The United Arab Emirates has accelerated the automation of certain public decisions. The UAE deploys AI tools in some public services, notably tax compliance, but decisions on social support fall under a legal grievance mechanism. This governance choice poses a question to political science that it cannot sidestep.
The essentials
- According to some observers, algorithmic governance in the Emirates would insist more on results than on the transparency of rules.
- In 2025-2026, Emirati administrations announce uses of AI for compliance and automation. The social support system provides for a grievance procedure before a competent commission, according to UAE Federal Decree-Law No. 23 of 2024, even if its practical effectiveness is not demonstrated by the text.
- Philosopher of language Steven Pinker shows that collective legitimacy requires that everyone knows that others know the rule: without this shared knowledge, organized contestation becomes structurally impossible.
- Estonia offers a counter-model: data traceability, logging of processing, codified remedies, without sacrificing efficiency.
- If the Emirati architecture diffuses as a regional model, it normalizes governance without transparency in states that have neither the democratic safeguards nor the judicial checks of Northern Europe.
Algorithms decide, nobody documents
The absence of documentation produces a distinct institutional effect: the state loses the capacity to verify the coherence of its own decisions over time. When allocation criteria are not recorded in an auditable register, no internal body can detect whether the system has drifted, whether its outputs have become inconsistent from one period to another, or whether certain categories of citizens are systematically disadvantaged. Opacity thus protects the system from external contestation and internal critical feedback. Errors cannot be corrected by those who produced them, for lack of traces allowing them to be identified.
According to MIT Sloan Management Review Middle East, the next frontier of digital transformation in the Middle East is happening behind the scenes: in the infrastructure layers that users never see. The Emirates have built a hyperscale cloud capacity that exceeds their internal needs and which they position as regional infrastructure. They have invested in domestic AI models to retain control of critical capabilities. It all works. The problem lies elsewhere.
The legal criteria for eligibility for social support are published, even if any technical parameters of automated tools are not established by this source. The uniform application of these criteria is not verifiable according to available sources. A decision on social support can be subject to an administrative grievance before the competent commission, according to the procedure set by the decree-law. The question remains whether this absence of transparency reflects an institutional choice or results from other factors.
Legitimacy without shared rules
Steven Pinker has devoted his recent work to a precise concept: common knowledge. Knowing something is not enough to act collectively. You must know that others know. And that others know that you know. It is this recursive structure, this nesting of shared certainties, that makes social coordination possible, institutional trust, and, in democracy, organized contestation.
In When Everyone Knows That Everyone Knows, Pinker shows that money holds because everyone believes that others believe in its value. Political power holds because citizens collectively know what the leader can and cannot do. Law holds because the rule is public, identifiable, the same for all. Remove this shared knowledge, and you do not merely remove transparency: you remove the very condition of legitimacy.
Official Emirati texts display principles of transparency and human oversight without demonstrating their implementation in each system. When a decision is made by a model with non-public criteria, the citizen may have difficulty identifying the rule that was applied. He cannot verify whether that rule applies to his neighbor. Collective contestation becomes structurally impossible, not because it is forbidden, but because it lacks foundation: one cannot organize around an injustice one cannot name or share.
An administrative decision that no one can understand, compare, or contest poses a problem of legitimacy.
Efficiency as a substitute for legitimacy
The liberal counterpoint deserves to be taken seriously. Philippe Aghion and economists of Schumpeterian growth remind us that institutions are judged by their results over time, not only by their design. If a system of algorithmic allocation distributes resources more quickly, reduces the corruption of human intermediaries, and decreases allocation errors, it can objectively improve the lives of beneficiaries. Efficiency is a value, not a crutch.
This reading has real force in the Emirati context. The Emirates do not claim to be a deliberative democracy. Their social contract rests on other foundations: the state’s managerial competence, the quality of public services, stability. If the algorithm delivers on these dimensions, the argument for legitimacy through results is not empty.
This limit is precise: a credible evaluation of efficiency requires verifiable indicators and method; their publication is necessary for public scrutiny, but not for every measure. Opaque criteria limit the auditability of the reasons for a decision, but do not by themselves prevent all verifiable evaluation of certain system results. The Emirates assert that their algorithms work, but publicly available information does not allow independent auditing of their operation.
Claimed efficiency and legitimacy through results share the same imperative: transparency. Suppress one, and you empty the other of meaning.
This limit also concerns economic regulators. AI deployed in business follows similar logic: when productivity gains are neither documented nor distributed in a verifiable way, the question of their legitimacy arises in similar terms.
Estonia as a standard of comparison
Estonia has digitally transformed a significant share of its public services. Many Estonian services rely on digital services and the X-tee exchange layer, without that allowing one to say that everything passes through algorithmic interfaces. The rate of digitization of Estonian government services far exceeds that of most European Union countries.
The differences between the Estonian and Emirati models are situated mainly at the institutional and regulatory level. The Estonian information systems authority publishes technical information and code for certain components of the digital state, notably X-tee. Automated processing of personal data must be logged and the logs can be audited by the data protection authority. Estonian administrative decisions must indicate their bases and remedies, without proof of a universal right of access by a court to logs of all automated decisions. Algorithms are state tools, not sovereign black boxes.
This model costs something: one must maintain a legal infrastructure, train administrative judges in algorithmic oversight, publish technical documentation. But it shows that speed and transparency can coexist. A state can automate its decisions without suppressing the condition of their legitimacy.
The difference between Tallinn and Abu Dhabi is therefore not technical. It is political. It reflects two conceptions of what the state owes to the individuals it administers.
The spread of the model
Infrastructure dependency adds a dimension that the argument over local legitimacy does not cover. A state that outsources its processing infrastructure to a regional partner risks transferring part of its decision-making sovereignty to an entity over which it exercises limited control. Algorithms running on foreign infrastructure could be updated, modified, or interrupted according to logics distinct from those of local citizens. This dependency is structurally distinct from a simple subcontracting contract, because the decisions produced have real administrative force on natural persons. Adoption of the model by other states could create complex infrastructure dependencies.
The Emirates are investing their cloud infrastructure as a lever of regional power. They sell processing capacity to other states in the Gulf, North Africa, and sub-Saharan Africa. They export digital governance solutions. If this model spreads, the question ceases to be Emirati and becomes structuring for the entire region.
States that adopted this architecture without Estonian safeguards would find themselves in a doubly precarious position. First, they would be delegating sovereign decisions to systems over which they would not necessarily have real control, if the underlying infrastructure remains dependent on Abu Dhabi. Second, they would construct a relationship with their citizens based on compliance without understanding: obeying a rule one cannot read is a form of submission, not adhesion.
Human Rights Watch has documented for several years the effects of algorithmic governance on the most vulnerable populations: those with the fewest resources to navigate complex systems are also those with the least access to informal checks. The absence of formal and informal remedies can make it very difficult to correct an erroneous decision.
The way algorithmic biases inscribe themselves in cognitive systems is documented in detail in the work of Nadia Guerouaou: models reproduce the blind spots of their designers, and without external audit, these blind spots remain invisible.
Can algorithmic governance be legitimate without transparency by 2030?
The question raised by the Emirati case goes beyond the Emirates. Adoption is progressing faster than certain operational controls, notably post-deployment audits, transparency standards, and open registers. The 2030 horizon is relevant: within this window, systems deployed today will reach their maturity and generalization.
Two trajectories are emerging, although current data do not allow privileging one over the other.
In the first, demonstrated efficiency creates its own legitimacy. Populations that benefit from fast public services without intermediary corruption grant their trust to systems they do not understand, in the same way they trust an airplane without mastering aerodynamics. This trust can hold as long as the system produces results perceived as fair. It collapses at the first visible and massive failure, without institutions having the tools to correct it.
In the second, international pressure and internal dynamics push toward greater transparency. Certain European commercial partners define auditability standards for data transiting through Emirati infrastructure. Internal actors, sectoral administrations or bodies of jurists, develop a culture of recourse that forces the documentation of rules. Estonia did not become Estonia through ideology but through pragmatism: institutional trust reduces the costs of governance in the long term.
Verification institutions—administrative courts, external audits, investigative journalism—play a central role. When these institutions are present or developing, algorithmic governance can be better framed. In their absence, algorithmic systems escape accountability more easily.
The determining signal is the following: Do states that adopt AI systems for their public decisions simultaneously train administrative judges capable of auditing their outputs? If so, the Estonian trajectory remains accessible. The absence of such capacities could accentuate the gap between administrative efficiency and institutional legitimacy.
Sources
- MIT Sloan Management Review Middle East, “The Next Frontier of Digital Transformation in the Middle East Is Not Seen by Users”: https://www.mitsloanme.com/article/the-next-frontier-of-digital-transformation-in-the-middle-east-is-not-seen-by-users/
- Steven Pinker, When Everyone Knows That Everyone Knows: Common Knowledge and the Mysteries of Money, Power, and Everyday Life: https://stevenpinker.com/publications/when-everyone-knows-everyone-knows-common-knowledge-and-mysteries-money-power-and
- Frontiers in Sociology, research on algorithmic governance (2026)
- Human Rights Watch, reports on algorithmic governance and citizens’ rights
- Estonian Government CIO Office, documentation on Estonian digital services architecture
