The problems that our society is facing right now, such as the increasing pollution and the global warming, demand us to change our habits, adopting new technological solutions. For this reason studying the processes regulating their adoption could support the formulation of public policies designed to promote them. However, in recent years these efforts resulted ineffective, failing to achieve the expected goals and perpetrating inequalities between the population. Ignoring social consensus and diversity in individuals is key to foster the use of innovations. In addition, all the different elements entering the process, such as the influence of external opinions, the presence of biases toward certain categories and the interference of random factors, should be accounted for in the design of the policies. To address this challenges, we propose an extended Linear Threshold model that describes the dynamics of technology adoption incorporating the intrinsic stochasticity of individuals' behavior. Moreover, structural biases against marginalized groups, which might hinder the spread the innovations, were modeled by assigning heterogeneous and asymmetric weights to social ties. Building on this, we introduce a fairness-aware Model Predictive Control (MPC) strategy that incorporates principles of equity and equality, as well as budget constraints, to optimize the diffusion process while ensuring a fair distribution of resources and opportunities among agents. The proposed framework has been tested on synthetic and real-world networks, exploring how the presence of bias affects the evolution of the adoption processes and what are the possible trade-offs between fairness and efficiency objectives. This work contributes to the emerging intersection between control theory, algorithmic fairness, social sciences and network analysis, offering insights into designing equitable interventions for technology adoption.
I problemi che la nostra società sta affrontando in questo momento, come il crescente inquinamento e il riscaldamento globale, ci impongono di cambiare le nostre abitudini, adottando nuove soluzioni tecnologiche. Per questo motivo, studiare i processi che ne regolano l'adozione potrebbe supportare la formulazione di politiche pubbliche volte a promuoverle. Tuttavia, negli ultimi anni questi sforzi si sono rivelati inefficaci, non riuscendo a raggiungere gli obiettivi attesi e perpetuando disuguaglianze tra la popolazione. Ignorare il consenso sociale e la diversità degli individui è fondamentale per promuovere l'uso delle innovazioni. Inoltre, tutti i diversi elementi che entrano in gioco nel processo, come l'influenza di opinioni esterne, la presenza di pregiudizi verso determinate categorie e l'interferenza di fattori aleatori, dovrebbero essere considerati nella progettazione delle politiche. Su questa base, introduciamo una strategia di controllo predittivo (MPC) che incorpora i principi di equità ed uguaglianza, nonché vincoli sulle risorse, per ottimizzare il processo di diffusione garantendo al contempo un'equa distribuzione di risorse e opportunità tra gli agenti. Il framework proposto è stato testato su reti sintetiche e reali, esplorando come la presenza di pregiudizi influenzi l'evoluzione dei processi di adozione e quali siano i possibili compromessi tra obiettivi di equità ed efficienza. Questo lavoro contribuisce all'intersezione emergente tra teoria del controllo, equità algoritmica, scienze sociali e analisi di rete, offrendo spunti per la progettazione di interventi equi per l'adozione della tecnologia.
Shaping opinion dynamics under epistemic injustice via fairness-aware control
Vitale, Edoardo
2024/2025
Abstract
The problems that our society is facing right now, such as the increasing pollution and the global warming, demand us to change our habits, adopting new technological solutions. For this reason studying the processes regulating their adoption could support the formulation of public policies designed to promote them. However, in recent years these efforts resulted ineffective, failing to achieve the expected goals and perpetrating inequalities between the population. Ignoring social consensus and diversity in individuals is key to foster the use of innovations. In addition, all the different elements entering the process, such as the influence of external opinions, the presence of biases toward certain categories and the interference of random factors, should be accounted for in the design of the policies. To address this challenges, we propose an extended Linear Threshold model that describes the dynamics of technology adoption incorporating the intrinsic stochasticity of individuals' behavior. Moreover, structural biases against marginalized groups, which might hinder the spread the innovations, were modeled by assigning heterogeneous and asymmetric weights to social ties. Building on this, we introduce a fairness-aware Model Predictive Control (MPC) strategy that incorporates principles of equity and equality, as well as budget constraints, to optimize the diffusion process while ensuring a fair distribution of resources and opportunities among agents. The proposed framework has been tested on synthetic and real-world networks, exploring how the presence of bias affects the evolution of the adoption processes and what are the possible trade-offs between fairness and efficiency objectives. This work contributes to the emerging intersection between control theory, algorithmic fairness, social sciences and network analysis, offering insights into designing equitable interventions for technology adoption.| File | Dimensione | Formato | |
|---|---|---|---|
|
Thesis_Edoardo_Vitale.pdf
accessibile in internet per tutti
Dimensione
4.32 MB
Formato
Adobe PDF
|
4.32 MB | Adobe PDF | Visualizza/Apri |
|
Executive_Summary_Edoardo_Vitale.pdf
accessibile in internet per tutti
Dimensione
819.21 kB
Formato
Adobe PDF
|
819.21 kB | 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/10589/240502