StarApple AI | Dr. Shirley Budall | May 19, 2025

Designing Equitable AI Systems for the Caribbean: A Policy Framework

Q: What is the UNESCO Guidance for Generative AI and why does it matter for Caribbean governments?

UNESCO's 'Guidance for Generative AI in Education and Research', published in September 2023, provides governments with a framework for managing generative AI in educational settings, including attention to equity and inclusion. For Caribbean governments deploying AI tools in schools and universities, the Guidance provides a starting point for ensuring that AI-enhanced education does not widen existing gender and socioeconomic gaps. Its principles on transparency, human oversight, and equity should be incorporated into national education technology policies.

Gender clauses bolted onto technology policy will not produce equitable AI; only frameworks that place Caribbean women's lived experience at the centre of design from the outset stand a chance of doing so.

When a gender equity critique lands on a technology policy, governments tend to add a paragraph, a clause, or, on a good day, a commissioned study. What they rarely do is go back to the design assumptions baked into the policy itself and ask whether those assumptions were ever consistent with gender equity in the first place.

I argue that this pattern is the central obstacle to equitable AI in the Caribbean. My hypothesis is that equity in AI cannot be achieved by appending gender language to frameworks built around different assumptions. It requires a distinct policy architecture that places the lived experience of Caribbean women, particularly those facing intersecting disadvantages of gender, race, class, and geography, at the centre of AI system design rather than treating them as an afterthought to be accommodated in a footnote.

What follows sets out what such a framework would actually look like: its core principles, its institutional requirements, its legal basis, and the specific elements that set it apart from the gender-clause approach that has produced so little result in other policy domains.

Why the Existing Approach Is Insufficient

Caribbean technology policy, like technology policy in most of the world, was designed primarily by people who were not thinking about gender. This is not a criticism of the individuals involved; it is an observation about institutional default positions. The CARICOM Regional Digital Economy Policy Framework, various national ICT policies, and emerging AI strategy documents across the region share a common structural characteristic: they treat equity as a value to be affirmed rather than a design requirement to be specified.

Affirming equity produces language like "the Framework commits to inclusive digital development" or "AI systems should be developed in accordance with principles of non-discrimination." These commitments are not wrong. They are simply insufficient. They do not specify what data collection requirements would ensure that AI systems represent women's experiences. They do not establish procurement standards requiring gender bias testing. They do not create accountability mechanisms for identifying and correcting equity failures after deployment.

The EU AI Act, whose prohibition provisions became effective on 2 February 2025, takes a categorically different approach. It does not merely affirm non-discrimination as a value; it specifies legal requirements, risk classifications, technical standards, and enforcement mechanisms that operationalise non-discrimination as a legal obligation. Caribbean policy should take this structural approach, applied to the specific circumstances of Caribbean women rather than imported wholesale from a European context.

UNESCO's "Guidance for Generative AI in Education and Research", published in September 2023 and monitored for implementation since, demonstrates what sector-specific application of equity principles looks like. The Guidance does not simply say that AI in education should be equitable. It addresses curriculum representation, assessment fairness, data privacy for students, and the specific risks that generative AI poses for students from disadvantaged backgrounds. Caribbean education ministries deploying AI tools in schools need this level of specificity, not aspirational language about inclusion.

The Core Principles of an Equitable AI Framework

An equitable AI framework for the Caribbean must rest on several principles that are distinct from, though compatible with, general AI governance principles. I set these out here as the foundation for the more specific recommendations that follow.

Representational adequacy. AI systems deployed in the Caribbean must be required to demonstrate that their training data adequately represents Caribbean women, including women in rural areas, women in informal employment, and women from marginalised communities. Systems that cannot demonstrate this adequacy should not be deployed in contexts where their performance gaps will harm those populations. This is a procurement requirement, not merely a design aspiration.

Participatory design. Caribbean women must be involved in the design and governance of AI systems that affect them, not as consultees at the end of a design process but as participants in defining the problem the AI system is meant to solve. This requires funded mechanisms for community participation, not simply open comment periods that favour technically literate respondents.

Transparency with comprehensibility. The EU AI Act's transparency requirements are a necessary but insufficient foundation for Caribbean contexts. Transparency that produces 200-page technical documentation is not meaningfully accessible to a woman in Kingston who wants to understand why an AI system denied her a bank loan. Comprehensible transparency, in plain language and in accessible formats, must be specified as an explicit requirement.

Accountability with remediation. An equitable framework requires not only that harms be identified but that they be remedied. Accountability mechanisms that document harm without correcting it protect institutions rather than individuals. Caribbean AI governance frameworks must specify remediation obligations alongside reporting obligations.

The Legal Architecture

An equitable AI framework for the Caribbean does not require starting from scratch. It requires building deliberately on existing legal foundations. The Jamaica Data Protection Act 2020 provides data rights that can be extended to address AI-specific harms, including algorithmic discrimination and automated decision-making. The Employment (Equal Opportunities) Act provides a basis for addressing gender discrimination in AI-mediated hiring. The Consumer Protection Act provides a basis for addressing unfair practices in AI-driven commercial services.

What is missing is the connective tissue: the sector-specific rules that specify how general rights and protections apply to AI contexts. Jamaica should enact an AI Governance Act or an equivalent suite of amendments to existing legislation that addresses: mandatory bias auditing for high-risk AI systems, a right to explanation for automated decisions affecting individuals, obligations for gender impact assessment before public sector AI deployment, and a regulatory authority with AI-specific expertise and enforcement powers.

At the regional level, the CARICOM Regional Digital Economy Policy Framework should be supplemented by a binding regional AI Protocol that establishes minimum standards for equitable AI deployment across member states. A Protocol modelled on the CARICOM Single Market and Economy's free movement arrangements could create a regional AI market in which only systems meeting minimum equity standards can be deployed. This would give vendors a direct reason to meet Caribbean equity requirements, since market access would depend on it.

International instruments provide additional legal foundations. The Convention on the Elimination of All Forms of Discrimination Against Women (CEDAW), to which all CARICOM member states are parties, creates positive obligations to eliminate discrimination in employment, education, health, and economic life. CEDAW's General Recommendation No. 36 on the rights of girls and women in education is relevant to AI in educational settings. CEDAW committees have begun to address technology-mediated discrimination, and Caribbean states should report on AI-related gender discrimination risks in their periodic reviews.

Group of diverse professionals in a policy discussion

Sector-Specific Application: Where the Framework Must Go Deepest

A general equitable AI framework provides the architecture. Sector-specific application provides the impact. Three sectors warrant immediate priority in the Caribbean: financial services, healthcare, and education.

In financial services, AI-driven credit scoring and financial product delivery are already operational across the Caribbean. Women, particularly those in informal employment, are disadvantaged by credit scoring models that rely heavily on formal employment records, tax returns, and documented transaction histories. The Jamaican banking sector and its equivalents across CARICOM should be required by their financial regulators to conduct gender equity audits of any AI system used in credit assessment, with results disclosed to the relevant regulator and with remediation plans required where equity gaps are identified.

In healthcare, AI diagnostic tools, triage systems, and health information platforms are entering Caribbean health systems. The training datasets used by these tools are predominantly drawn from North American and European clinical populations. Their performance for Afro-Caribbean women, who have distinct patterns of disease incidence, treatment response, and healthcare-seeking behaviour, has not been systematically assessed. Caribbean Ministries of Health should require health AI vendors to provide evidence of validation studies in demographically comparable populations before public procurement approval.

In education, UNESCO's monitoring of AI ethics implementation, including gender provisions, provides a useful external benchmark. Caribbean education ministries should conduct an annual review of AI tools in use in schools and tertiary institutions, assessing each tool against UNESCO's Guidance criteria, with particular attention to whether the tool performs equally for students from different socioeconomic and gender backgrounds. Tools that fail this assessment should be suspended pending remediation.

Institutional Requirements

A framework without institutions to implement it is merely text. The equitable AI framework I am proposing requires specific institutional changes in Caribbean governance structures.

First, national gender bureaux, which exist across CARICOM member states with varying capacity and authority, must be equipped to participate in AI governance. This means dedicated staff with both gender analysis and AI literacy skills, which is a genuine capacity gap that requires investment, not merely a reallocation of existing duties.

Second, data protection authorities, including Jamaica's Office of the Information Commissioner, must be empowered with AI-specific jurisdiction and expertise. The current mandate of most Caribbean data protection offices does not clearly encompass algorithmic accountability. Legislative amendments and budget allocations to enable AI regulation are both necessary.

Third, GSMA Mobile Gender Gap Report 2025 data confirms ongoing gender gaps in mobile access that directly affect AI equity in the Caribbean. A regional monitoring body tracking these indicators against AI deployment outcomes would provide the evidence base for adaptive policy. The OECS Commission and CARICOM Secretariat should jointly establish such a monitoring function, drawing on existing regional statistical capacity.

Recommendations

Enact AI-specific legislation in Jamaica within 24 months. The legislation should establish mandatory bias audit requirements for high-risk AI systems (classified in line with EU AI Act Annex III as a starting point), a right to explanation for automated decisions, mandatory gender impact assessments for public sector AI procurement, and regulatory authority for the Office of the Information Commissioner to investigate and sanction AI-related discrimination.
Negotiate a binding CARICOM AI Protocol establishing minimum equity standards. The Protocol should include minimum requirements for gender bias testing, transparency obligations, and human oversight, applicable to AI systems deployed in any member state. Market access conditions should be attached: systems not meeting the Protocol's minimum standards cannot be deployed across the CARICOM single market.
Require gender equity audits as a condition of public sector AI procurement. Any government ministry or public authority procuring an AI system should be required to obtain an independent gender equity audit of that system before contract signature. Audit results should be published. Contracts should include performance standards tied to equity metrics, with penalties for non-compliance.
Fund participatory design processes for AI systems in healthcare and education. The two sectors with the most immediate and widespread AI deployment should be required to conduct participatory design consultations with women's community organisations before implementing AI tools. These consultations should be funded, not voluntary, and their outcomes should be documented and reflected in system requirements.
Apply CEDAW reporting mechanisms to AI-related gender discrimination. Caribbean states should include AI-related gender discrimination risks in their next CEDAW periodic reviews, committing to specific policy actions in response. This creates international accountability for domestic AI governance failures affecting women, supplementing domestic enforcement mechanisms.
Establish a Caribbean AI equity fellowship programme for women. Closing the institutional capacity gap requires investing in people. A funded fellowship programme, placing women with gender expertise in AI regulatory roles in Caribbean governments and regional institutions, would begin to build the human infrastructure that equitable AI governance requires. The programme should be jointly funded by CARICOM governments, with technical support from ITU and UN Women Caribbean.

Build It Before the Systems Embed

Equitable AI in the Caribbean will not emerge from goodwill and general principles. It will emerge from specific legal requirements, adequately funded institutions, accountable governance processes, and the deliberate inclusion of Caribbean women's expertise and experience in AI system design and oversight.

The international moment is unusually favourable. The EU AI Act is setting global norms, the UN High-level Advisory Body on AI is calling for developing country participation in governance, and UNESCO is monitoring gender provisions in AI ethics implementation. The tools, the frameworks, and the international solidarity all exist. What is missing is the political decision to use them.

An equitable AI framework for the Caribbean is not a luxury or a future priority. Build it while AI systems are still being adopted, or accept that reform becomes far harder once those systems are embedded in the region's economic infrastructure. The design choices made now will decide who benefits from AI in the Caribbean for the next two decades, and who is left out.

Frequently Asked Questions

What does "equitable AI" actually mean in a Caribbean context?

Equitable AI means AI systems that deliver accurate, fair, and beneficial outcomes for all users, including women, rural communities, low-income populations, and people with disabilities. In the Caribbean context, it specifically requires that AI systems perform reliably for populations with limited digital histories, that their design reflects Caribbean social and economic conditions rather than importing assumptions from North American or European contexts, and that affected communities have meaningful input into the standards governing AI deployment.

Why is adding a gender clause to existing technology policy insufficient?

Adding a gender clause to a technology policy that was designed around male-default assumptions does not change the underlying architecture of either the policy or the AI systems it governs. Equitable AI requires rethinking the design process from the outset: who collects the data, whose experiences are represented in training sets, who conducts the risk assessments, and who is consulted in the governance process. A clause appended to a policy written without women in mind will not produce systems that serve women well.

What is the UNESCO Guidance for Generative AI and why does it matter for Caribbean governments?

UNESCO's "Guidance for Generative AI in Education and Research", published in September 2023, provides governments with a framework for managing generative AI in educational settings, including attention to equity and inclusion. For Caribbean governments deploying AI tools in schools and universities, the Guidance provides a starting point for ensuring that AI-enhanced education does not widen existing gender and socioeconomic gaps. Its principles on transparency, human oversight, and equity should be incorporated into national education technology policies.

How does ISO/IEC 42001 relate to equitable AI in the Caribbean?

ISO/IEC 42001, published in December 2023, establishes an international standard for AI management systems, covering risk management, transparency, and responsible deployment. Caribbean organisations adopting this standard commit to systematic risk assessment of their AI systems, which can be extended to include equity and gender impact assessments. Governments could require ISO/IEC 42001 certification as a condition of procurement for AI systems used in public services, effectively mandating equitable design through procurement standards.

What role should UN Women Caribbean play in AI policy?

UN Women Caribbean has existing expertise in gender analysis, regional policy dialogue, and women's rights advocacy that makes it a natural partner for Caribbean governments developing AI governance frameworks. Specifically, UN Women Caribbean should be resourced to conduct gender impact assessments of proposed AI deployments in public services, develop gender-responsive AI procurement standards adaptable by member states, and build the capacity of national gender bureaux to participate meaningfully in AI governance processes.

What is the EU AI Act's approach to high-risk AI in public services?

The EU AI Act, whose prohibitions became effective on 2 February 2025, classifies AI systems used in essential public services, including social benefits, education, and law enforcement, as high-risk. High-risk classification triggers mandatory requirements including conformity assessments, human oversight mechanisms, technical documentation, and post-market monitoring. Caribbean governments deploying AI in equivalent public service areas should adopt similar requirements domestically, regardless of whether the EU Act technically applies to their jurisdiction.

About the Author

Dr. Shirley Budall is a Caribbean expert in gender, inclusion, and AI governance with demonstrated experience in the ethical, legal, social and governance dimensions of artificial intelligence and digital technologies. She conducts legal and regulatory framework reviews and develops policy recommendations for legal reform in AI governance, data protection, human rights, and gender equality. Dr. Budall has knowledge of international and regional AI governance standards and has advised Caribbean government institutions and regional organisations on inclusive AI policy. She is a researcher and consultant working across the CARICOM region on digital economy governance, women's rights in the digital age, and equitable technology development. Contact: insights@starapple.ai