Research
Computational iconology, embodied spatial interfaces, and the survival of visual patterns
Visual pattern analysis. Embodied experience. Human-AI systems. Three axes, one conviction: scholarship and practice inform one another.
Research axes
Visual Pattern Analysis
Investigating how visual patterns carry meaning across time and media, from art historical survival in AI-generated imagery to cinematic representations of power and resistance.
Embodied Spatial Interfaces
Developing VR installations that transform abstract data into navigable space, where bodily movement generates understanding.
Human-AI Systems & Knowledge Production
Designing frameworks for human-AI collaboration and examining how knowledge flows shape innovation, from creative pedagogy to bibliometric analysis of technological asymmetries.
Current research
Active projects across 3 research axes, from early concept to publication.
Visual Pattern Analysis
Nachleben Computacional
Pathosformeln in Latent Space
Nachleben Computacional
Pathosformeln in Latent Space
Can emotional gestures that survived millennia of art history (Warburg's Pathosformeln) reappear in AI-generated images? This project investigates the computational survival (Nachleben) of gestural patterns in diffusion models, testing whether historically recurrent configurations emerge preferentially when models are prompted to represent intense emotional states.
Target: Leonardo (MIT Press)
Hypotheses
- H1: High-affective gestural configurations (Pathosformeln) appear disproportionately in generative model output when prompted for emotional representations.
- H2: Poses generated by contemporary models converge toward configurations documented in historical art (WikiArt, Web Gallery of Art).
- H3: The same Pathosformel appears in model output regardless of stylistic markers in the prompt.
- H4: Pathosformeln maintain their emotional valence (Warburg's antithetical principle) in generated images.
Methodology
Model: Stable Diffusion 1.5
Pose Extraction: MediaPipe
Beyond the Black Box
Ontogenesis of AI-Generated Images via Simondon
Beyond the Black Box
Ontogenesis of AI-Generated Images via Simondon
This paper develops a novel framework for understanding AI-generated images through Gilbert Simondon's philosophy of individuation. Combining computational experiments with artistic case studies (Refik Anadol, Sougwen Chung, Holly Herndon), we demonstrate how latent space functions as a preindividual field and reconceptualize 'hallucinations' as alternative individuations.
Status: Published
Published in: Philosophy & Technology
Methodology
Model: Stable Diffusion 2.1
- Latent Space Analysis (10,000 vectors)
- Denoising Trajectory Analysis (500 runs)
- Variation Analysis (100 images × 15 prompts)
Key Findings
- Effective dimensionality ~127 vs 16,384 nominal
- Phase transitions at 20%, 50%, 80% of denoising
- Semantic coherence 0.847 within prompt variations
Archaeology of Latent Space
Visual Patterns in Generative Model Outputs
Archaeology of Latent Space
Visual Patterns in Generative Model Outputs
Building on the Nachleben Computacional findings, this project extends computational iconology to analyze broader patterns of visual culture encoded in generative models. Depends on completion of Nachleben for methodological foundation.
Target: TBD
The Colonial Situation on Screen
Balandier, Fanon, and the Anthropology of Power in The Battle of Algiers
The Colonial Situation on Screen
Balandier, Fanon, and the Anthropology of Power in The Battle of Algiers
An anthropological analysis of Gillo Pontecorvo's The Battle of Algiers (1966) through two theoretical frameworks central to the political anthropology of colonialism: Georges Balandier's concept of the 'colonial situation' and Frantz Fanon's theory of decolonizing violence. The article establishes a dialogue with African and Afro-Brazilian studies, drawing on Kabengele Munanga's contributions on negritude and Clóvis Moura's work on resistance.
Status: Submitted
Target: Journal of Postcolonial Writing
Embodied Spatial Interfaces
Inside Sound
Embodied Navigation Through Timbral Constellations
Inside Sound
Embodied Navigation Through Timbral Constellations
A room-scale VR installation transforming spectral audio analysis into navigable three-dimensional space. Proposes 'navigational listening' as a distinct epistemic mode where physical locomotion through timbral data creates embodied understanding of acoustic properties typically inaccessible to non-expert listeners.
Status: ACAH: Accepted (Tokyo, May 2026) / ISEA: Submitted
Target: ACAH 2026 / ISEA 2026
Data Constellations
Navigating Extinction Through Embodied Spatial Memory
Data Constellations
Navigating Extinction Through Embodied Spatial Memory
A VR installation that transforms ecological loss into navigable virtual space, positioning extinct species as luminous points in an immersive constellation that viewers traverse with their bodies. Drawing on Indigenous wayfinding traditions and embodied cognition theory, the work frames navigation as an epistemological act: making knowledge through movement rather than simply accessing information.
Status: Accepted
Target: ISEA 2026 — International Symposium on Electronic Art, Dubai
Hẻm: Three Generations
VR Heritage Documentation in Vietnamese Alleys
Hẻm: Three Generations
VR Heritage Documentation in Vietnamese Alleys
An immersive VR experience documenting the endangered cultural heritage of Vietnamese hẻm (alleyways) across three generations. Combines photogrammetry, oral history, and spatial storytelling to preserve disappearing urban landscapes.
Status: Seeking Funding
Target: TBD
Cartographies of Refusal
When Colonial Archives Look Back
Cartographies of Refusal
When Colonial Archives Look Back
A room-scale VR installation transforming French Indochina colonial photographs into active agents that resist viewer engagement through algorithmically determined refusal behaviors. Positioning 50–100 colonial-era images as a navigable three-dimensional constellation, the work spatializes each photograph according to temporal, geographic, and violence indices, literalizing the concept of 'ethnographic refusal' through embodied interaction. Created and situated in Ho Chi Minh City (formerly colonial Saigon), the project enacts geographic reversal: the archive looks back from the colonized territory.
Status: Abstract accepted; full manuscript invited
Target: Arts (MDPI) — Special Issue: Performative Imaging
Human-AI Systems & Knowledge Production
Sympoietic Pedagogy
Human-AI Co-Creation in Creative Education
Sympoietic Pedagogy
Human-AI Co-Creation in Creative Education
Drawing on Haraway's sympoiesis, postdigital theory, and decolonial computing, this paper proposes five principles for creative pedagogy in the age of generative AI: relational creativity, situated technology, epistemic justice, productive friction, and reflexive practice.
Status: Revised (v2)
Target: Postdigital Science and Education
SBBrasil TrainSheets
Human-in-the-Loop Dental Imaging Pipeline
SBBrasil TrainSheets
Human-in-the-Loop Dental Imaging Pipeline
A reproducible pipeline transforming public PDF manuals from Brazil's national oral health program into structured image datasets for computer vision research. Achieves 96.4% accuracy on dental view classification using lightweight methods with human-in-the-loop refinement.
Status: Published
Published in: IEEE Access
Capturing the Immersive
Knowledge Flows and Innovation Asymmetries in VR/XR Technology (2010–2024)
Capturing the Immersive
Knowledge Flows and Innovation Asymmetries in VR/XR Technology (2010–2024)
A bibliometric study examining global knowledge flows and innovation asymmetries in Virtual Reality and Extended Reality technologies. Introduces the DICE Framework (Dependency Ratio, Integration Index, Innovation Capture Index, and Efficiency Ratio) to quantify asymmetries in knowledge production and appropriation between regions.
Status: Data Collection
Target: Scientometrics
Methodology
Book in progress
From Caravaggio to Cyberpunk
Artistic References for Game Artists
A comprehensive visual guide connecting 600 years of art history to contemporary game design. From Medieval illuminations to Hyperrealism, each chapter pairs canonical artists with iconic games, demonstrating how historical techniques inform modern concept art, environment design, and visual storytelling.
- Years of Research
- 5
- Chapters
- 12 + Intro
- Images
- 250+
- Words
- 80,000
- Games Analyzed
- 80+
- Status
- In Revision
Table of Contents
01 Medieval and Gothic Art Bestiaries, Cathedrals and the Aesthetics of the Fantastic
Explores the art of the 5th to 15th centuries as a symbolic and theological system. Analyzes illuminations, Gothic architecture, bestiaries and religious iconography, highlighting how the absence of linear perspective created symbolic rather than physical spaces.
Artists Studied
Games Analyzed
02 Renaissance Perspective, Anatomy and the Construction of the Gaze
Investigates the visual revolution of the Italian Quattrocento and Cinquecento, focusing on the invention of linear perspective by Brunelleschi and its codification by Alberti. Examines how the anatomical studies of Leonardo and Michelangelo established standards that still inform character design.
Artists Studied
Games Analyzed
03 Baroque Drama, Light and the Theater of Emotions
Analyzes Counter-Reformation art as emotional propaganda, focusing on Caravaggio's tenebrism and theatricality as a persuasion strategy. Explores the use of diagonals, spirals and unstable compositions, along with architectural illusionism (trompe-l'œil).
Artists Studied
Games Analyzed
04 Rococo and Neoclassicism Between Elegance and Reason
Contrasts two opposing movements of the eighteenth century: the ornamental lightness of Rococo and the moral austerity of Neoclassicism. Introduces Edmund Burke's concept of the Sublime as a bridge between the two.
Artists Studied
Games Analyzed
05 Romanticism Nature, Emotion, and Ruins
Explores the reaction to Enlightenment rationalism and industrialization, focusing on nature as a secular cathedral and the artist as a solitary genius. Analyzes the iconography of ruins, cemeteries, and storms as expressions of melancholy and transcendence.
Artists Studied
Games Analyzed
06 Realism and Naturalism The World Without Idealization
Investigates art that 'comes down from the pedestal' after the 1848 revolutions, focusing on everyday themes and the rejection of idealization. Analyzes how Courbet painted peasants and workers with the same dignity reserved for kings.
Artists Studied
Games Analyzed
07 Impressionism and Post-Impressionism Light, Color and the Liberation of Vision
Analyzes the break with the Academy through the Salon des Refusés and plein air painting. Examines complementary color techniques, dissolution of contour and series as method. Explores the transition to Post-Impressionism: Seurat's pointillism, Cézanne's geometry, Van Gogh's emotional expression.
Artists Studied
Games Analyzed
08 Modernisms Fragmentation, Dream and the Destruction of Rules
Maps the simultaneous ruptures of the early twentieth century: Cubism (multiple viewpoints), Futurism (speed and machine), Dadaism (against art), Surrealism (access to the unconscious). Analyzes Picasso's spatial fragmentation and Dalí's paranoiac-critical method.
Artists Studied
Games Analyzed
09 Contemporary Art Pop, Minimalism and the Street as Gallery
Examines the dissolution between high and low culture in the postwar era. Analyzes Pop Art (mass culture as raw material), Minimalism (reduction to essentials) and Street Art (city as canvas).
Artists Studied
Games Analyzed
10 Hyperrealism and Digital Realism The Aesthetics of Simulation
Investigates the pursuit of perfect illusion, from Photorealism to contemporary Hyperrealism. Introduces the concepts of simulacrum (Baudrillard) and uncanny valley (Mori). Analyzes PBR techniques, subsurface scattering, ray tracing, and facial capture.
Artists Studied
Games Analyzed
11 Global Visual Traditions Beyond the Western Canon
Expands the repertoire beyond Europe, addressing responsibility in cultural representation. Examines Japanese art (ukiyo-e, sumi-e), Islamic art (arabesques, infinite geometry), African art (masks, rhythmic abstraction), and pre-Columbian art (cosmological iconography).
Artists Studied
Games Analyzed
12 Concept Art and the Creative Process From Idea to Image
Defines concept art as a communication language between departments, distinct from final illustration. Examines the process: thumbnails, silhouettes, rapid iteration, color keys, mood boards. Analyzes masters of visual design.
Artists Studied
Games Analyzed
Publications
Under Review / Submitted
Sympoietic Pedagogy: Toward a Framework for Human-AI Co-Creation in Creative Education
Postdigital Science and Education · 2025
Navigating Timbre: Locomotion as Epistemic Practice in Immersive Acoustic Environments
ACAH 2026 — Asian Conference on Arts & Humanities (IAFOR) · 2026
Inside Sound: Embodied Navigation Through Timbral Constellations in Virtual Reality
ISEA 2026 — International Symposium on Electronic Art (Dubai) · 2026
Data Constellations: Navigating Extinction Through Embodied Spatial Memory in Virtual Reality
ISEA 2026 — International Symposium on Electronic Art, Dubai · 2026
The Colonial Situation on Screen: Balandier, Fanon, and the Anthropology of Power in The Battle of Algiers
Journal of Postcolonial Writing (Taylor & Francis) · 2025
Measuring Cultural Representation Bias in CLIP Embeddings: A Large-Scale Empirical Study Across Five Museum Collections
Expert Systems with Applications (Elsevier) · 2026
Published
Beyond the Black Box: Mapping the Ontogenesis of AI-Generated Images Through Latent Space Analysis
Philosophy & Technology (Springer) · 2026
SBBrasil TrainSheets: From PDF Records to Dental View Classification with a Human-in-the-Loop Imaging Pipeline
IEEE Access · 2026
Zero-Shot Video Anomaly Detection via Cross-Modal Vision-Language Alignment
ICCiT 2025 — International Conference on Computing and Information Technology (IEEE) · 2025
An Evolving Fuzzy Neural Network Based on Or-Type Logic Neurons for Identifying and Extracting Knowledge in Auction Fraud
Mathematics · 2022
An Explainable Evolving Fuzzy Neural Network to Predict the k Barriers for Intrusion Detection Using a Wireless Sensor Network
Sensors · 2022
Evolving Fuzzy Neural Network Based on Null-Unineurons for the Identification of Coronary Artery Disease
IEEE International Conference on Systems, Man, and Cybernetics (SMC) · 2022
Humans and Machines: A Study of the Impacts of the Technological Advances in the Light of Generative Art Theory
ISEA 2022 — 27th International Symposium on Electronic Art · 2022
A Method to the Madness: modeling the creative process of new media artists through the analysis of knowledge exchanges
MULTIMODUS 2022 — 1st International Conference on Sound and Image in Art & Design (Instituto Politécnico de Portalegre, Portugal) · 2022
A Pipeline for Facial Animations on Low Budget VR Productions
IEEE Conference on Virtual Reality and 3D User Interfaces (VRW) · 2021
'If We Only Knew How You Feel' — A Comparative Study of Automated vs. Manual Classification of Opinions of Customers on Digital Media
Social Networking · 2018
Opinion Mining Framework Applied to a Social Networks Data for Small and Medium Enterprises
Revista Gestão & Tecnologia · 2020
Simulador de Realidade Virtual Aplicado à Educação Patrimonial para Experiências Imersivas Gamificadas
RENOTE — Revista Novas Tecnologias na Educação · 2018
Co-Creation in the Game Industry: A Case Study in a Brazilian Developer
Social Networking · 2018
Propagação de Boatos em Empresas: Modelo Baseado em Autômatos Celulares
14th CONTECSI — International Conference on Information Systems and Technology Management · 2017
Engajamento como Mecânica para a Gestão do Conhecimento do Consumidor com Apoio de Ferramentas da Web 2.0 para PMEs
14th CONTECSI — International Conference on Information Systems and Technology Management · 2017
Qualitative Analysis of T-Shaped and K.S.A Models
SODEBRAS · 2017
Leader Enterprise Content Management (ECM) Systems Market Report Comparison
SODEBRAS · 2017
Intellectual property
CACL: Cultural-Aware Contrastive Learning System
Patent application · INPI (Instituto Nacional da Propriedade Industrial), Brazil
A contrastive-learning system that incorporates cultural context into visual representation.
Theses
Doctoral and master's research, in full.
Towards a Design Methodology for Specialized Immersive Training utilizing Virtual Reality, Serious Games, Biofeedback, and UX Design
Ph.D. · Information Technology & Knowledge Management
The doctorate is the high point of a long interest in immersive technology and real-time interaction. It set out a design methodology that binds four elements (VR immersion, serious games, real-time biofeedback, and UX design) into one framework for building training simulators, and it backed the method with a complete VR prototype.
Towards a Design Methodology for Specialized Immersive Training utilizing Virtual Reality, Serious Games, Biofeedback, and UX Design
Ph.D. · Information Technology & Knowledge Management
Universidade Nove de Julho (UNINOVE) / University of Applied Sciences Upper Austria, Hagenberg
Prof. Dr. Marcos Antonio Gaspar (Advisor), Prof. Dr. Ulrich Norbisrath (Co-advisor), Prof. Jeremiah Diephuis (Local advisor in Austria)
The question
How should immersive training be designed so that it prepares people for high-pressure work rather than simulating the look of it? The thesis treated this as a design problem as much as a technical one.
Method
The methodology rests on four pillars: virtual reality immersion, serious games mechanics, real-time biofeedback monitoring, and human-centered UX design. Instead of treating these as separate features, the work specified how they fit together while a training simulator is being built. The approach was tested against practitioner and academic judgment through an international survey. To show it could produce something real, the research built a complete VR prototype with character animation, environmental storytelling, and a custom facial animation system driven by phonetic alignment. That prototype was aimed at law enforcement training, one of the clearer cases of preparing for pressure.
Contribution
The main contribution is the methodology itself: a framework that ties immersion, game mechanics, biofeedback, and user-centered design into one process for building specialized training simulators. Its value sits in the integration rather than the individual ingredients. Real-time biofeedback carries particular weight, treating the trainee's physiological state as a live signal the design has to respond to. The facial animation pipeline is a separate and more technical contribution, using phonetic alignment to generate lip-synced speech for the prototype's characters.
Findings
The methodology held up under external review, assessed by 141 professionals and academics from eleven countries, which gave it standing beyond a single author's claim. On the technical side, the facial animation system cut production time by 70% while keeping lip-sync accuracy intact, a real gain for teams on tight budgets.
Legacy
What carries forward from the doctorate is a disposition and a set of technical instincts: comfort with immersion, real-time feedback, and systems that run live. It is not continuity of subject. The thesis produced a training simulator, and the current lines of research did not exist in 2021. The thread between then and now is the maker, not the research question.
Social Media Opinion Mining Framework for Customer Knowledge Discovery
M.Sc. · Information Technology & Knowledge Management
The master's is where the formal training in artificial intelligence and machine learning took shape. It produced an opinion mining framework that combines sentiment analysis, topic modeling, and natural language processing to pull structured customer knowledge out of reviews and conversations, built to adapt across domains and languages.
Social Media Opinion Mining Framework for Customer Knowledge Discovery
M.Sc. · Information Technology & Knowledge Management
Universidade Nove de Julho (UNINOVE)
Prof. Dr. Marcos Antonio Gaspar (Advisor), Prof. Dr. Renato José Sassi (Co-advisor)
The question
Can the unstructured mass of social media opinion be turned into knowledge a company can act on, without a dedicated analytics team to sift it by hand? That practical question sits beneath the framework.
Method
The framework brings together three techniques from computational intelligence: sentiment analysis to read tone, topic modeling to surface what people are actually discussing, and natural language processing to handle the raw text. It operates on online reviews and social conversations, the kind of unstructured data most businesses collect and rarely use. The pipeline was built to be modular, so an organization can fit each stage to its own domain and language instead of accepting one fixed configuration.
Contribution
The contribution is the framework as an integrated pipeline, not a single algorithm. Sentiment, topic, and language analysis are combined into one process aimed specifically at customer knowledge: the recurring themes, complaints, and preferences buried in ordinary text. The modular design is part of that contribution, since it lets a small firm assemble a working analysis from open techniques rather than buy a finished platform.
Findings
What the framework demonstrated was feasibility: that the three techniques could be combined into one pipeline pulling customer knowledge from unstructured social media text, and that the pipeline could be retargeted across domains and languages. The source abstract does not report quantitative validation figures, so no accuracy or sample sizes are claimed here.
Legacy
The value of the master's is in the training, not in any line it started. Building an opinion mining pipeline on real social media data is where artificial intelligence and machine learning stopped being reading and became working tools. It is education, and the honest connection to the current work is only that the foundation was laid here, not that this project pointed toward it.
Download PDF Available only in Brazilian Portuguese