education - INEP
@ Federal University of Ceará (UFC)
@ Federal University of São Paulo (Unifesp)
@ Federal University of Rio Grande do Norte (UFRN)
@ Computing Education Workshop (WEI@CSBC)
AutoML with Python - machine learning made easy(ish) @ Python Brasil 2018
A component-wise approach to multi-objective evolutionary algorithms: from flexible frameworks to automatic design
Automated algorithm engineering
@ Federal University of Rio Grande do Sul (UFRGS)
@ Federal University of Paraíba (UFPB)
pagmo C++ scientific library @ European Space Agency (ESA)
A practical introduction to irace
@ Federal University of Rio Grande do Norte (UFRN)
@ Federal University of Rio Grande do Norte (UFRN)
@ Université Libre de Bruxelles (ULB)
Automatic Generation of Multi-Objective ACO Algorithms for the Biobjective Knapsack
@ Alpen-Adria-Universität Klagenfurt / Lakeside Labs
Community
In addition to my formal collaborations through supervision and authoring, I have also met incredible people along my research career.
- Wider network, researchers that always wholeheartedly assist me when I reach out to them.
- IRIDIAns, the whole family of researchers that graduated with my from the IRIDIA lab.
- Alumni, the amazing students I had the privilege to supervise.
Check their whereabouts (updated eventually) at the interactive map at the beginning of this CV :blush: