- AI technology must be trained using unbiased data sets to prevent discrimination
- Ethical framework needed for algorithms to reduce filter bubbles and protect diversity
- Teachers must retain control over decisions impacting students’ future opportunities
In a resolution adopted on Wednesday European Parliament calls to ensure that AI technologies are trained in a way that prevents gender, social or cultural bias and protects diversity.
In a resolution adopted by 623 votes in favour, 12 against and 61 abstentions, MEPs say that the use of artificial intelligence (AI) technologies in education, culture and the audiovisual sector could have an impact on “the cornerstones of the fundamental rights and values of our society”. Therefore all technologies must be regulated and trained so to protect non-discrimination, gender equality, pluralism, as well as cultural and linguistic diversity.
Human control in education
Addressing the different aspects of AI technology use in education, MEPs stress that teachers must always be able to correct decisions taken by the AI’s, especially regarding student selection and evaluation. At the same time, they highlight the need to enhance digital skills across Europe and train teachers to prepare for AI technology use in education.
Teachers must never be replaced by AI technologies, especially in early childhood education, they warn.
Regulate media algorithms
To prevent algorithm-based content recommendations from negatively affecting the EU’s cultural and linguistic diversity, MEPs ask for specific indicators to be developed to measure diversity and ensure that European works are being promoted.
The Commission must establish a clear ethical framework for how AI technologies are used in EU media to ensure people have access to culturally and linguistically diverse content. Such a framework should also address the misuse of AI to disseminate fake news and disinformation, they add.
Teaching EU values to Artificial Intelligence
The use of biased data that reflect already existing gender inequality or discrimination should be prevented when training AI, MEPs urge. Instead, inclusive and ethical data sets must be developed, with the help of stakeholders and civil society, to be used during the “deep learning” process.
"For years we have fought against discrimination and to retain our cultural and linguistic diversity, and these values need to be reflected in algorithms and uses of AI technologies", said the rapporteur Sabine Verheyen (EPP, DE) in the plenary debate preceding the vote.
"Artificial intelligence should be trained to not replicate discrimination". She also spoke about the need to reduce the so called “filter bubbles” and enable transparency for consumers, so that “people can get a better idea of the culture on offer."
Speaking of the need to ensure human-centered approach in the education sector, she said: "The human being needs to be in the center. They need to always have responsibility and control and the final say".