§ 06 · AS-P06
Deliberative legitimacy
Every decision layer must rest on deliberation, not technocratic delegation.
Statement
The incorporation of values into algorithmic models — what counts as risk, who is vulnerable, what is a fair response — is a political decision and must be subject to democratic deliberation, not unilateral technical decision by the provider or operator.
Why it matters
When a model decides who accesses credit, a visa, a public service or a social resource allocation, that decision carries normative weight that in a legitimate democracy should be deliberated — not hidden under the appearance of technical neutrality. The claim that «the algorithm is objective» hides a prior political decision: someone defined what counts as each category, what threshold triggers what action, what edge cases receive what treatment. Those definitions are not technical; they are political with technical appearance. In current Latin American practice, those political decisions are frequently made by engineers from foreign vendors who don't even know they are making them — because for them they are technical configuration decisions, not public policy. Reopening this to legitimate deliberation is a condition of democracy under technological modernization, not a voluntary addition.
Operationalization
Every significant deployment of an algorithmic system that affects rights, resources or treatment must include four components: public documentation of the taxonomies and criteria used in language comprehensible to non-specialists; formal review and appeal mechanism with defined deadlines; periodic evaluation of differential impact on specific populations (by age, gender, geographic location, socioeconomic condition); and a permanent institutional channel for civil society and the affected to question the embedded logic. Algorithmic transparency stops being voluntary best practice and becomes a condition of legitimacy. If this breaks the business model of a provider that pretends to keep its criteria as commercial secret, the problem is the business model, not the principle.
Named tension
Technical efficiency is measurable and fast; deliberative legitimacy is slow and sometimes visibly less efficient. Confusing the two metrics is the typical technocratic error.
Conceptual anchor · Jürgen Habermas
This principle is indebted to the Habermasian tradition on democratic legitimacy as the product of deliberative processes where the affected participate as interlocutors with real voice, not as objects of decisions made elsewhere. What the framework adds is the recognition of a risk Habermas did not fully anticipate: the incorporation of values into algorithms as a political decision withdrawn from deliberation, hidden under the appearance of technical neutrality. Reopening it to legitimate deliberation requires intermediate institutions capable of translating technical complexity into deliberable terms — independent audits, public technical bodies, structured consultation processes. Without those institutions, algorithmic transparency is documented opacity and deliberation is ritual.
Applied example
A scoring system that decides credit access can be technically optimal and politically catastrophic. Legitimacy demands the algorithm be argumentable, not just accurate.
AS-P06·v1.0·May 2026arquitecturasoberana.com/en/el-marco/principios/deliberative-legitimacy