Activities of Maria da Graça CARVALHO related to 2020/2012(INL)
Shadow opinions (1)
OPINION with recommendations to the Commission on the framework of ethical aspects of artificial intelligence, robotics and related technologies
Amendments (21)
Amendment 6 #
Draft opinion
Paragraph 1
Paragraph 1
1. Underlines the importance of an EU regulatory framework of ethical aspects of artificial intelligence , robotics and related technologies being applicable where consumers within the Union are users of or subject to an algorithmic system, irrespective of the place of establishment of the entities that develop, sell or employ the system in order to bring legal certainty to business and citizens alike;
Amendment 11 #
Draft opinion
Paragraph 2
Paragraph 2
2. Notes that the regulatory framework should apply to algorithmic systems, including the fields of artificial intelligence, machine learning, deep learning, automated decision making processes and robotics;
Amendment 14 #
Draft opinion
Paragraph 2 a (new)
Paragraph 2 a (new)
2a. Stresses that an EU regulatory framework of AI shall have a human- centric approach and lead to development of systems which incorporate European ethical values by-design; considers that an EU regulatory framework that focuses on European values would be an added value providing Europe with a unique competitive advantage and make a significant contribution to the well-being and prosperity of European citizens and businesses, and boost our internal market;
Amendment 16 #
Draft opinion
Paragraph 2 b (new)
Paragraph 2 b (new)
2b. Underlines that an ethical framework of AI is an added value to promote innovation on the market;
Amendment 20 #
Draft opinion
Paragraph 3
Paragraph 3
3. Stresses that any future regulation should follow a differentiated risk-based approach, with clear criteria and indicators, followed by an impartial and regulated assessment based on the potential harm for the individual as well as for society at large, taking into account the specific use context of the algorithmic system; legal obligations should gradually increase with the identified risk level; in the lowest risk category there should be no additional legal obligations; algorithmic systems that may harm an individual, impact an individual’s access to resources, or concern their participation in society shall not be deemed to be in the lowest risk category; this risk-based approach should follow by clear and transparent rules;
Amendment 25 #
Draft opinion
Paragraph 3
Paragraph 3
3. Stresses that any future regulation should follow a differentiated risk-based approach, based on the potential harm for the individual as well as for society at largethe whole society, taking into account the specific use context of the algorithmic system; legal obligations should gradually increase with the identified risk level; in the lowest risk category there should be no additional legal obligations; algorithmic systems that may harm an individual, impact an individual’s access to resources, or concern their participation in society shall not be deemed to be in the lowest risk category; this risk-based approach should follow clear and transparent rules;
Amendment 35 #
Draft opinion
Paragraph 4
Paragraph 4
4. Underlines the importance of an ethical and regulatory framework including in particular provisions on the quality of data sets used in algorithmic systems, especially regarding the representativeness of training data used, on the de-biasing of data sets, as well as on the algorithms themselves, and on data and aggregation standardshigh-quality and, where possible, unbiased data sets in order to improve the output of algorithmic systems and boost consumer trust and acceptance;
Amendment 42 #
Draft opinion
Paragraph 5
Paragraph 5
5. Believes that consumers should be adequately informed in a timely, impartial, easily-readable, standardised and accessible manner about the existence, process, rationale, reasoning and possible outcome and reasoning of algorithmic systems, abouton how to reach a human with decision- making powers, and about how the system’s decisions can be checked, meaningfully contested and corrected;
Amendment 50 #
Draft opinion
Paragraph 5
Paragraph 5
5. Believes that consumers should be adequately informed in a timely, impartialaccurate, easily-readable, standardised and accessible manner about the existence, process, rationale, reasoning and possible outcome of algorithmic systems, about how to reach a human with decision- making powers, and about how the system’s decisions can be checked, meaningfully contested and corrected;
Amendment 53 #
Draft opinion
Paragraph 5 a (new)
Paragraph 5 a (new)
5a. Considers that a value-sensitive design approach is strongly needed to allow a widespread social acceptance of AI for consumers; considers that ethical values of fairness, accuracy, confidentiality and transparency should be the basis of AI which in this context entails that the system’s operations cannot generate unfairly biased outputs;
Amendment 58 #
Draft opinion
Paragraph 6
Paragraph 6
6. Recalls the importance of ensuring the availability of effective remedies for consumers and calls on the Member States and national market surveillance authorities to ensure that accessible, affordable, independent and effective procedures and review structures are available to guarantee an impartial human review of all claims of violations of consumer rights through the use of algorithmic systems, whether stemming from public or private sector actors;
Amendment 64 #
Draft opinion
Paragraph 7
Paragraph 7
7. Stresses that where public money contributes to the development or implementation of an algorithmic system, the code, the generated data -as far as it is non-personal- and the trained model should be public by default, to enablguarantee transparency and reuse, among other goals, to maximise the achievementenable their reuse to foster innovation; stresses that in this way, the full potential of the Single Market, and to can be unlocked avoiding market fragmentation;
Amendment 65 #
Draft opinion
Paragraph 7
Paragraph 7
7. Stresses that where public money contributes to the development or implementation of an algorithmic system, the code, the generated data -as far as it is non-personal- and the trained model should be public by defaultaccessible, to enable transparency and reuse, among other goals, to maximise the achievement of the Single Market, and to avoid market fragmentation;
Amendment 67 #
Draft opinion
Paragraph 7 a (new)
Paragraph 7 a (new)
7a. Recalls that an examination of the current EU legal framework, including the consumer law acquis, data protection legislation, product liability legislation, product safety and market surveillance legislation, is needed to check that it is able to respond to the emergence of AI and automated decision-making and that it is able to provide a high level of consumer protection;
Amendment 85 #
Draft opinion
Paragraph 8
Paragraph 8
8. Underlines the importance of ensuring that the interests of marginalised and vulnerable consumers and groupconsumers in vulnerable situations are adequately taken into account and represented in any future regulatory framework; notes that for the purpose of analysing the impacts of algorithmic systems on consumers, access to non-personal data shcould be extended to appropriate parties notably independent researchers, media and civil society organisations, while fully respecting Union data protection and privacy law; recalls the importance of training and giving basic skills to consumers to deal with algorithmic systems in order to protect them from potential risks and detriment of their rights;
Amendment 86 #
Draft opinion
Paragraph 8
Paragraph 8
8. Underlines the importance of ensuring that the interests of marginalised and vulnerable consumers and groupconsumers are adequately taken into account and represented in any future regulatory framework; notes that for the purpose of analysing the impacts of algorithmic systems on consumers, access to data should be extended to appropriate parties notably independent researchers, media and civil society organisations, while fully respecting Union data protection and, privacy law and business secrecy rules; recalls the importance of training and giving basic digital skills to consumers to deal with algorithmic systems in order to protect them from potential risks and detriment of their rights;
Amendment 89 #
Draft opinion
Paragraph 9
Paragraph 9
9. Underlines the importance of training highly skilled professionals in this area and ensuring, the need of having diverse teams of developers and engineers working alongside with key actors to prevent gender and cultural bias of being inadvertently included in AI algorithms, systems and applications and ensure the mutual recognition of such qualifications across the Union;
Amendment 104 #
Draft opinion
Paragraph 10
Paragraph 10
10. Calls for the Union to establish a European market surveillance structure for algorithmic systemcomposed of national market surveillance authorities issuing guidance, opinions and expertise to Member States’ authorities;
Amendment 110 #
Draft opinion
Paragraph 11
Paragraph 11
11. Notes that it is essential for the software documentation, the algorithms and data sets used to be fully accessiexplainable to market surveillance authorities, while respecting Union law; invites the Commission to assess if additional prerogatives should be given to market surveillance authorities in this respect;
Amendment 115 #
Draft opinion
Paragraph 12
Paragraph 12
12. Calls for the designation by each Member State of a competent national authority for monitoring the application of the provisions related to algorithms;
Amendment 122 #
Draft opinion
Paragraph 13
Paragraph 13
13. Calls for the establishment of a European market surveillance board for algorithmic systems, to ensure a level playing field and to avoid fragmentation of the internal market, to decide with a qualified majority and by secret vote in case of different decisions on algorithmic systems used in more than one Member State, as well as at the request of the majority of the national authorities;