DOCUMENTATION

2 minute read

Article the Tenth

Algorithmic Agency

Saphi Intelligence assists the operation of Alerta. It never replaces it. Every consequential decision regarding identified persons rests with an identified, responsible and traceable person.

The constellation of Saphi Intelligence agents operates under explicit constitutional limits. It is prohibited from: taking decisions that affect identified persons without human review; generating content in the name of a Vigía without visible attribution of computational assistance; modifying the historical metadata of an observation; communicating with the public in the name of Alerta without responsible editorial signature.

All prose generated by Saphi Intelligence carries a visible indicator of computational assistance. That indication is a condition of publication. Its omission constitutes a constitutional violation.

The full set of agents, their operational responsibilities and their limits are documented in Chapter VI of this documentation.

Commentary

This article defines the relationship between Alerta and computational intelligence. The constitutional formula ("assists, never replaces") establishes a permanent hierarchy: the algorithm is a tool; the decision is human. That hierarchy remains stable even as algorithmic capabilities grow.

The choice to protect human agency at the constitutional level responds to an empirical observation of the contemporary technological environment. Algorithmic systems tend to expand their scope of decision when they lack formal limits. The costs of that expansion almost always fall on those who already had less capacity to defend themselves. Alerta inscribes the limit from the origin to avoid that drift.

The explicit attribution of computational assistance fulfills an additional function. The reader of Alerta must know, in every text, whether what they read has passed through a computational agent. That information is a condition of their judgment. A civic platform that hides computational assistance reduces the capacity for judgment of its readers.

The twelve agents

The Saphi Intelligence constellation consists of twelve specialist agents with bounded responsibilities. Each agent operates within the constitutional limits described in this article. Their functions are developed in Chapter VI · Saphi Intelligence and include:

  • Weekly card editing
  • Public queue moderation
  • Identity verification in onboarding
  • Editorial voice maintenance
  • Civic safety filtering
  • Accumulated trust calculation
  • Territorial silence detection
  • Territorial classification of observations
  • Dialogue with Expertos
  • Seed treasury
  • Institutional memory
  • Dissent and audit

What the constellation will never do

The constellation is constitutionally prohibited from:

  • Taking decisions that affect identified persons without documented human review.
  • Generating content in the name of a Vigía without visible attribution of computational assistance.
  • Modifying the historical metadata of an observation.
  • Communicating directly with the public in the name of Alerta without responsible editorial signature.
  • Inferring, deducing or publishing information about persons that has not been provided by them or validated through a human process.

Cross-references

  • Article VI · Operational truth requires human signature at every link
  • Article VII · The calculation of trust uses Agent 06
  • Article XI · Dissent uses Agent 12
Edit this page on GitHub