From Data Asymmetry to Equitable Stewardship:
The Gender Imperative for Development
CSTD 2026–2027 Inter-sessional Panel
November 5 | Palais des Nations, Geneva
Gender Advisory Board to the CSTD
Data is becoming foundational infrastructure for artificial intelligence, public services and economic development. Yet the data record on which these systems rely is profoundly uneven. Women’s bodies remain underrepresented in clinical evidence; their unpaid and informal economic activity is largely absent from official statistics; and women in low-resource language communities are excluded both by language and by gender within the limited data that exists.
These are not static gaps that can be corrected gradually. Under current AI development practices, data asymmetries can compound across successive model generations. Research on recursive training demonstrates that sparse and underrepresented parts of a data distribution are lost first, while existing distortions can be amplified. Continued access to fresh, provenanced human data is the identified safeguard, but it cannot recover knowledge that was never documented in machine-usable form. Each new generation built on an incomplete record increases the cost of correction, until correction becomes a problem of rebuilding the foundation rather than improving the data.
This side event will present the CSTD Gender Advisory Board’s proposed framework of data symmetry, comprising four interdependent dimensions:
Representational symmetry: The populations affected by a system must be present in the data used to build and evaluate it, at levels sufficient to support comparable performance. Representativeness is a condition of scientific validity, not an optional fairness measure. A model trained principally on one sex is not a general model with a gender deficit; it is a model of that sex, misdescribed.
Distributive symmetry: Economic and social value must return to the people and communities whose knowledge, labour and data make digital systems possible. This requires recognition and remuneration of data work, meaningful ownership and governance at source, and equitable benefit-sharing where community-derived data generates downstream commercial value.
Temporal symmetry: The representational quality of the data record must improve, or at minimum not deteriorate, across successive uses and model generations. This requires provenance documentation distinguishing human-generated from synthetic data, together with longitudinal measurement of which populations and forms of knowledge are being preserved or lost.
Procedural symmetry:Procedural symmetry, where women have equal voice in conceptualizing, implementing, and overseeing the mechanisms, structures, roles, institutional arrangements and policies that constitute data governance.
The session will examine the implications of these four dimensions for the CSTD’s wider work on science, technology and innovation for development. It will consider how questions of representation, ownership, benefit-sharing, interoperability and data flows affect the ability of countries to build inclusive innovation capacity and derive equitable value from data and AI.
Particular attention will be given to developing countries, which are frequently sources of data and sites of technology deployment, but remain underrepresented in decisions about standards, infrastructure, ownership and the distribution of economic value. The discussion will also consider why gender-responsive data governance is not a sectoral or specialist concern, but a condition for valid science, effective public policy and sustainable economic transformation.
The session will discuss practical implications and possible directions for CSTD consideration, including:
- recognising representativeness as a scientific validity requirement for AI systems used in public and high-risk functions;
- strengthening demographic coverage and data-provenance documentation;
- measuring representational coverage and degradation across successive model generations;
- preserving provenance and disaggregation information when data moves across systems and borders;
- supporting community data stewardship as development infrastructure; and
- ensuring that those who generate, collect and steward data participate in its governance and share in the value it creates.
The objective is not simply to add gender language to data governance. It is to ensure that the CSTD’s consideration of data, AI and development addresses four foundational questions: who is present in the data record, who has authority over it, who receives the value it creates, and whose knowledge survives into the future.