
Questa tesi esplora lo sviluppo di un framework adattivo per il supporto alle decisioni automatizzate, ispirato ai processi cognitivi umani descritti nella psicologia comportamentale. Il sistema proposto, denominato CASS (Context Aware Selection System), è progettato per selezionare dinamicamente l’algoritmo di machine learning più appropriato in base alla distribuzione dei dati e al contesto operativo, massimizzando la fiducia nella decisione presa. L'architettura del CASS si fonda sull'integrazione di tecniche di ensemble learning e modelli di scoring basati sulla fiducia, con l’obiettivo di aumentare la trasparenza e l’adattabilità del processo decisionale. Il sistema prende ispirazione dal Sistema 2 di Kahneman, implementando un approccio deliberativo e analitico alla scelta dell’algoritmo. La validazione sperimentale è stata condotta su diversi dataset reali, sia in ambito di classificazione che di regressione, dimostrando miglioramenti significativi in termini di robustezza, affidabilità e interpretabilià rispetto a metodi tradizionali e a baseline naïve. La tesi include inoltre una proposta metodologica per l’analisi del livello di fiducia nei modelli, utile per applicazioni in domini critici come la cybersecurity, la sanità e l’intelligenza artificiale responsabile. Per ulteriori informazioni o approfondimenti sul progetto, è possibile contattarmi all’indirizzo mpietri82@gmail.com oppure visitare il sito marcopietri.framer.website.
This thesis presents an adaptive framework for decision support, inspired by human cognitive processes. The proposed system, CASS (Context Aware Selection System), dynamically selects the most suitable algorithm based on the characteristics of the input data and the context. Combining ensemble learning and confidence-based evaluation, the system aims to enhance the reliability and transparency of machine learning decisions. The implementation includes a modular architecture influenced by Kahneman's System 2 and demonstrates improved performance across various datasets and learning scenarios. For further information or insights regarding this project, feel free to contact me at mpietri82@gmail.com or visit marcopietri.framer.website.
Machine Learning/ethics, Kahneman System 2, Dataset analysis, Fairness in machine learning, Adaptive systems, Model reliability, Confidence-based evaluation, Cognitive architectures, Decision-making support, Algorithm selection, AI transparency, trustworthy AI, Context-aware reasoning, Machine learning, Machine Learning/classification, Man-Machine Systems, Bias mitigation, Psychology and AI, Trust-based systems, Explainable AI (XAI), Classifier orchestration, Dynamic model selection, CASS framework, Human-centered AI, Supervised Machine Learning, Human-inspired AI
Machine Learning/ethics, Kahneman System 2, Dataset analysis, Fairness in machine learning, Adaptive systems, Model reliability, Confidence-based evaluation, Cognitive architectures, Decision-making support, Algorithm selection, AI transparency, trustworthy AI, Context-aware reasoning, Machine learning, Machine Learning/classification, Man-Machine Systems, Bias mitigation, Psychology and AI, Trust-based systems, Explainable AI (XAI), Classifier orchestration, Dynamic model selection, CASS framework, Human-centered AI, Supervised Machine Learning, Human-inspired AI
| selected citations These citations are derived from selected sources. This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically). | 0 | |
| popularity This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network. | Average | |
| influence This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically). | Average | |
| impulse This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network. | Average |
