The thesis originates from an apparently simple question, one that must be protected from overly immediate answers: what happens to architectural design when generative tools are no longer understood merely as means of acceleration, optimisation, or formal production, but as epistemic devices, that is, as instruments capable of intervening in the very ways in which design is read, criticised, and produced? Recent debates on artificial intelligence in architecture often tend to unfold along two opposing trajectories. On the one hand, AI is interpreted as a technical apparatus of efficiency: a tool for automating operations, reducing time, and optimising solutions. On the other, it is approached as a machine of drift, capable of producing unexpected images and formal variations that are difficult to control. This research try to position itself within a third, non-conciliatory space: that of a designer who works with systems they do not fully master, transferring compositional, analogical, and interpretative skills into computational environments. It is within this unstable zone that the figure of the computational bricoleur, or “engineered savage”, takes shape: a subject who is neither the sovereign technician of the machine nor the passive user of its results. Rather, their work consists in constructing protocols, selecting corpora, interpreting errors, and forcing transitions between images, models, descriptions, and material verifications. Algorithmic error, or, more precisely, the algorithmic unexpected, is therefore not treated as a mere malfunction, nor is it naively celebrated as creative chance. Instead, it becomes a point of friction: the place where the system [un]veils a different organisation of the possible. The central question of the research is thus not what the machine produces, but how the relationship between designer and computational system redefines what can be recognised as hybrid, monstrous, or unexpected. The hypothesis is that generative tools are not neutral supports. They incorporate operative logics, regimes of selection, models of learning, and latent spaces that orient the field of design in advance. Starting from this premise, the thesis unfolds through three movements: first, a critical rereading of the hybrid as an architectural category, from its roots in hybris and monstrum to contemporary forms of generative error; second, an operational archaeology of computational systems, from cybernetics to neural networks, examined not in terms of their efficiency but through the implicit epistemologies they put to work; and finally, the construction of experimental protocols of uncertain verification, conceived also as pedagogical tools. Within these protocols, ambiguity is not something to be resolved, but what allows design to open up unforeseen directions.
La tesi nasce da una domanda apparentemente semplice, che occorre sottrarre a risposte troppo immediate: che cosa accade al progetto architettonico quando gli strumenti generativi non vengono più considerati soltanto come mezzi di accelerazione, ottimizzazione o produzione formale, ma come dispositivi epistemici, capaci cioè di intervenire sul modo stesso in cui il progetto viene letto, criticato e prodotto? Il dibattito recente sull’intelligenza artificiale in architettura tende spesso a disporsi lungo due traiettorie opposte. Da un lato, l’AI è interpretata come apparato tecnico di efficienza: uno strumento per automatizzare passaggi, ridurre tempi, ottimizzare soluzioni. Dall’altro, è assunta come macchina di deriva, capace di produrre immagini inattese e variazioni formali difficilmente controllabili. Questa ricerca prova a collocarsi in uno spazio terzo, non conciliatorio: quello di un progettista che lavora con sistemi che non domina completamente, trasferendo dentro ambienti computazionali competenze compositive, analogiche, interpretative. È in questa zona instabile che prende forma la figura del bricoleur computazionale, o del “selvaggio ingegnerizzato”: un soggetto che non coincide né con il tecnico sovrano della macchina né con l’utente passivo dei suoi risultati. La sua operazione consiste piuttosto nel costruire protocolli, scegliere corpus, interpretare errori, forzare passaggi tra immagini, modelli, descrizioni e verifiche materiali. L’errore algoritmico, o per meglio dire l’inatteso algoritmico, in questo senso, non viene assunto come semplice malfunzionamento, ma nemmeno celebrato ingenuamente come casualità creativa. Diventa invece un punto di attrito: il luogo in cui il sistema [dis]vela una diversa organizzazione del possibile. La domanda centrale della ricerca non è dunque che cosa produca la macchina, ma come la relazione tra progettista e sistema computazionale ridefinisca ciò che può essere riconosciuto come ibrido, mostruoso, inatteso. L’ipotesi è che gli strumenti generativi non siano supporti neutri. Essi incorporano logiche operative, regimi di selezione, modelli di apprendimento e spazi latenti che orientano in anticipo il campo del progetto. A partire da questa premessa, la tesi si articola lungo tre movimenti: una rilettura critica dell’ibrido come categoria architettonica, dalle sue radici nella hybris e nel monstrum fino alle forme contemporanee dell’errore generativo; un’archeologia dei sistemi computazionali, dalla cibernetica alle reti neurali, osservati non per la loro efficienza ma per le epistemologie implicite che mettono al lavoro; infine, la costruzione di protocolli sperimentali di verifica incerta, pensati anche come strumenti didattici. In questi protocolli, l’ambiguità non è ciò che deve essere risolto, ma ciò che consente al progetto di aprire direzioni non previste.
HYBRId S : on hybridisation as an epistemic and operational condition in architectural design
MAHI, HOUSSAM
2025/2026
Abstract
The thesis originates from an apparently simple question, one that must be protected from overly immediate answers: what happens to architectural design when generative tools are no longer understood merely as means of acceleration, optimisation, or formal production, but as epistemic devices, that is, as instruments capable of intervening in the very ways in which design is read, criticised, and produced? Recent debates on artificial intelligence in architecture often tend to unfold along two opposing trajectories. On the one hand, AI is interpreted as a technical apparatus of efficiency: a tool for automating operations, reducing time, and optimising solutions. On the other, it is approached as a machine of drift, capable of producing unexpected images and formal variations that are difficult to control. This research try to position itself within a third, non-conciliatory space: that of a designer who works with systems they do not fully master, transferring compositional, analogical, and interpretative skills into computational environments. It is within this unstable zone that the figure of the computational bricoleur, or “engineered savage”, takes shape: a subject who is neither the sovereign technician of the machine nor the passive user of its results. Rather, their work consists in constructing protocols, selecting corpora, interpreting errors, and forcing transitions between images, models, descriptions, and material verifications. Algorithmic error, or, more precisely, the algorithmic unexpected, is therefore not treated as a mere malfunction, nor is it naively celebrated as creative chance. Instead, it becomes a point of friction: the place where the system [un]veils a different organisation of the possible. The central question of the research is thus not what the machine produces, but how the relationship between designer and computational system redefines what can be recognised as hybrid, monstrous, or unexpected. The hypothesis is that generative tools are not neutral supports. They incorporate operative logics, regimes of selection, models of learning, and latent spaces that orient the field of design in advance. Starting from this premise, the thesis unfolds through three movements: first, a critical rereading of the hybrid as an architectural category, from its roots in hybris and monstrum to contemporary forms of generative error; second, an operational archaeology of computational systems, from cybernetics to neural networks, examined not in terms of their efficiency but through the implicit epistemologies they put to work; and finally, the construction of experimental protocols of uncertain verification, conceived also as pedagogical tools. Within these protocols, ambiguity is not something to be resolved, but what allows design to open up unforeseen directions.| File | Dimensione | Formato | |
|---|---|---|---|
|
houssam_mahi_hybri[d]s_on hybridisation as an epistemic and operational condition in architectural design.pdf
solo utenti autorizzati a partire dal 21/05/2029
Dimensione
110.85 MB
Formato
Adobe PDF
|
110.85 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/257937