Written Evidence Submitted by Center for Feminist Artificial Intelligence

(GAI0013)

Introduction into the AI Act:

The use of Artificial Intelligence (AI) opens a field of tension between innovational potential and the need for regulation. However, though heavily debated, harmonized guidelines or standards for regulating this technology are currently in its infancy. At the same time, there are cases of discrimination in which the need for appropriate regulatory frameworks becomes clear. A political milestone in the development of a set of rules comes from the European Union. A Draft of the Artificial Intelligence Act of the European Commission has been proposed in April 21, 2021.[1] The suggested law attempts a balancing act to further promote innovation while at the same time safeguarding ethical principles such as non-discrimination.

Scope:

Artificial intelligence is seen as the growth lever of the 21st century. With an estimated market volume of €277.9 billion worldwide and a five-year compound annual growth rate of 17.5 % in 2021, AI can no longer be excluded from discussions about progress.[2] If this forecast is correct, services and products related to AI will grow to a market volume of € 621.5 billion in 2026.

Meanwhile, it is estimated that the risk-based approach of the AI Act (ranges from no risk, low, high risk, and forbidden techniques) will have tremendous influences on EU and international businesses. Killian Gross, Head of Unit for AI Policy and responsible for the development and coordination of AI policy in EU, estimates that the AI high risk definition will affect 10 - 15 % of all European companies. This demonstrates the scope of impact of this upcoming regulation.

Problem statement:

Analyzing the current AI environment regarding the EU AI Act developments, two main issues can be identified.

(1)   In its current version, as concluded by NGOs and civil society, the protection of human rights does not go far enough.[3] Indeed specific requirements related, for example, to environmental sustainability, accessibility for persons with a disability, stakeholders participation in the design and development of the AI systems and diversity of development teams on the basis of clear objectives and key performance indicators to measure the achievement of those objectives are only encouraged on a voluntary basis (cf. proposed Codes of Conduct, Art. 69 AIA). Moreover, neither gender mainstreaming nor specific articles on gender or other minoritized groups are part of the current draft.

(2)   At the same time, the EU AI Act in its current version focusses due to its risk-based approach on mitigation (negative business case) only.[4] Incentive systems for the promotion of positive business cases in line with the goal to establish an ecosystem of trustworthy AI applications are largely missing. This is especially worrisome as several experts have pointed out AI’s potential to achieve the Sustainable Development Goals.[5]

Solutions:

(1)   We draw attention to the risk that the human rights-based approaches aiming for equality may dilute in the field of tension between excellence and trust and point out that current discussions about discrimination of AI Systems do not take an inclusive, feminist perspective in most cases. Intersectional discrimination i.e. due to gender and class, gender and ethnicity, sexual orientation do not get the attention required to solve major human-rights injuries.

Benchmarking the landscape of digital policy with foreign policy, a feminist lens led to a human-centric and inclusive way to bring foreign policy into practice[6]. Hence a feminist approach in addition to existing human-rights based approaches in digital policy and AI is mandatory to enable more equality.

(1)   Although taking a risk-based approach is mandatory, the positive business case of AI is underestimated. AI is a main force to support the SDGs. AI and digital policy can be an enabler for equality as it analyzed biases in data that decode our world. Thus, AI can leverage the opportunity to support human-rights in different manners as it showcases where change is required. Furthermore, taking a feminist lens on AI allows to extend business models and costumer groups. If a product launch is based on historical data that include i.e. white cis-male people only, optimizing a product for an additional target group with specific needs will be overlooked. As a result, the positive business case of AI should be in focus for companies.

Definition of FemAI:

A definition of feminist AI is required to base our work on a solid fundament and to be able deriving a holistic product and service offering:

Feminist Artificial Intelligence (AI) describes an inclusive, intersectional approach to lift the potential of AI to create equality and a better life for all. It particularly focuses on the role of women, LGBTIQ+, and other marginalized and underrepresented groups regardless of their gender, age, religion and belief, disability, sexual identity, ethnicity, and appearance.

The feminist AI approach seeks to establish a contemporary and holistic relationship between human-machine systems and machine-machine systems.

It focuses on overcoming patriarchal structures, systemic privileges, and power imbalances in society, politics, and economy.

Through a feminist approach, AI can be a driver of the Sustainable Development Goals (SDGs) and make a critical contribution to a more sustainable future. Feminist AI should be understood as part of the effort to achieve the SDGs.

Products and Services of FemAI:

Based on this analysis, the following products and services will be offered by our thinktank (all aspects in italic are ongoing, currently bro-bono activities):

A)     Network: feminist AI and digital policy roundtable to bring all relevant perspectives together to work on ideas and concepts to fulfill our mission of a feminist digital world.

B)     Public Sector: Attending conferences, writing positioning papers, speaking to key stakeholders in order to influence the EU AI Act process and its execution towards a feminist lens (This perspective is currently completely missing!)

C)     Civil Society: Educational events to enlighten about the impact of AI focusing on both risks but also on opportunities.

D)     Private Sector: offering workshops to sensitize for discriminatory AI systems beyond gender, engage in digitalization projects to represent a feminist lens, working on AI Governance (Product Idea: Feminist AI Playbook)

Annex 1

(November 2022)