The pervasiveness of algorithms in everyday decisions has raised growing concerns about their potential discrimination and superficiality. This thesis aims to replicate and extend the analyses conducted in a well-known 2016 ProPublica study on the COMPAS algorithm, which is widely used in the U.S. justice system to predict criminal recidivism. In addition, the Aequitas audit toolkit will be used to explore and compare the implications of algorithmic justice in the results produced by COMPAS. This work not only offers a critical analysis of the fairness of COMPAS, but also demonstrates the effectiveness of tools such as Aequitas in promoting greater transparency and accountability in developing decision-making algorithms and their dissemination. Finally, this study aims to offer itself as a contribution to pursuing a society in which fairness and justice drive the evolution of technology.
La pervasività degli algoritmi nelle decisioni quotidiane ha suscitato crescenti preoccupazioni riguardo la loro potenziale discriminazione e superficialità. Questa tesi si propone di replicare ed ampliare le analisi condotte in un noto studio di ProPublica del 2016 sull’algoritmo COMPAS, ampiamente utilizzato nel sistema giudiziario statunitense per predire la recidiva criminale. Inoltre, verrà utilizzato il toolkit di audit Aequitas per esplorare e confrontare le implicazioni di giustizia algoritmica nei risultati prodotti da COMPAS. Questo lavoro non offre solo un’analisi critica della fairness di COMPAS, ma dimostra anche l’efficacia di strumenti come Aequitas nel promuovere una maggiore trasparenza e responsabilità nello sviluppare algoritmi decisionali e nella loro diffusione. Infine, questo studio vuole proporsi come un contributo per perseguire una società in cui l’equità e la giustizia guidino l’evoluzione della tecnologia.
Misurare la fairness negli algoritmi: il caso di studio COMPAS
BONATO, GIULIA
2023/2024
Abstract
The pervasiveness of algorithms in everyday decisions has raised growing concerns about their potential discrimination and superficiality. This thesis aims to replicate and extend the analyses conducted in a well-known 2016 ProPublica study on the COMPAS algorithm, which is widely used in the U.S. justice system to predict criminal recidivism. In addition, the Aequitas audit toolkit will be used to explore and compare the implications of algorithmic justice in the results produced by COMPAS. This work not only offers a critical analysis of the fairness of COMPAS, but also demonstrates the effectiveness of tools such as Aequitas in promoting greater transparency and accountability in developing decision-making algorithms and their dissemination. Finally, this study aims to offer itself as a contribution to pursuing a society in which fairness and justice drive the evolution of technology.File | Dimensione | Formato | |
---|---|---|---|
Bonato_Giulia.pdf
accesso aperto
Dimensione
2.02 MB
Formato
Adobe PDF
|
2.02 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/67352