Reframing the AI era not around who can use technology, but around who can shape it, and introducing the Participation Gap.
By Sana Nasir, Founder & Chair, Alliance for Equitable AI · Policy Paper Series No. 001 · Published 6 June 2026
Abstract
For more than two decades, digital inclusion has been understood mainly through access: internet connectivity, mobile ownership, broadband coverage, affordability, and digital literacy. These remain essential, particularly for emerging democracies where access gaps still shape social and economic opportunity. However, artificial intelligence introduces a new layer of exclusion. A citizen may be connected to the internet, own a smartphone, and use digital services, yet remain excluded from the systems that design, train, govern, and profit from AI.
This paper argues that emerging democracies must move from digital inclusion to AI inclusion. The central challenge of the AI era is not only who can use technology, but who can shape it. The paper introduces the concept of the Participation Gap: the divide between societies, institutions, and communities that actively contribute to AI development and governance, and those that mainly consume AI systems developed elsewhere.
Using Pakistan as a policy reference point, the paper proposes an A4EAI Participation Framework built around five dimensions: access, representation, capability, governance, and ownership. It argues that Pakistan’s National AI Policy direction, Digital Pakistan agenda, and growing technology ecosystem create an opportunity to build a more inclusive AI future, provided participation is treated as a policy objective rather than a by-product of adoption.
01. Introduction
The history of digital development has largely been written as a story of access. For governments, development institutions, and technology companies, the central questions have been familiar: who is connected, who can afford devices, who has broadband coverage, who can use digital services, and who remains offline? These questions matter. In countries such as Pakistan, digital access has shaped education, employment, public service delivery, entrepreneurship, and civic participation.
Over the past decade, Pakistan has made serious efforts to expand the foundations of a digital economy. The Digital Pakistan Policy placed emphasis on ICT-enabled growth, innovation, e-government, entrepreneurship, and improved public service delivery. Universal service initiatives have extended connectivity to underserved areas. Digital skills programmes, freelancing initiatives, and public sector digitisation efforts have helped widen the country’s digital base.
Yet artificial intelligence changes the meaning of inclusion.
In the AI era, being connected is not the same as being represented. Using a digital service is not the same as shaping the system behind it.
Having access to a chatbot is not the same as having one’s language, culture, laws, realities, and risks reflected in the model’s design and deployment. This distinction is important for emerging democracies.
If artificial intelligence becomes a layer through which citizens access information, education, finance, healthcare, government services, and employment, then exclusion will no longer be limited to those who are offline. It will also affect those who are online but invisible within AI systems.
A farmer may have mobile internet but receive poor AI-supported agricultural advice because the system does not understand local cropping patterns. A woman may access a digital service but be affected by gendered assumptions embedded in automated decision-making. A student may use generative AI in Urdu, Punjabi, Sindhi, Pashto, Balochi, or Roman Urdu, yet receive weaker educational support than an English-speaking peer. A public institution may deploy AI tools without sufficient capacity to evaluate bias, privacy risks, accountability, or local relevance.
These are not distant concerns. They are the next generation of digital inequality.
This paper argues that the policy conversation must move beyond digital inclusion towards AI inclusion. Digital inclusion asks whether people can access and use technology. AI inclusion asks whether people, communities, and institutions can meaningfully participate in the AI systems that increasingly shape their lives. The distinction is not semantic. It has serious implications for governance, equity, and development.
02. From the Digital Divide to the Participation Gap
The traditional digital divide was relatively easy to describe, even if difficult to solve. It referred to the gap between those with access to digital technologies and those without. Over time, this understanding became more sophisticated. Policymakers began to distinguish between access, affordability, skills, safety, and meaningful use.
This evolution was necessary. A person may technically be connected but still unable to benefit from digital services due to cost, lack of literacy, poor quality of service, unsafe online environments, or limited content in their language.
Artificial intelligence adds another layer. AI systems do not merely provide access to information. They increasingly organise, interpret, recommend, classify, translate, predict, and decide. They are embedded in search engines, educational tools, hiring systems, financial products, health platforms, customer service systems, and government workflows. This means that exclusion can occur even when access exists.
The divide is between those who participate in creating, governing, and benefiting from AI systems and those who are merely subject to them. This paper calls this divide the Participation Gap, and it has five dimensions.
- Data gap: communities with limited digitised knowledge or low representation in training data are more likely to be misunderstood or ignored.
- Language gap: AI performs best in well-represented languages. English dominates; regional languages and informal expression remain underdeveloped.
- Capability gap: countries lacking AI skills, research capacity, compute access, and regulatory expertise are less able to build or evaluate AI systems.
- Governance gap: many emerging democracies adopt AI faster than they build institutions to govern it, creating risks to accountability and trust.
- Ownership gap: the economic value of AI may flow disproportionately to those who own the models, infrastructure, data pipelines, and IP.
Taken together, these gaps create a new development challenge. A country may succeed in expanding digital access but still remain dependent on AI systems designed elsewhere, governed elsewhere, and optimised for someone else’s priorities.
03. Pakistan’s AI Moment
Pakistan is entering this debate at an important time. The country has a young population, a large technology workforce, a growing freelancing economy, a developing startup ecosystem, and increasing public sector interest in digital transformation. The National AI Policy direction reflects recognition that artificial intelligence can support economic growth, productivity, innovation, and national competitiveness. This is a positive development.
Pakistan should not approach AI only through fear. Artificial intelligence can support education, agriculture, healthcare, climate resilience, financial inclusion, disaster response, legal information access, and citizen services. It can help small businesses improve productivity, support teachers with learning material, assist public institutions in service delivery, and create new forms of employment.
However, AI adoption without inclusion can reproduce older inequalities in newer forms.
Pakistan’s digital divide has always had social dimensions. Rural communities, women, low-income groups, persons with disabilities, and speakers of regional languages often face greater barriers to meaningful digital participation. If AI systems are layered on top of these existing inequalities, they may deepen exclusion rather than reduce it.
For example, an AI-enabled public service system may improve efficiency for digitally confident citizens but exclude those who cannot navigate formal language, English interfaces, or complex digital workflows. AI-based hiring tools may increase speed but reproduce bias if trained on historical employment patterns. Educational AI tools may benefit elite students first if they are available mainly in English or require paid subscriptions. Financial AI systems may widen access to credit but also create opaque forms of risk scoring that citizens cannot challenge.
Pakistan must encourage AI innovation, investment, and adoption, while ensuring that AI systems are fair, locally relevant, accountable, and inclusive. If policy focuses only on risk, it may discourage innovation. If it focuses only on innovation, it may neglect rights, equity, and public trust. The challenge is to build an AI ecosystem that is both enabling and responsible.
04. Why Access Is No Longer Enough
Pakistan cannot build an inclusive AI future without affordable connectivity, reliable infrastructure, digital skills, and broad public access to technology. However, AI inclusion requires asking deeper questions:
- Who creates the datasets?
- Who validates the models?
- Who decides what counts as accuracy?
- Who audits harms?
- Who benefits economically?
- Who is consulted before deployment?
- Who can challenge an automated decision?
- Who is responsible when a system fails?
These questions are particularly important in public sector contexts. When AI is used by private companies, consumers may have some choice, though often limited. When AI is used by government, citizens may have no realistic alternative. This creates a higher duty of care.
Public sector AI should therefore meet stronger tests of accountability, transparency, fairness, and accessibility. It should be evaluated not only for technical performance but also for social impact. A system that is efficient but exclusionary should not be considered successful.
Digital inclusion often focuses on bringing people to technology. AI inclusion also requires bringing people into governance.
That means public consultation, civil society participation, community testing, local language evaluation, independent audits, grievance mechanisms, and clear institutional responsibility.
05. The A4EAI Participation Framework
The framework is designed for emerging democracies, public institutions, development partners, universities, civil society organisations, and private sector actors seeking to evaluate whether AI systems are inclusive in practice. It is built around five dimensions: access, representation, capability, governance, and ownership.
5.1 Access
Access remains the first requirement. Without connectivity, devices, affordability, and digital literacy, citizens cannot benefit from AI tools. But AI access must be understood more broadly than internet access, it includes affordable access to AI services, local language interfaces, disability-inclusive design, safe usage environments, and tools that meet local needs.
For Pakistan, this means ensuring AI tools are not available only to English-speaking, urban, elite users. AI-enabled services must work for people using regional languages, voice interfaces, low-cost devices, and low-bandwidth environments. Access also includes institutional access: universities, startups, public sector bodies, and civil society need computing resources, open datasets, and technical support.
5.2 Representation
Representation asks whether people and communities are visible within AI systems, in datasets, languages, design teams, governance bodies, expert consultations, and testing processes. An AI system may be technically advanced but socially narrow if it is trained, tested, and governed without diverse participation.
For Pakistan, representation means investing in Urdu and regional language datasets, local legal and policy knowledge, culturally relevant educational content, and community-informed evaluation, including women, rural communities, persons with disabilities, minorities, and non-elite users. Representation is not symbolic; it directly affects system performance. If a model does not understand local language, context, or social realities, it will produce weaker outputs, and in high-stakes contexts weak outputs can cause real harm.
5.3 Capability
Capability refers to the skills, institutions, and infrastructure needed to participate meaningfully in AI. This includes technical skills such as machine learning, data science, cybersecurity, and model evaluation, and policy skills such as risk assessment, procurement oversight, ethical review, legal analysis, and public communication.
A country cannot govern AI effectively if only engineers understand it. Nor can it innovate effectively if policy discussions are disconnected from technical realities. AI capability must therefore be interdisciplinary. Public officials need training on AI risks; journalists need capacity to report responsibly; civil society needs tools to evaluate rights impacts; universities need support for applied research; regulators need expertise to audit systems; and citizens need awareness of when AI is being used. Capability also includes institutional memory, durable institutions, standards, and accountability mechanisms, not individual champions alone.
5.4 Governance
Governance is the dimension that determines whether AI systems are trusted. Responsible AI governance should include transparency, accountability, privacy protection, human oversight, fairness, security, explainability where appropriate, and grievance redress, principles widely reflected in international frameworks, including UNESCO and OECD approaches.
For emerging democracies, governance must also be practical. Many institutions do not have large AI ethics teams or advanced audit capacity, so governance tools must be usable by ministries, regulators, local governments, universities, and small firms, not designed only for large technology companies.
- Risk classification for AI systems
- Local language performance testing
- Privacy and data protection safeguards
- Grievance mechanisms for affected citizens
- Public sector AI procurement standards
- Bias and discrimination assessments
- Human review for high-impact decisions
- Public disclosure where AI is used publicly
Governance should not be treated as a brake on innovation. It should be treated as the infrastructure of trust.
5.5 Ownership
Ownership is often missing from AI inclusion debates. A society may use AI extensively while owning very little of the underlying value chain, models, cloud systems, data infrastructure, intellectual property, and platforms may remain controlled by foreign firms, creating dependency and limiting national capacity.
Ownership does not mean every country must build frontier models from scratch. But countries can build strategic forms of ownership: local datasets, domain-specific models, public-interest AI tools, open-source collaborations, national research capacity, sovereign data infrastructure, and locally governed applications for priority sectors such as education, agriculture, health, climate resilience, and public services.
The goal is not technological isolation. The goal is negotiated participation.
Pakistan should collaborate globally while developing enough local capacity to shape AI according to its own development needs.
06. Policy Recommendations
6.1 Treat AI inclusion as a national policy objective
AI policy implementation should explicitly include AI inclusion as a measurable objective; it should not be assumed that innovation automatically produces inclusion. Indicators should go beyond the number of startups, trained professionals, or investment to include language coverage, gender participation, public sector readiness, regional inclusion, accessibility, and adoption by underserved groups.
6.2 Build public-interest Urdu and regional language AI resources
Language is foundational to AI inclusion. Pakistan should support high-quality, ethically sourced datasets for Urdu and regional languages, educational material, public service information, legal and regulatory content, health guidance, agricultural knowledge, and culturally relevant material, developed with attention to consent, privacy, copyright, and community representation.
6.3 Create an AI public sector readiness programme
Before deploying AI widely in government, Pakistan should invest in readiness, training officials in AI basics, procurement risks, data governance, privacy, ethics, citizen communication, and evaluation, and providing templates for risk assessment and procurement. Public institutions should not procure AI systems they cannot understand, evaluate, or govern.
6.4 Establish independent AI impact assessment mechanisms
High-impact AI systems should undergo assessment before deployment, especially in policing, welfare, health, education, credit, recruitment, and citizen services, examining discrimination, privacy risks, exclusion, accuracy, explainability, human oversight, and grievance redress. Where public rights are affected, citizens should know when AI is being used and how to challenge decisions.
6.5 Support interdisciplinary AI research
AI governance cannot be left only to computer scientists, nor innovation only to policymakers. Pakistan needs interdisciplinary research spanning technology, law, ethics, sociology, economics, public administration, media studies, gender studies, and development studies. Universities should build applied AI governance labs and policy clinics that work with government, civil society, and industry.
6.6 Create an AI Inclusion Index
Pakistan should develop an AI Inclusion Index to measure who benefits from AI adoption and who remains excluded, moving the policy conversation from aspiration to measurement.
- AI tools available in Urdu & regional languages
- Regional distribution of AI skills programmes
- Public sector AI readiness
- Local dataset availability
- Gender participation in AI training & employment
- Accessibility for persons with disabilities
- Civil society participation in AI governance
- Grievance mechanisms for AI-affected citizens
6.7 Include civil society and communities in AI governance
AI governance should not be a closed conversation between government, industry, and technical experts. Civil society organisations, journalists, educators, women’s groups, disability rights organisations, consumer protection bodies, and community representatives should be included in consultations, particularly in emerging democracies where institutional trust may already be fragile. Public engagement is not cosmetic; it improves the quality, legitimacy, and acceptance of AI systems.
07. Role of Development Partners
For many years, development support in technology focused on connectivity, digital skills, e-government, and entrepreneurship. These remain important. However, AI requires new forms of support. Development partners can help emerging democracies build:
- AI governance capacity
- Local language resources
- Independent research capacity
- Cross-country learning platforms
- Ethical data infrastructure
- Public sector readiness
- Inclusive innovation ecosystems
They should also avoid promoting AI adoption without safeguards. The wrong kind of AI assistance can create dependency, vendor lock-in, surveillance risks, or poorly governed automation. The right kind can strengthen public institutions and expand opportunity.
08. Pakistan as a Test Case for AI Inclusion
Pakistan is well positioned to become a test case for AI inclusion. It has the scale, demographic diversity, linguistic complexity, public sector need, and technology talent to make AI inclusion both necessary and possible. The country also has a history of digital inclusion work through connectivity expansion, mobile adoption, digital skills, freelancing, and public sector digitisation. These experiences provide valuable lessons, and the next step is to apply them to AI.
Pakistan should not wait until AI systems are deeply embedded before asking whether they are fair or inclusive. The window for shaping governance is now. If Pakistan can build a model that combines innovation, inclusion, and accountability, it can contribute not only to its own development but also to global debates on AI in emerging democracies.
09. Conclusion
The future of artificial intelligence will not be determined only by the countries that build the largest models or spend the most money. It will also be shaped by the societies that ask the right questions early. For emerging democracies, the central question is not simply how to adopt AI. It is how to participate in AI.
Digital inclusion brought millions online. AI inclusion must ensure that they are not invisible once they arrive.
Access remains essential, but it is no longer enough. Pakistan’s AI journey should be judged not only by how many tools are deployed, how many startups are funded, or how many professionals are trained. It should also be judged by whether AI systems understand local realities, serve diverse communities, protect rights, and create shared value.
Artificial intelligence can widen inequality, but it can also expand opportunity. The difference will depend on governance, representation, capability, ownership, and public trust.
Emerging democracies do not need to remain passive consumers of AI systems built elsewhere. With the right frameworks, they can become active contributors to a more equitable AI future.
References
- Government of Pakistan, Ministry of Information Technology and Telecommunication. (2018). Digital Pakistan policy. MoITT.
- Government of Pakistan, Ministry of Information Technology and Telecommunication. (2025). National artificial intelligence policy 2025. MoITT.
- UNESCO. (2022). Recommendation on the ethics of artificial intelligence (Adopted 23 November 2021). United Nations Educational, Scientific and Cultural Organization.
- OECD. (2024). Recommendation of the Council on artificial intelligence (OECD/LEGAL/0449; originally adopted 2019, updated May 2024). Organisation for Economic Co-operation and Development.
- World Bank. (2023). Digital progress and trends report 2023. World Bank.
- World Bank. (2025). Digital progress and trends report 2025: Strengthening AI foundations. World Bank.
- GSMA. (2025). The mobile gender gap report 2025. GSMA Connected Women.
- Digital Rights Foundation. (2023). Feedback on Pakistan’s draft national artificial intelligence policy. Digital Rights Foundation.
- Pakistan Institute of Development Economics. (2025). Will AI transform Pakistan? Assessing the 2025 national policy (PIDE Knowledge Brief No. KB-136). PIDE.
- Institute of Strategic Studies Islamabad. (2025). Pakistan’s National AI Policy 2025: Developing AI ecosystem and ethical data governance [Issue Brief]. ISSI.
Alliance for Equitable AI · Policy Paper Series No. 001 · A4EAI.com