 |
Status: concluded Period: September 2018 - May 2020Funding: 30.000 € Funding organization: TIM S.p.A. Person(s) in charge: Antonio Vetrò, Antonio Santangelo, Eleonora Bassi, Giacomo Conti, Elena Beretta , Flavio Cannavò |
Executive summary:
This project is part of a context of ongoing transformation towards the digitalization of society and industry based on technological developments in the ICT sector.
Background:
The context of Telecom operators, as often happens, has two sides: on the one hand telecommunication systems are among the enabling technologies for the diffusion of AI solutions, on the other hand operators are themselves large users (e.g. for network automation solutions, network protection and security systems, customer interactions).
The project is part of a context of ongoing transformation towards the digitalization of society and industry based on technological developments in the ICT sector. Communication networks will represent an increasingly relevant component for every industrial and governmental sector, also thanks to the development of 5G networks. This technology, in fact, is not only a radio interface and a faster and more performing network, but thanks to the use of a set of software-based technologies that make the network easily configurable and flexible, allows the development of new business models in the service offering.
Big data, statistical techniques of AI and Machine Learning will be key components of 5G.
Objectives:
Starting from the experience of the use of AI in TLC environment, the project covered the ethical and privacy aspects focusing on the issues of discrimination that may arise, analyzing the complex concept of equity aimed at mitigating such discrimination and examining the emerging guidelines at European and national level and the ethical codes adopted by companies.
A part of the project focused on data ethics and algorithms, and the following topics:
• Fairness and justice: mathematical statistical definitions of the concept of fairness and reflections on recommendation systems.
• Ethical and socially aware data labels: how to label databases, so that it is clear what has been done to consider the ethical issues raised by their use.
• Ethical guidelines for the development of AI: analysis of the documents produced by the various committees set up by States and the European Commission.
Another part of the project focused on the following legal aspects:
• "AI: between corporate governance and legal responsibility in the telecommunications sector".
• Examination of use cases in TIM.
• Problems of legal liability: contractual and non-contractual.
• Governance between business ethics, organizational processes and corporate responsibility.
• The role of codes of ethics.
• Possible cases.
Results:
The results of the collaboration were the following:
• A selection and analysis of real cases of discrimination resulting from the incorrect design of automatic systems and/or decision-making algorithms or the use of databases with bias problems, with a focus on the European context.
• The identification of metrics for the measurement of bias in databases, which can be used as "risk labels" of discrimination resulting from their use.
• A legal analysis of philosophical, legal principles, guidelines and codes of ethics concerning the development of Artificial Intelligence systems.
The deliverables have been presented and discussed with two workshops at TIM. In addition, the results of the legal part have been presented at the 77° Nexa Lunch Seminar (https://nexa.polito.it/lunch-77)
