Skip to content

Human authorship and stewardship

Why authorship is the better invariant

“Humans must remain better than AI” is not a stable educational principle. “Humans must remain meaningfully responsible for the worlds they authorize and inhabit” is much more durable.

Human authorship means that people remain participants in forming the ends, commitments, standards, permissions and responsibilities that shape their lives and institutions. This is compatible with delegating enormous amounts of analysis and execution. It is incompatible with treating machine capability as automatic legitimacy.

Stewardship extends authorship through time and across shared systems. A person is not only the author of private preferences; they inherit relationships, institutions, infrastructure, culture and nature that they did not create and can damage or improve for others.

Four kinds of human relation to a task

Every important Obelisk capability should specify which of these applies:

  1. Internalize - the learner should carry enough knowledge or skill without external assistance for orientation, fluency or safety.
  2. Reconstruct / evaluate - the learner need not memorize every detail but must be able to rebuild the mechanism or critically inspect claims.
  3. Retrieve / collaborate - the learner should know how to find trustworthy current knowledge or the right human/AI expertise.
  4. Delegate with accountability - the learner may hand off execution but must establish purpose, permissions, evidence, monitoring and responsibility appropriate to stakes.

These categories are more useful than a blanket “AI allowed / AI forbidden” rule.

Domains of enduring accountability

AI can contribute to all of these. The point is that capability alone should not dissolve human or legitimate institutional ownership:

  • purposes and commitments;
  • rights, consent and coercion;
  • epistemic standards for consequential decisions;
  • education and human formation;
  • care and relationships;
  • taste and cultural transmission;
  • institutional authority;
  • high-stakes irreversible action;
  • embodied self-care and practical life;
  • intergenerational stewardship;
  • responsibility for delegated systems;
  • maintenance of the common physical, social and epistemic world.

A stronger acceptance question

The Obelisk should eventually produce people about whom an independent examiner, collaborator, community member or institution could reasonably say:

I trust this person to understand enough to know what matters; to learn what they do not know; to distinguish evidence, values and incentives; to use expertise and AI without becoming passive before them; to make and revise judgments with taste and humility; to care for people and shared systems; to act rather than merely comment; and to remain accountable for the consequences of what they authorize.

That is not a claim of perfection. It is a demanding conception of adulthood and citizenship under high technological capability.