Measuring what matters: Embedding gender equality in AI governance through gender-specific indicators

Nandini argues that measurement is a precondition for accountability and policy attention. If gender is absent from monitoring frameworks, it is effectively sidelined in digital governance implementation, including in AI-related processes.

She states directly that "currently, we don't measure progress on gender equality at all" and links this to the broader principle that what is not measured is not taken seriously and therefore does not progress . She also gives a concrete example: the UNGIS draft joint implementation roadmap presented to the CSTD in April contains the word "gender" zero times, which she uses to show institutional omission at a high policy level .

Nandini argues that measuring access alone is too narrow for assessing gender equality in AI. She says governance should be the central lens because the real question is whether women can shape innovation, benefit from it, and share in the value created by AI systems.

She says that "governance is the meta indicator that we need to track" because the issue is not merely whether women access AI services, but whether they can "meaningfully realize the value potential of AI innovation" and whether AI trajectories improve women's lives, including different groups of women across social inequalities .

Nandini proposes practical, actionable indicators that can be inserted into WSIS action lines, especially those related to governance and enabling environments. Her focus is on making gender visible in national strategies and in leadership structures, rather than leaving it as an afterthought.

Under Action Line C1, she proposes measuring whether gender equality appears as a standalone resource chapter in national digital and AI strategies rather than merely as a cross-cutting footnote, and notes that UNDESA's national digital strategy review could be used for this purpose . She also proposes an indicator on the percentage of WSIS action line facilitating bodies and national delegations meeting gender parity in leadership, suggesting that ITU Secretariat records and facilitator appointment data could supply the information .

Nandini argues that relying on private markets will not correct gender inequality in innovation. She contends that public funding tied to gender equity is needed because women-led innovation receives very little venture capital, while stronger public research systems are associated with better female participation in patenting.

She proposes measuring the percentage of public R&D and innovation funding in ICTs and AI that has gender equity conditions attached, drawing on OECD DAC gender markers and national science and technology budgets . She supports this by noting that women-led start-ups receive less than 2% of global venture capital funding . She further cites WIPO-based analysis showing low women's inventor rates in digital sectors and observes that countries performing better on female patent participation tend to have robust public R&D systems, with more filings coming from universities and state-owned enterprises where gender ratios are more balanced .

Nandini argues that gender measurement in AI should not stop at sectoral or technical indicators. It must also capture wider labour-market and social changes, especially how AI-driven transformation may deepen unequal work patterns and unpaid care burdens for women.

She calls for tracking economy-wide statistics on women's share in the working-age population, employment, management, and unpaid care work using ILO statistics, UN Women databases, and national time-use surveys . She explains that this is necessary to assess how digital servicification and AI-driven labour restructuring affect gender power relations, and warns that evidence from majority world countries already suggests labour-market polarisation, gigification, and reinforcement of unequal unpaid care work burdens .

Nandini argues for a much broader conception of AI harm than simple representational bias in models. She says monitoring should cover the whole AI production chain, including environmental and labour impacts that disproportionately affect vulnerable women.

In her closing intervention, she refers to UNESCO's Outlook Study on AI and Gender, saying it highlights the need for incidence reporting mechanisms with robust taxonomies of gendered harms in AI innovation systems . She adds that IT4Change believes this analysis must expand beyond model-level representation harms to include environmental and labour rights violations across the AI stack, specifically naming data centres and critical mineral mining as examples of material layers where vulnerable women may be affected .

Nandini argues that indicator design should not be treated as a purely technical exercise. The purpose and users of indicators matter because they shape what is measured and how the information will be used in policy, advocacy, and accountability processes.

She explicitly says that it is important to ask not only what is being measured, but also why the indicators are being measured, how they will be used, and by whom . She then gives examples of possible users, including governments for policymaking and civil society and academia for advocacy and advancing the message .

Marina argues that the need for gender indicators follows directly from commitments already made in the WSIS Plus 20 process. Since the outcome document affirms gender equality and requires action line facilitators to address it, indicators are necessary to make those commitments operational and trackable.

She explains that the WSIS outcome document reaffirms gender equality and, in paragraph 112, calls on all Action Line facilitators to address gender equality and the empowerment of women and girls as a core theme within their work . She then says that to do this, it is very important to develop specific gender indicators, especially because these indicators are currently being drafted by WSIS action line facilitators .

Marina presents the coalition's indicators as an initial framework rather than a rigid universal template. She emphasises that they can provide a common baseline within WSIS while still being adjusted by countries and regions according to local priorities and realities.

She says the coalition has proposed indicators for all action lines and describes them as "a baseline" and "an initial proposal" . She also stresses that gender-responsive policies must be contextually grounded and localised, and that countries and regions should consider additional specific matters to measure beyond general indicators in their own implementation processes .

Marina argues that the proposed indicators have a wider practical use than formal state reporting. While governments may be the main implementers through WSIS action lines, companies and other actors can also use the indicators as guidance for more inclusive and feminist AI practices.

Responding to a question from the Zoom chat, she says the indicators were designed to be embedded into the action lines, so governments and national bodies are likely to be the main recipients . She immediately adds that private companies and other stakeholders can also use them in practice, alongside feminist principles, as a path for building gender-responsive frameworks, even if not every indicator will apply to every company .

Mrinalini argues that positive claims about AI are hollow if women and girls cannot participate safely and equally in digital spaces. In her view, exclusion, insecurity, and disproportionate harm fundamentally undermine the idea that AI is universally beneficial.

She says that if women and girls cannot participate safely online, they cannot equally benefit from digital infrastructure, education, innovation, or entrepreneurship . She also recounts hearing country representatives describe AI as a catalyst for change and a reliable technology, but rejects that framing if it leaves half the global population unsafe and disproportionately treated .

Mrinalini presents the coalition's Feminist Guiding Principles as a concrete framework that governments and action line facilitators can use to turn gender commitments into implementation. She stresses that the principles are broad enough to address key AI governance issues while also including recommendations and accountability tools.

She identifies the coalition's Feminist Guiding Principles for AI Governance as a starting point for making indicators a reality and says they can be used by governments and action line facilitators . She describes the seven principles as covering safety, dignity, interoperability by design, algorithmic bias, human rights, tech-facilitated gender-based violence, inclusive and equitable AI, investment for feminist AI, and feminist AI governance mechanisms, with recommendations including independent mechanisms, pre- and post-human rights assessments, and public audits .

Mrinalini argues that the coalition is not proposing something unprecedented. She says there are already multiple international legal and policy frameworks that recognise the need for proactive gender equality measures in digital and AI governance.

She explicitly points to CEDAW, including Article 4.1 on temporary special measures, the Commission on the Status of Women, the EU Gender Equality Strategy, and the Council of Europe's Framework Convention on AI as relevant precedents . She concludes from this that frameworks already exist which address AI and gender equality and which action line facilitators should begin examining .

Mrinalini argues that the process of designing and implementing indicators must include a far wider set of actors than conventional technical experts. She sees inclusive expertise as necessary for legitimacy, accountability, and for ensuring that indicators reflect lived realities of harm and exclusion.

She says a multi-stakeholder approach is key because it helps build shared values and a shared agenda . She then lists the kinds of expertise that should be included under the coalition's governance principle: survivors of technology-facilitated gender-based violence, feminist technologists, labour advocates, human rights defenders, advocates working on gender-based violence, indigenous leaders, persons with disabilities, anti-racist advocates, public policy experts, and community organisations .

Mrinalini argues that implementation should begin immediately through existing consultation processes rather than waiting for a later stage. She encourages civil society actors to take the coalition's proposals into current meetings with action line facilitators and to broaden support through alliances.

In her closing remarks, she notes that action line facilitators are hosting consultations during the week and encourages people to take the coalition's message and proposed indicators into those meetings . She also says the coalition is small and therefore needs to build allies and partners beyond its own membership in order to move the proposals forward .

Pavitra argues that gender cannot be treated as a side issue in digital governance. Because inequality is built into technological systems and governance structures, gender equality has to be integrated as a core organising principle across connectivity, AI, infrastructure, and data governance.

She says that although there is broad recognition of commitments to gender equality in processes such as the Global Digital Compact and WSIS Plus 20, these processes often fail to acknowledge that the barriers are structural and embedded in technologies . She lists unequal access, algorithmic discrimination, tech-facilitated gender-based violence, exclusion from decision-making, and disproportionate impacts from data-driven systems as continuing realities . She therefore argues that gender mainstreaming must be embedded across digital governance processes and that gender equality should be seen as foundational rather than merely cross-cutting .

Pavitra argues that effective gender-responsive indicators must assess substantive outcomes, not just whether women are present or connected. She proposes indicators that examine women's real influence in governance, their role in knowledge production, and whether AI systems are assessed for multiple kinds of harm before deployment.

She says indicators must go beyond measuring participation or access alone and should capture outcomes . As examples, she proposes measuring the percentage of women in decision-making roles in national digital or AI governance authorities under Action Line C1, the gender gap in who produces versus consumes digital content and knowledge under C3, and the percentage of AI systems deployed in public services that have undergone integrated human rights, gender, environmental, and labour impact assessments with gender separately reported under C10 .

Pavitra argues that one common objection to gender-responsive indicators is overstated. She says substantial internationally standardised data already exists, and the real challenge is connecting and adapting those datasets to WSIS and related digital governance commitments.

She says it is "not entirely true" that a whole new data infrastructure is needed, because some relevant data already exists and is published internationally in standardised form, even if it is not yet linked to WSIS . She gives three examples: WIPO Patentscope, which provides gender-disaggregated data on inventors and patent filings in technology fields ; the ITU gender dashboard, which tracks representation, decision-making, access, and leadership, including women ICT ministers and regulators ; and the WHO Global Digital Health Monitor, which includes a specific indicator on whether digital health strategies and governance explicitly account for gender considerations . She concludes that these existing tools mean governments can start measuring now and later refine the framework .

Pavitra argues that gender indicators will be inadequate if they treat women as a uniform category. To be meaningful, measurement must reflect how gender interacts with other social hierarchies and forms of exclusion.

She states that indicators must adopt an intersectional approach because gender does not operate in isolation . She then lists race, ethnicity, disability, age, migration status, socioeconomic background, and sexual orientation as factors shaping exclusion, and warns that indicators ignoring these realities will overlook those most affected by digital inequalities .

Pavitra argues that the growing number of digital governance processes creates a risk of fragmented implementation. She proposes a joint roadmap with shared priorities and coordinated reporting so that overlapping commitments on gender equality reinforce one another instead of producing parallel systems.

She says that broad commitments across the GDC, WSIS Plus 20, and the Global Dialogue on AI risk becoming fragmented if each process develops its own principles, reporting mechanisms, and implementation pathways . She argues that stakeholders should identify overlapping commitments, develop common priorities and shared indicators where appropriate, and coordinate reporting mechanisms to reduce duplication, ease the reporting burden on governments, and preserve policy coherence and a common baseline for collective progress .

Eva argues that the proposed indicators have value beyond WSIS and can support accountability under regional legal systems. She is interested in adapting them for use in the Inter-American framework to monitor how states implement obligations relating to digital violence against women and AI.

She explains that at the Organization of American States they need to develop indicators to measure implementation by states parties, particularly in light of developments in artificial intelligence and digital violence against women . She says they would like to review the coalition's indicators and potentially work together to adapt them for their fifth multilateral evaluation round under the Inter-American Convention, covering 33 countries and supporting state accountability .

Esperanza argues that institutional work is already under way to align indicator systems across major global frameworks. She stresses that stakeholders should participate in the consultation process so that the resulting indicators are robust and internationally comparable.

She describes the work of the Partnership on Measuring ICT for Development as a global initiative tasked by the UN General Assembly resolution on measurement and monitoring . She says the partnership is mapping current core ICT indicators against the GDC, WSIS action lines, and the SDG indicators, and invites all stakeholders to join a consultation session and written questionnaire that will inform the report to be submitted to the CSTD in April 2027 .

Esperanza argues that indicator design must balance political ambition with statistical usability. A key requirement is that the resulting measures produce data that can be meaningfully compared internationally, otherwise they will be less useful for global monitoring.

She underlines "the importance of international comparability" and says that when discussing indicators, stakeholders must consider the need for data that can be compared across countries .

Nagla argues that indicators and indices are only useful when they faithfully reflect realities on the ground. She warns against shallow methodologies that encourage procedural compliance rather than substantive understanding.

She says indices are "super useful" but only if they are accurate, representative, nuanced, deep, and expressive of what is happening on the ground, rather than being used simply to tick boxes in response to methodologies . She also commends the coalition's work precisely because few indices incorporate gender equality by design .

Nagla argues that global indices can create perverse incentives if governments pursue higher rankings instead of genuine reform. She therefore recommends supplementing top-down measures with alternative approaches that better capture local realities.

She references her own experience working with indices and governments, including the Global Innovation Index, and says she has argued in research that top-down indices do not capture realities and should be complemented by alternative measures . She warns of a danger that governments focus on ticking boxes so they can move up global rankings, rather than solving underlying problems . She also points to the Global Index on Responsible AI, which includes dimensions such as inclusion and diversity, ethics and sustainability, labour and skills, trust and safety, and public delivery, with subdimensions including gender equality, children's rights, disability rights, and cultural and linguistic diversity .

The audience speaker argues that, whatever the normative case for gender indicators, implementation on the ground can face institutional resistance. In sectors like cybersecurity, collecting governance-related gender data may be politically or administratively difficult, so a staged strategy may be more realistic.

The speaker, identifying as an ITU project lead working on capacity-building focused on women, says that when trying to include indicators such as the percentage of women in leadership roles within ICT and AI governance authorities, it is "hard to get" and faces strong pushback . The speaker adds that they have been unable to obtain gender-specific indicators in a cybersecurity index and therefore suggests envisioning several steps before full implementation or adding more granular layers to make stakeholders more comfortable .