The introduction of Artificial Intelligence (AI) into healthcare practice raises questions that extend beyond system performance and ethical acceptability, challenging the ways in which knowledge is produced, interpreted, and used in clinical reasoning. This thesis investigates the conditions under which knowledge produced through AI systems can reliably contribute to the epistemic processes of rehabilitation practice without reducing the complexity of the person to a computable representation. The study originates from an internship experience focused on the ethical assessment of voice datasets used for the prediction and monitoring of Parkinson’s disease. The challenges encountered in applying an assessment framework highlighted some of the limitations involved in operationalizing AI Ethics, shifting the focus of the research toward a more specifically epistemological question. Through a theoretical-conceptual and interdisciplinary approach, the thesis examines the characteristics of algorithmic inference, with particular attention to the relationship between prediction, explanation, and understanding, as well as to issues of opacity, causality, generalizability, and reliability. In parallel, it explores the distinctive nature of knowledge in rehabilitation, where scientific evidence, clinical reasoning, professional expertise, knowledge of the person, and the therapeutic relationship jointly contribute to understanding the individual case. The comparison between these two epistemic processes highlights the need for an epistemic alignment between the clinical question and the computational question. Integrating AI into clinical practice requires clarifying the clinical problem being addressed, which part of that problem has been made computationally tractable, what the resulting output represents, and what conclusions can legitimately be drawn from it. From this perspective, AI can be conceived as an epistemic mediator within a distributed process of knowledge production, in which algorithmic knowledge, scientific evidence, professional expertise, and the patient’s experiential knowledge contribute in different ways. The central issue is therefore not to establish the superiority of human or algorithmic knowledge, but to define their respective roles, limitations, and conditions for integration, so that AI may expand the possibilities of knowing without reducing the complexity of the person and of care.
L’introduzione dell’Intelligenza Artificiale (AI) nella pratica sanitaria non solleva soltanto questioni relative alle prestazioni dei sistemi o alla loro accettabilità etica, ma interroga il modo in cui la conoscenza viene prodotta, interpretata e utilizzata nel ragionamento clinico. La presente tesi indaga a quali condizioni la conoscenza prodotta attraverso sistemi di AI possa contribuire in modo affidabile al processo conoscitivo della pratica riabilitativa, senza ridurre la complessità della persona alla sua rappresentazione computabile. Il lavoro prende avvio da un’esperienza di tirocinio dedicata alla valutazione etica di dataset vocali utilizzati per la predizione e il monitoraggio della Malattia di Parkinson. Le criticità incontrate nell’applicazione di un framework valutativo hanno evidenziato alcuni limiti dell’operazionalizzazione dell’AI Ethics, orientando la ricerca verso una questione più propriamente epistemologica. Attraverso un’impostazione teorico-concettuale e interdisciplinare, vengono analizzate le caratteristiche dell’inferenza algoritmica, con particolare attenzione al rapporto tra predizione, spiegazione e comprensione e ai problemi dell’opacità, della causalità, della generalizzabilità e dell’affidabilità. Parallelamente, viene ricostruita la specificità della conoscenza in riabilitazione, nella quale evidenze scientifiche, ragionamento clinico, esperienza professionale, conoscenza della persona e relazione terapeutica concorrono alla comprensione del caso particolare. Dal confronto tra questi due processi emerge la necessità di un allineamento epistemico tra domanda clinica e domanda computazionale: integrare l’AI nella pratica richiede di chiarire quale problema clinico si intenda affrontare, quale parte di esso sia stata resa computabile, che cosa rappresenti l’output prodotto e quali conclusioni esso consenta effettivamente di formulare. In questa prospettiva, l’AI può essere concepita come un mediatore epistemico all’interno di un processo conoscitivo distribuito, nel quale conoscenza algoritmica, evidenze scientifiche, competenza professionale e sapere esperienziale del paziente contribuiscono in forme differenti. La questione non consiste quindi nello stabilire una superiorità tra conoscenza umana e algoritmica, ma nel definirne funzioni, limiti e condizioni di integrazione, affinché l’AI possa ampliare le possibilità di conoscenza senza ridurre la complessità della persona e della cura.
Conoscere per curare nell'era dell'Intelligenza Artificiale: elementi epistemologici per il ragionamento clinico in riabilitazione.
PIGLIACAMPO, LAURA
2025/2026
Abstract
The introduction of Artificial Intelligence (AI) into healthcare practice raises questions that extend beyond system performance and ethical acceptability, challenging the ways in which knowledge is produced, interpreted, and used in clinical reasoning. This thesis investigates the conditions under which knowledge produced through AI systems can reliably contribute to the epistemic processes of rehabilitation practice without reducing the complexity of the person to a computable representation. The study originates from an internship experience focused on the ethical assessment of voice datasets used for the prediction and monitoring of Parkinson’s disease. The challenges encountered in applying an assessment framework highlighted some of the limitations involved in operationalizing AI Ethics, shifting the focus of the research toward a more specifically epistemological question. Through a theoretical-conceptual and interdisciplinary approach, the thesis examines the characteristics of algorithmic inference, with particular attention to the relationship between prediction, explanation, and understanding, as well as to issues of opacity, causality, generalizability, and reliability. In parallel, it explores the distinctive nature of knowledge in rehabilitation, where scientific evidence, clinical reasoning, professional expertise, knowledge of the person, and the therapeutic relationship jointly contribute to understanding the individual case. The comparison between these two epistemic processes highlights the need for an epistemic alignment between the clinical question and the computational question. Integrating AI into clinical practice requires clarifying the clinical problem being addressed, which part of that problem has been made computationally tractable, what the resulting output represents, and what conclusions can legitimately be drawn from it. From this perspective, AI can be conceived as an epistemic mediator within a distributed process of knowledge production, in which algorithmic knowledge, scientific evidence, professional expertise, and the patient’s experiential knowledge contribute in different ways. The central issue is therefore not to establish the superiority of human or algorithmic knowledge, but to define their respective roles, limitations, and conditions for integration, so that AI may expand the possibilities of knowing without reducing the complexity of the person and of care.| File | Dimensione | Formato | |
|---|---|---|---|
|
PIGLIACAMPO_LAURA_TESI_PDFA.pdf
non disponibili
Dimensione
1.22 MB
Formato
Adobe PDF
|
1.22 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/29338