Artificial intelligence is reshaping multilateral decision-making by giving voice to previously unheard stakeholders. Based on my recent book, Political Automation: An Introduction to AI in Government and Its Impact on Citizens (Oxford University Press, 2025), I argue that AI agents can help international organizations better incorporate marginalized communities into global governance frameworks.
AI-powered Systems go beyond Traditional Representation in Multilateralism
The United Nations family and other multilateral institutions continually seek innovative approaches to make their processes more inclusive. However, a persistent "perspective deficit" affects collective decisions globally, with certain voices – particularly vulnerable or conflict affected communities, or the very young/elderly – remaining systematically underrepresented in policy forums.
A promising solution has emerged: AI-powered systems that create virtual representations of underrepresented groups, allowing them to "participate" in multilateral processes. Generative AI can produce virtual representations through which traditionally unheard stakeholders gain "voice" – what Scheuermann and Aristidou call "AI Voice" – in critical discussions ranging from climate action to sustainable development.
AI Proxies can Support Rapid Assessments, Negotiations and Virtual Dialogues
This approach stands apart from traditional AI-based sentiment analytics. While conventional methods merely mine and measure existing opinions, AI agents can actively engage in dialogue about hypothetical scenarios, offering insights into how marginalized communities might respond to proposed policies before implementation.
Researchers at UNU-Macau speculate that such “AI simulations can also be useful in decision making. A strong example is the use of AI system proxies for representing future generations, for instance, in climate-related decision making. But any kind of marginalized group could be represented by AI, on any kind of matter of relevance for the SDGs by using AI-generated outputs to ‘speak’ for under-represented groups. This can allow for inclusion opportunities for under-represented groups, such as including an AI proxy of their perspectives in decision-making scenarios.”
The UN system, with its commitment to inclusive governance, stands at a pivotal moment for adopting these technologies. If they can be designed to ethically and authentically simulate the thought patterns of specific demographic groups, then this technology may serve three critical functions for international organizations:
- Facilitating rapid needs assessments in crisis zones where traditional surveys are impractical, dangerous, or time-consuming
- Supporting negotiation training by creating realistic counterparts for diplomats preparing for complex peacekeeping discussions
- Enabling virtual policy dialogues between AI representatives of conflicting groups, testing potential policies before engaging communities
Early experiments in this field demonstrate promising results. In one project I am working on at UNU-CPR called "Ask Amina/Ask Abdalla," we are testing how AI-powered agents may represent refugee and combatant experiences in the Sudan-Chad border region. This experiment provides a concrete example of a system that combines 3D avatars, large language models, retrieval-augmented generation, and a specially designed AI “anthropologist agent” that can curate a representative knowledge base, to help deliver responses that closely mirror the perspectives of affected populations.
Ethical Implementation of AI Agents in Humanitarian Contexts is Key
Significant governance questions remain. How do we ensure that real people, particularly those from marginalized communities, are not misrepresented, or that their digital avatars are not used to justify policies that may ultimately not be in their interest? Approaches that safeguard the authenticity of real people’s voices must be insisted upon.
It's important to recognize that public participation serves multiple objectives beyond merely gathering perspectives to develop better policies. Participation is itself a right, and the process builds trust in governance systems. If affected communities perceive they are being completely replaced by AI proxies rather than meaningfully included, this could ultimately undermine their trust in multilateral institutions and processes. AI agents should therefore augment rather than replace direct human participation, serving as bridges to inclusion where traditional participation faces obstacles.
For multilateral organizations to responsibly implement these technologies, local stakeholder involvement must remain central. AI agent outputs must align with authentic perspectives rather than algorithmic approximations, that are often biased. Indeed, transparency in AI deployment is essential, particularly regarding the origins and composition of training data.
When representing vulnerable communities through AI proxies, clear documentation helps both decision-makers and members of the communities that may be affected by the outcome of those decisions, understand potential limitations and address biases. Affected communities must be informed, and must be able to opt-out, engendering important conversations around both individual data rights and group rights.
As I discuss in my book, we may be witnessing the emergence of an altogether new form of citizen expression. AI-animated digital versions of people may soon start to engage in discussions on their behalf in a large array of public policy matters. It is crucial that real citizens maintain ownership over their digital counterparts and be able to adjust them so that they prioritize the same values that they would.
AI Proxies as Democratic Catalysts can Redefine Representation in Global Governance
The implementation of AI agents as proxies for underrepresented voices marks a significant evolution in how multilateral institutions conceptualize participation. The relationship between human stakeholders and their AI proxies establishes a new paradigm for inclusive governance that extends beyond traditional participation frameworks.
These developments represent more than mere technical innovations – they embody a fundamental rethinking of representation in the digital age. Successful deployment hinges on maintaining human agency and oversight, where the balance between technological capability and authentic human experience remains critically important.
By establishing guidance structures that prioritize community ownership of digital representations while leveraging technological advantages, international institutions can pioneer a more inclusive form of multilateralism that amplifies previously unheard voices without diminishing their agency or humanity.
About the author
Eduardo Albrecht is a Senior Fellow at United Nations University Centre for Policy Research, Adjunct Associate Professor at Columbia SIPA, and author of “Political Automation: An Introduction to AI in Government and Its Impact on Citizens” (Oxford University Press, 2025). His research examines how artificial intelligence technologies are transforming governance and citizen participation globally