The project MORE-GREEN (Mechanism Design, Online Learning, Robust Optimization, and Sentiment Extraction Tools for Adjustable Green-Aware Agents) is supported by the PNRR project FAIR—Future AI Research (PE00000013) under the NRRP MUR program funded by NextGenerationEU.

This project aims at providing tools for designing intervention plans for boosting green-aware behaviours in agents. However, the agents’ behaviours can depend on multiple features, on the interactions among agents, and the influence that they lean on each other. And these features are often unknown to the designer, or their knowledge is noisy or partial.

In this project we will use techniques from sentiment analysis, social (hyper)-network mining for extract the relevant features from data observed by the interaction of user in social networks. We will use online learning techniques to tune the intervention plans to the unknown and evolving features of the agents and their interactions. We will use mechanism design techniques for designing simple and opportune incentives leading agents to take the desired behaviour, regardless of their own features. We finally use techniques from robust optimization and the design of algorithms with predictions, to design interventions that are guaranteed to achieve good performances when agents’ features are correctly known, and, at the same time, that these performances do not degrade too much when the inputs are noisy.

We will apply these tools to three problems of interest: facility location (i.e., where to install “green” facilities), task allocation (e.g., allocation of “green” tasks), and influence of opinion formation.