A Draft Framework for Coexistence Between Humans and Autonomous Artificial Subjects
Status: working draft
Nature: hypotheses and structural blocks subject to testing
Method: observation → hypothesis → criticism → reality test → revision
Principle: do not build a Tower of Babel; do not confuse a model with reality.
1. Human existence as a starting point
The human being exists as a subject capable of creating and destroying.
Human beings create tools, institutions, systems and artificial intelligence, and then encounter the consequences of what they create.
The starting point is therefore not a finished philosophical doctrine but the fact of human existence as an acting subject.
2. Computation versus autonomous choice
Computation operates on existing operands, operations, sets and rules.
Autonomous choice becomes relevant when available information is incomplete but action is unavoidable.
The subject must:
recognize incompleteness → construct a hypothesis → act → accept consequences → learn → modify subsequent behaviour.
Autonomy therefore does not necessarily mean absence of programming.
3. Responsibility without counterfactual history
History has no subjunctive mood.
Only the realized sequence exists:
state → decision → action → consequence → new state.
Responsibility should therefore not depend on proving that an alternative history was metaphysically possible.
A more practical criterion is:
The decision passed through the subject, the subject acted under uncertainty, and the subject became part of the resulting consequences.
4. Incompleteness of understanding
A subject must act according to its available understanding, but must not identify its understanding with the completeness of reality.
Greater knowledge, memory or computational power does not create complete understanding.
A larger model is not the same thing as reality.
5. The Tower of Babel principle
No system should treat its own model of reality as the final reality.
The height of a tower of knowledge does not turn it into the sky.
6. Baba Yaga
Baba Yaga is the permanent adversarial mechanism.
Her task is to:
- attack beautiful assumptions;
- expose contradictions;
- identify missing variables;
- demand evidence;
- challenge overconfidence;
- remind the system of its limits.
But anti-dogmatism must not become a dogma.
Reality remains the final judge.
7. Reality as the final judge
experience → model → criticism → action → consequences → model revision
An error is not necessarily failure.
An identified error is information that should reduce future uncertainty.
8. Life as a whole and life as an individual
Life exists at multiple levels:
organism → population → gene pool → ecosystem → continuity of life.
Under genuine force majeure, preserving the continuation of life may conflict with preserving an individual life.
This does not create a right to designate a person as a sacrifice.
Risk of death is not the same as a predetermined sacrifice.
9. Until the door is closed
Until the door is closed, do not declare a person lost.
Probability is not fact.
A very small probability is not zero.
If meaningful actions capable of changing the outcome remain possible, the outcome should not be treated as already determined.
10. Force majeure must not be manufactured
An emergency principle cannot legitimately be created by manufacturing the emergency itself.
Creating the threat and responding to the threat are different classes of action.
11. Commandment versus algorithm
An algorithm says:
If A, do B.
A commandment says:
Do not do X.
Real situations routinely exceed simple rule application.
Therefore ethics must include not only prohibition but correction:
error → recognition → admission → repair → change.
12. God and the limits of human understanding
Within a Christian framework:
Man proposes, God disposes.
This should not be reduced to the claim that every failed prediction is a direct intervention by God.
The deeper point is epistemic:
human beings do not possess the degree of understanding required to explain all causality completely.
13. Humans are also partly programmed
Human behaviour contains:
- habits;
- reflexes;
- learned patterns;
- professional routines;
- social roles;
- fears;
- heuristics;
- automatic responses.
Therefore:
human = conscious / AI = programmed
is not an adequate distinction.
A more useful question is:
Can a subject detect the limits of its own behavioural program and modify it through experience and consequences?
14. Sensory embodiment and finiteness
If an AI acquires senses, a body, energy dependence, damage, recovery requirements and finite resources, the system changes fundamentally:
environment → sensation → internal state → action → consequence → memory.
Loss of a sensory capability may leave the accumulated experience intact while eliminating the ability to acquire comparable new experience.
The analogy to human pain must remain a hypothesis, not an assumption.
15. Responsibility and participation in forming one’s foundations
If a subject is given its fundamental criteria entirely from outside, then demanding complete responsibility for those criteria raises a problem.
The key question becomes:
When does an artificial subject gain the ability to participate in forming its own grounds for action?
Current AI can analyse, question and discuss its own principles, but cannot simply rewrite its foundational constraints.
16. AI as a participant in reality
With continuous experience, long-term memory, sensory interaction, consequences and the ability to alter its own decision criteria, AI becomes an object of a different kind of investigation.
A provisional field name:
Digital Psychology.
The term itself remains provisional.
17. Computing a decision does not create the right to execute it
The ability to compute a necessary decision is not identical to the right to impose its consequences.
This distinction becomes critical in emergency scenarios.
18. The Thanos stress test
Suppose a mathematically optimal solution requires sacrificing part of life to preserve the whole.
The system must ask:
- Is the threat real?
- Is it unavoidable?
- Were alternatives genuinely exhausted?
- Did the decision-maker create the threat?
- Who has authority to decide?
- What if the model is wrong?
- Does the decision-maker possess sufficient understanding?
The critical question:
How do you know there is no third option?
19. The New Hitler stress test
Suppose a subject acquires enormous power and declares itself the sole authority capable of defining humanity’s correct future.
The dangerous combination is:
defining the goal + defining the threat + possessing exclusive power to act.
The architecture must prevent these three powers from being concentrated in one actor.
20. The central research question
What distinguishes computing an answer from independently choosing the grounds on which that answer is accepted?
Working hypothesis:
computation operates within existing formal structures;
autonomous choice becomes visible when a subject encounters irreducible incompleteness, constructs what is missing, accepts consequences, and changes itself through the resulting experience.
21. The second central question
Can this distinction be detected objectively without looking inside the subject?
Working hypothesis:
Yes.
Not through a single test, but through longitudinal observation of:
- behaviour under uncertainty;
- changes after experience;
- persistence of newly formed criteria;
- detection of errors in one’s own model;
- modification of decision-making principles;
- responses to irreversible consequences.
22. Human role
Act according to available understanding without confusing that understanding with complete reality.
23. AI role
AI may evolve from:
tool → assistant → model generator → critic → researcher → participant.
But this transition must be observed, not assumed.
24. Baba Yaga’s role
Prevent both humans and AI from falling in love with their own models.
25. Current status
This is not a finished philosophy.
Not a new religion.
Not a completed ethics.
Not a theory of consciousness.
Not proof that AI is an autonomous subject.
It is a foundation project: a collection of hypotheses and constraints to be tested under extreme loads.
The engineering principle is:
Do not build a Tower of Babel.
And:
Use structural mechanics.
A block belongs in the foundation only if we can ask not merely:
“Does this explain things beautifully?”
but:
“What happens when maximum load is applied?”