Incubating Feminist AI*

Incubating Feminist AI*

*Feminist AI= Algorithmic Decision-Making Systems and Artificial Intelligence harnessed to deliver equality outcomes, designed with inclusion at the core, creating new opportunities and proactive, innovative correction of inequities.

<A+> Alliance seeks to collaborate with communities, practitioners, researchers, scholars and innovators working to springboard from the established ‘what’ of describing harm and ‘why’ harms exist socially or technically, to research focussed on ‘how’ course correcting future harms can happen, and most urgently ‘how’ new data, algorithms, models, policies + systems could be created, then prototyped, and piloted. The 2021 call for proposals from multi-disciplinary, global south teams of feminist thinkers is intended to create opportunities for new research engagement. Novel approaches and new models, with a focus on practical application in prototype and pilot phases will be prioritized. Papers will focus on multi-disciplinary work researching, creating, or iterating new AI framework models and designs for new or current social policy allocations in order to re-engineer system thinking and algorithmic solving for proxy problems and old assumptions buried in the data and now active in the new technology. This may include research on the efficacy of including or hiding protected characteristics, or finding social and economic models of allocation/distribution now made more possible and equitable through the creation or innovative use of new or big data in Algorithmic Decision-Making (ADM).

We are interested particularly in how Algorithmic Decision-Making systems (ADM) in developing countries function to inhibit rights and amplify unequal power relations, and how AI and ADM can be harnessed to deliver equality outcomes and new opportunities. How can feminist methodologies and approaches be applied when developing AI systems? How can AI innovation and social systems innovation be catalyzed concomitantly to create a positive movement for social change larger than the sum of the data science or social science parts. Some indicative areas that intersect with core public policy agendas in the Global South such as Economic / Social Allocation: Health, Education, Justice, Social Protection/Benefits (including Subsidy, Digital ID, Land Tenure/Use, Pensions, Conditional Cash Transfer, Housing Lotteries), new algorithmic applications of Indigenous, Decolonial, Traditional models, etc, Geospatial Mapping,Climate Change, and Feminist Data collection or community-driven data stewardship models, Accountable Local, National or Regional AI Policy or Governance models are foreseen.

Eligibility, Research Directions, Criteria


Eligibility

  • Teams with two or more researchers from different disciplines will be very strongly encouraged, and in the case of thematic research collaboration across regions will be encouraged.
  • ​​The 2 Principal Investigators for each project, from different disciplines, must be domiciled in the Global South, although team mates not receiving funding may reside anywhere in the world.
  • f < a + i > r is dedicated to helping untraditional academic teams or community groups find partners, and will support teams with capacity building or technical assistance where appropriate.
  • Principal Investigators from each selected paper team will be required to attend bi-weekly meetings to discuss their approaches, insights and research progress throughout the duration of the project (including prototyping and pilot phases if selected). In addition to this South-South network knowledge sharing, teams (and selected applicants) will be invited to attend f < a + i > r regional and/or global meetings happening bi-monthly.

Call for Proposals

Opens: 15 September, 2021

Closes: 31 October, 2021 (Anywhere on Earth)

Determination: 15 November, 2021

Bi Weekly Meetings throughout process

Delivery: 15 May, 2022

Papers, Models, Prototypes, Pilots

  • 9 proposals will be selected to be developed into papers each year of the three-year project with best efforts to equally divide the annual 9 papers between the 3 focus regions of LAC, MENA, and SEAsia.

  • Out of the 9 papers delivered each year, 3 will be selected to go to Prototype.

  • One of those 3 prototypes will go to an official pilot. The process will be repeated each year through 2024.

Grants

  • CAD 8,000 per paper to the team with 2 Principal Investigators from 2 different disciplines, both based in the Global South

  • Prototype if selected: CAD 75,000 per prototype plus partner funding

  • Pilot if selected: CAD 50,000 to seed each partnered pilot

Research Directions

Incubating Feminist AI seeks to catalyze the creation of Feminist AI taking current social policy allocations to either redefine algorithmic allocation/justification or re-engineer system thinking and algorithmic solving for proxy problems buried in the data and emerging in the tech. We seek to focus on multi-disciplinary work that creates, or iterates new framework models and designs for:

Economic / Social Allocation

  • Articulating models of allocation/distribution that are made more possible or more equitable through the creation and/or innovative use of new or big data in ADMs.

  • Some indicative areas: Social Policy & Protection Design (p.ex., Social insurance, social assistance and safety nets: from Subsidy, Digital ID, Land Tenure/Use, Pensions, Conditional Cash Transfer, Lotteries), Health, Housing and Education services, Labour activation policies, Social Welfare, Justice, new algorithmic applications of Indigenous, Decolonial, Traditional models, etc, Geospatial Mapping, and Climate Change.

Feminist Data collection & deployment

  • Methodology for Inclusive Data Collection for broader representation, that adheres to Feminist Principle

  • Community-driven data stewardship models

  • Protocol documenting AI-relevant characteristics of datasets

Accountable Local National or Regional AI Policy or Governance models

  • Methodology for How small a dataset could be and still be effectively used in machine learning.

  • Intersectional Fairness for multiple sensitive attributes (e.g. more than 2, race and gender and class/disability/indigenous/ethnicity etc) mathematical models.

  • Design of a Feminist Digital ID

Feminist research

  • What would feminist research methodologies look like in AI? How can feminist methodologies result in different approaches to AI in developing countries? What questions would emerge, who would now be included, how would they be engaged, with what epistemologies?
  • How could this methodology influence the knowledge that was developed?

Data Collection

  • Where does traditional data collection go wrong? From a feminist point of view, what would inclusive data collection look like Before, During, After data is gathered?
  • What would be a methodology to create this? How could governments (and other actors) ensure/facilitate

Social Protections

  • How do social protections work in the developing world? And which ones specifically affect women and girls?
  • How is bias mitigated or amplified by underlying social protection assumptions, or development aid assumptions and in AI / Automated Decision-Making systems?
  • What would an ADM social protection system look like if designed with a feminist perspective?
  • What are private sector uses of ADM in developing country contexts that are discriminatory/biased

Cultural Norms

  • What are the combinations of norms, history, and procedure that perpetuate ways of working and that have led to amplification of existing inequalities?
  • How do these norms constrain or promote patterns of behaviour in communities/organizations generally, and AI / tech organizations specifically?
  • What are the forces and environments necessary for norm change for AI sector outcomes, and the AI sector itself?
  • How can the dynamics of norm change be incorporated into a feminist agenda for AI?

Partners

Supported by

Partners

Supported by