Italy

Authored by: 

Organization: 

Eurovisioni
AttachmentSize
PDF icon gisw2019_itlay_Challenges_for_activists673.45 KB

Challenges for activists pushing for an ethical approach to AI in Italy

Introduction

Italy is the second biggest manufacturing country after Germany in the European Union (EU), and one of the main university and research centres. Because of this, the use and application of artificial intelligence (AI) in the country has advanced and spread. While this is most notable in robotics applied to the industrial sector, there are a number of experimental projects using AI in the country, including in the media sector. Recently, another area that has been developing fast has to do with building so-called “smart cities”, with a vast number of start-ups specialised in making services more accessible to citizens in various fields (health, education, social assistance, unemployment benefits, etc.). There is also strong and lively research activity on the ethical consequences of AI, including decision-making processes assisted by AI, and on related issues such as the regulation and self-regulation of media and social media, especially in the context of elections, hate speech, fake news, and child protection online.

Governments in the country, over the years, have had different approaches to and interest in AI. The centre-left government that was in power from 2013 until the beginning of 2018 was pushing hard (both in the country and in international forums such as the G7) to develop common policies on AI. In 2017, the Ministry of Economic Development launched a special fund (called “Industria 4.0”) to finance the introduction of AI and digital innovation in the traditional industrial sectors.[2] The plan was a huge success, and increased investment, especially in the automotive industry, but also in the pharmaceutical and energy sectors.[3]

Outlining some AI initiatives in the media sector in Italy, this report argues that the current political climate in the country and a fragmented civil society are making the ethical regulation of AI very difficult. It nevertheless identifies two potential policy windows for activists to push for a human-rights approach to AI implementation.

Laboratories for innovation in the media sector

In Italy, AI applications in general (and particularly in the media sector) have been mainly developed in areas where there are a lot of repetitive tasks and where using a human for a task adds little value to the job at hand. The two AI-based services of particular social relevance in the media sector have been developed in the areas of increasing accessibility for people with disabilities and the automatic archiving of digital multimedia files.[4]

For example, Radiotelevisione Italiana (RAI, the national public service media), through its research centre based in Turin (CRITS), is working on numerous AI applications aimed at making accessible radio and TV programmes for the elderly and people with disabilities. This work is the result of a new 2018-2027 framework agreement between the government and RAI, which aims to increase the percentage of programmes for people who are hard of hearing and those with reduced vision, and even deaf and blind people.[5]

RAI has developed three prototypes for new services based on AI that are now in the testing phase: Virtual LIS (short for Lingua dei Segni Italiana, Italian Sign Language), a virtual weather forecast using LIS, and Stretch TV.

Virtual LIS is a system that automatically converts voice audio into sign language using a 3D virtual interpreter. The first tests using LIS have been for weather forecasts.

Stretch TV is a system that can slow down the speed of TV programmes – both images and audio – by 10% to 20%, so that even the hard-of-hearing can easily follow dialogue. As part of EU-funded projects, RAI has also developed an archiving system for video and film using AI.[6]

There are many of these kinds of activities in the country, often implemented as part of EU research consortia.

One particularly interesting project involving media research is called “Femicide Storytelling”, run by the University of Turin (within the PRIN project).[7] Using AI, the project gathers data on cases of femicide and violence against women that are reported in newspapers and on radio and TV. These are then categorised and mapped against existing socio-demographic data in order to create a framework for predicting crimes against women. This is one of the most advanced experiments in big data in the country. While preventing crimes against women, it can be used to raise public awareness, and produce meaningful stories with a social impact.[8]

The most important EU project currently ongoing in the country is SoBigData, of which the Istituto Superiore Sant’Anna, based in Pisa, is one of the major actors and coordinators.[9] This project also has applications in the field of media research, particularly online.

The idea behind the project is very simple: according to the dominant narrative, big data is the oil of future growth. According to the research project, this narrative is true, but is currently being undermined by the misappropriation of personal data by some big companies that use personal data to make profits.

SoBigData proposes to create the “Social Mining and Big Data Ecosystem”, a research infrastructure that would allow the use of data in a protected environment, and on a collaborative basis, aiming to create a totally new approach to the use of big data. The aim of the project is to create a research community that could use the research infrastructure as a “secure digital wind-tunnel” for large-scale “social data analysis and simulation experiments.” Establishing common ethical principles on the use of algorithms, AI and big data is the pre-condition of making such an ecosystem workable, reliable and sustainable.

The areas of potential applications are enormous. The project has identified some of them: from smart cities to health, from media and understanding the influence of social media, to migration, to sports. For instance, the consortium has produced an analysis of mobility within a metropolitan area and even a map of accessibility for people in wheelchairs for a town, based on the data collected through an app for mobile phones. Another ongoing test involves the correlation between the use of social media and the polarisation of political and societal debates (analysed through algorithms checking and monitoring social media debates). The consortium publishes a newsletter presenting some of its experiments and achievements.[10]

The context for policy advocacy on AI

Italy is a country where the debate about social media and its misuse (hate speech, discrimination, sexual harassment and so on) has been widespread in media and society for many years. The proliferation of local media (radio and TV) in the last decades of the 20th century has raised awareness of media-related issues and led to the establishment of specific jurisprudence, as well as a media-related authority set up in 1987. The Autorità per le Garanzie nelle Comunicazioni (AGCOM, the Authority for the Control of Media) is a multisectoral authority, now with a very broad mandate: it is in charge of regulating printed media, electronic media, telecoms and (by extension) social media.

In an already highly regulated media ecosystem, the arrival of social media was nevertheless a shock, because some of the protections previously ensured by the traditional media regulations proved to be inefficient when it came to the new online forms of communication. As a result, calls to regulate social media came from various sectors (the judiciary, human rights activists, the media, etc.) – calls that the two parties governing the country in a coalition since June 2018 have not supported at all.

On the contrary, the Five Star Movement, led by comedian Beppe Grillo, and the League, led by Matteo Salvini – the two political parties forming the government that collapsed at the end of August 2019 – have based a lot of their success on the use of social media. For example, Salvini, whose Facebook page has more than 3.75 million fans, prefers to talk to his supporters live on Facebook instead of going into Parliament, holding press conferences or being interviewed on TV.[11]

The Five Star Movement makes most of its decisions using a voting platform called Rousseau,[12] where registered members are called to vote on the selection of candidates for elections, on main legislative proposals and even on internal reform of the party’s bylaws.[13]

Consequently, both parties had no great interest to intervene on the matter of the regulation of the internet, because they position themselves as anti-establishment forces, and consider traditional media to be controlled by the main economic and political interest groups in the country. This refusal, however, creates a growing tension between government, civil society, opposition parties and traditional media, but also within civil society, and the same two ruling parties.

This also makes pushing for the ethical regulation of AI difficult in the country. Nevertheless, two potential policy windows remain.

Policy windows to advocate for the ethical regulation of AI

The push for regulating social media

Despite the government's reluctance to regulate social media, civil society pressured AGCOM to approve a proposed regulation against online hate speech on 15 May 2019. The proposed regulation is called the “Regolamento recante disposizioni in materia di rispetto della dignità umana e del principio di non discriminazione e di contrasto all’hate speech” (Regulation containing provisions regarding respect for human dignity and the principle of non-discrimination and combatting hate speech).[14]

In Article 9 of the Regulation, AGCOM requires that internet platforms distributing video content:

  • Adopt, publish and respect formal codes of self-regulation in which they commit themselves to remove hate speech content and other content violating human rights online.
  • Provide quarterly monitoring reports on content identified, removed and penalised.
  • Launch awareness campaigns among their users to promote diversity and to fight any kind of online discrimination.

The regulation was submitted for public consultation before it was adopted. Some of the consulted entities have argued that the media and telecom authority has no mandate to regulate (or even co-regulate) social media and the internet. AGCOM has submitted the proposed regulation officially to the Italian Parliament, which will now have to decide whether to enshrine the regulation as law and formally confirm AGCOM's de facto mandate on social media.[15]

This proposal potentially opens up the door for further public discussion on ethical principles with respect to the use and implementation of AI online.

The multistakeholder roundtable

Another potential avenue where activists could push for the regulation of AI is in a multistakeholder roundtable set up by AGCOM, which contributed to the development of the regulation approved in May. In November 2017, AGCOM created the roundtable in an attempt to guarantee pluralism and to ensure reliable news on digital platforms.[16] This informal, voluntary group has grown over the years, and today gathers around 30 companies, including Google, Facebook and Wikipedia (Twitter has not joined the roundtable yet), institutions and civil society organisations. The job of the roundtable is to come up with solutions to problems that are raised, based on in-depth research and knowledge of the Italian media market.[17]

Since January 2019, it has produced a regular newsletter (called the Observatory on Online Disinformation) and all relevant communication between the authority and the parties involved, such as an internet platform, are discussed publicly and published on its website. For instance, in March 2018 the secretariat of the roundtable addressed a detailed request for explanations from Facebook on the impact of the Cambridge Analytica case in Italy. Another area of concern is the impact of fake news and hate speech on the online world.

During the campaigns for the Italian national elections in March 2018 and for European elections in May 2019, the roundtable agreed on a special strategy to counter possible cyber interference,[18] while some guidelines to guarantee equal access for all candidates to the online platforms were issued.[19]

Conclusion

In the current political turmoil through which Italy is passing, civil society organisations are in a trap. On the one hand they are very involved with institutions in the debate about AI and algorithms, on their implications for society (as seen in some of the projects discussed here) and in the development of regulations. On the other hand, they are paralysed in their advocacy because criticism of how internet platforms operate is considered as an attack on both of the ruling parties (the Five Star Movement and the League), both of which use the internet as a bedrock for their campaigns, including in their attacks on traditional media.

Civil society in Italy is also currently fragmented, with different groups concentrating on their own problems: political parties are worried about trolls and elections; feminist movements are worried about femicide; the child protection movement is worried about the risks of social media for children; minorities and refugee organisations are worried about hate speech and its consequences for their constituencies.

Nevertheless, two avenues for pushing for regulations on AI remain: the current advocacy for the regulation of hate speech online, and the multistakeholder roundtable set up by AGCOM.

A further policy window has presented itself with the sudden and unexpected collapse in August 2019 of the populist and right-wing parties that make up the current government. This has ushered in a change of alliances (now populist and left-wing parties), and opened up new perspectives for civil society activism, and for the Parliament. However, what exactly these advocacy opportunities will be is still too early to predict.

Action steps

If the fragmented civil society movement is able to free itself from its current paralysis, then a lot could be achieved with respect to AI. They need to collectively start to push for ethical principles to be incorporated into AI, such as privacy by design in the algorithms that are developed. This could be a successful way to unify movements that at the moment are not cooperating and are working separately, focused on their own battles. In this sense, the Italian tradition of fighting for civil liberties – which resulted in a very detailed and in-depth juridical framework and a modern constitution – could contribute significantly to the European and global debate on AI.

Footnotes

[1] The author thanks Andrea Cairola, Roberto Masotti and Francesco Sciacchitano for their support in this work.

[2] https://www.mise.gov.it/images/stories/documenti/investimenti_impresa_40_eng.pdf

[3] The media and cultural sectors were not allowed to access these incentives, which were reserved for the industrial sector, but nevertheless many research initiatives have been launched since then with private and EU funding.

[4] www.crit.rai.it/CritPortal/wp-content/uploads/2018/11/2018-10-09_TheExperienceOfRaiDigimasterProjectInProgress_sitow.pdf

[5] Obligations are mentioned in the new service’s contract with RAI for 2018-2022, published in January 2018. The services to increase accessibility are covered in Article 25 of the contract. www.rai.it/dl/doc/1521036887269_Contratto%202018%20testo%20finale.pdf

[6] CRITS and RAI Teche presented the archiving system at the FIAT/IFTA World Conference in Venice on 9 October 2018. www.crit.rai.it/CritPortal/?notizia=fiat-ifta-world-conference-2018&lang=en

[7] https://cercauniversita.cineca.it/php5/prin/cerca.php?codice=2015BBLK7M

[8] www.crit.rai.it/CritPortal/?notizia=1553

[9] The project receives funding from the EU Horizon 2020 research and innovation programme. See sobigdata.eu/index for more information; a complete description of the project can be found here: www.sobigdata.eu/sites/default/files/www%202018.pdf

[10] sobigdata.eu/newsletter

[11] According to the last available statistics on 4 August: https://www.socialbakers.com/statistics/facebook/pages/detail/2523060331...

[12] https://rousseau.movimento5stelle.it/sso_home.php

[13] The platform has around 150,000 registered participants, but data is not certified. The Italian authority on privacy has recently (4 April 2019) criticised the platform for not being protected against piracy and third-party manipulation. See: https://www.garanteprivacy.it/web/guest/home/docweb/-/docweb-display/doc...

[14] https://www.agcom.it/documentazione/documento?p_p_auth=fLw7zRht&p_p_id=101_INSTANCE_ls3TZlzsK0hm&p_p_lifecycle=0&p_p_col_id=column-1&p_p_col_count=1&_101_INSTANCE_ls3TZlzsK0hm_struts_action=%2Fasset_publisher%2Fview_content&_101_INSTANCE_ls3TZlzsK0hm_assetEntryId=15055471&_101_INSTANCE_ls3TZlzsK0hm_type=document

[15] In particular, Article 1.31 of the law that instituted AGCOM (Legge 31 luglio 1997, n. 249 – Istituzione dell'Autorità per le garanzie nelle comunicazioni e norme sui sistemi delle telecomunicazioni e radiotelevisivo) grants it the authority to impose penalties on those who do not comply with its regulations, ranging from fines to revoking their licence to operate in Italy. https://temi.camera.it/leg18/post/OCD15-54606/approvato-dall-agcom-regolamento-contrasto-alle-espressioni-d-odio-hate-speech.html

[16] AGCOM. (2017). Tavolo tecnico per la garanzia del pluralismo e della correttezza dell’informazione sulle piattaforme digitali. https://www.agcom.it/documents/10179/8860814/Comunicato+stampa+16-11-2017/7977d222-b9bd-4763-b1e6-9ae7c960a0ee?version=1.0

[17] https://www.agcom.it/gli-studi-e-le-indagini-conoscitive

[18] AGCOM. (2019). Osservatorio sulla disinformazione online n. 4/2019. https://www.agcom.it/visualizza-documento/721fe550-fcb0-4a05-b3d9-30012a...

[19] AGCOM. (2018). Linee guida per la parità di accesso alle piattaforme online durante la campagna elettorale per le elezioni politiche 2018. https://www.agcom.it/documentazione/documento?p_p_auth=fLw7zRht&p_p_id=101_INSTANCE_2fsZcpGr12AO&p_p_lifecycle=0&p_p_col_id=column-1&p_p_col_count=1&_101_INSTANCE_2fsZcpGr12AO_struts_action=%2Fasset_publisher%2Fview_content&_101_INSTANCE_2fsZcpGr12AO_assetEntryId=9478309&_101_INSTANCE_2fsZcpGr12AO_type=document

Notes:
This report was originally published as part of a larger compilation: “Global Information Society Watch 2019: Artificial intelligence: Human rights, social justice and development"
Creative Commons Attribution 4.0 International (CC BY 4.0) - Some rights reserved.
ISBN 978-92-95113-12-1
APC Serial: APC-201910-CIPP-R-EN-P-301
978-92-95113-13-8
ISBN APC Serial: APC-201910-CIPP-R-EN-DIGITAL-302

Country: 

Themes: