Artificial Intelligence experts and policymakers discuss inclusion in AI policy development
Inclusive and participatory processes help mobilize the collective intelligence of communities, to ensure that voices and concerns of disadvantaged groups are heard and that they are involved in key decision making.
UNESCO and the Innovation for Policy Foundation (i4Policy) conducted a series of workshops with AI experts and policymakers from different parts of the world to exchange ideas on inclusion and participation in AI policy development.
AI policy design should be human-centered
In the workshops, the effects of AI on human rights and fundamental freedoms were emphasized. As per Jake Okechukwu Effoduh, Nigerian lawyer and activist, “technology is not neutral. It can either advance or infringe on human rights”. He added that, “a broad range of human rights protections are now affected by the deployment and use of AI”. For example, while it may be necessary for security reasons, the use of facial recognition technology in public may affect democracy by taking away the space to protest, according to Jibu Elias, Research Lead at INDIAai.
As such, policymakers should focus on the human rights aspect in designing AI policies, said Wanda Munoz, independent public policy and human rights expert from Mexico. In addition to policymakers, other actors in society have a part to play. “To guide legislative intervention, scientists must strive to paint an explainable picture of the technology’s impact. Civil society and the media are crucial towards the legitimization of the technology because they can influence public awareness and acceptance,” said Effoduh.
AI policy design should be inclusive
A major consideration in AI policy design that can exclude some communities from benefitting from its use is the under-representation of groups that are technologically disengaged. To bridge the digital divide, there should be a “creation of mutual language and understanding of AI”, said Dr. Katherina E. Höne who leads research at the Diplo Foundation that trains diplomats on technology policy issues. Presently, there is no fixed definition of AI or common technical language, making it difficult to achieve common understanding, according to Alexander Heußner from the Federal Ministry of Education and Research in Germany (BMBF).
Eddan Katz, a researcher based in the US, emphasized that “[d]iversity of datasets is of primary importance, and we should consider AI policy discussions an opportunity to make diversity inherent to ethics in technology.” However, achieving diversity and inclusion in AI Policy Design is not an easy feat. Isidora Edwards, an advisor to the Undersecretary of Economy at the Ministry of Economy, Development and Tourism of Chile, highlighted “inefficiencies in government and different political agendas” as another challenge.
In addressing these issues, experts warned against tokenism in inclusion.
AI Policy Design should be Multistakeholder-Driven
In light of UNESCO’s R.O.A.M (Rights, Openness, Access and Multistakeholder) framework which underlines multi-stakeholder approaches as an important aspect for adaptive governance and collaboration, the expert groups also discussed participatory processes for AI policy. “A multistakeholder approach is necessary because it leads to equality in the AI ecosystem where data flows are protected,” said Philip Sauerbaum, Policy Officer for Digitalization at European Commission.
José Gustavo Sampaio Gontijo, the Director of Brazil’s Department of Science, Technology and Digital Innovation, pointed out that “for now, AI itself is only a sequence of codes completely relying on data inputs. Therefore, the challenge for policymakers is to ensure open data access and accountability.” To illustrate, in Brazil, anonymized health data could be collected and used to train healthcare providers, making them publicly available for AI innovators to improve the design of AI-enabled telemedicine applications.
José Antonio Guridi, from the Ministry of Economy in Chile, noted that “the multi-stakeholder approach itself can be an education process.” A multi-stakeholder approach allows different stakeholders to come together and learn from their exchanges on AI issues.