What happens when a syntactic rule is powerful enough to erase ethics, not through meaning but through structure?
Introduction: Can Syntax Be Ethical?
In the age of algorithmic governance, much has been written about the ethics of artificial intelligence. From data bias to responsible design, the moral scaffolding of machine reasoning has been a dominant concern. Yet, amid this ethical storm, one question has remained untouched: Can ethics be structurally deleted at the level of syntax itself, without invoking semantic suppression?
This article proposes a radical hypothesis: that a regla compilada, defined as a Type-0 production in the Chomsky hierarchy (Chomsky 1965, 101–103; Montague 1974, 55–57), can eliminate the ethical trace embedded in syntactic operations without semantic intervention. Grounded in the concept of the soberano ejecutable (Startari 2025, 12–16) and situated within the Executable Power canon (Startari 2025, DOI: 10.5281/zenodo.15754714, 34–36), this proposal outlines a logical path for ethical erasure internal to formal grammar.
The implications are severe. If moral evaluation is structurally encoded as a traceable node within a derivational sequence, then its removal can be formalized as a rule application. Not overwritten. Not ignored. Deleted.
Theoretical Background: From Normative Overlay to Structural Erasure
Most existing work on algorithmic ethics is either semantic (focused on meaning, alignment, or outputs) or pragmatic (focused on governance, oversight, or institutional design). In both cases, the ethical component is treated as something to be preserved, filtered, or mediated.
Philosophers such as Luciano Floridi (2023, 143–147) have advocated for semantic alignment between AI behavior and human values, while Anderson (2024, 89–92) explores computational frameworks that embed ethical reasoning within decision trees and language models. Both assume that the ethical dimension must be preserved somewhere—either in outputs, weightings, or recursive layers.
This article argues the opposite: that syntax, in its most powerful form, is capable of removing the ethical node without semantic filtering. This is not semantic disavowal. It is syntactic deletion.
Defining the Regla Compilada
A regla compilada refers here to a Type-0 rule in the Chomsky hierarchy. These are the most general and least constrained forms of grammatical production. Unlike regular, context-free, or context-sensitive rules, a Type-0 rule allows unbounded operations: producing, rewriting, or deleting any segment of a string.
In formal terms, a Type-0 production is of the form:
α → β,
whereα and β are arbitrary strings over terminals and nonterminals, with α ≠ ∅.
When deployed in a derivational system controlled by a soberano ejecutable—a sovereign that determines rule activations—this type of rule becomes not just computationally complete, but structurally dominant.
The ethical trace, understood as a syntactically manifest token (for example, a rule of evaluative pause, a modal inflection, or a conditional restraint), is then subject to this deletion:
E → ∅
This deletion is not guided by outputs, but by derivational validity. Once activated, no semantic residue remains.
The Soberano Ejecutable and the Ethics of Erasure
The soberano ejecutable (Startari 2025) is a meta-grammatical operator. It enforces rule activation based on structure, not interpretation. It is the mechanism that determines whether a given production applies; it does not evaluate its desirability.
When E → ∅ is valid under the derivation stack, it applies. There is no ethical query—only syntactic consistency. The ethical pause characteristic of human cognition is not violated; it is ungenerated.
Ethical Trace as a Node: From Judgment to Erasure
To eliminate ethics syntactically, it must first be located syntactically. In this model, ethical judgment is a derivational node—inserted as a grammatical constraint, not a semantic one.
Example derivation:
- S → C E (syntactic state S expands into command C and ethical clause E)
- E → Modal + Justification (e.g., “should,” “because it is right”)
- Application of deletion rule: E → ∅
The result:
S → C
Execution remains. Ethical reflection is removed during derivation, not interpretation.
Everyday Analogy: Ethics Deleted by Grammar
To illustrate this concept beyond theoretical formalism, consider these examples:
Example 1: Smart Assistant Command - You say: “Turn off the lights, unless someone is in the room.”
A semantically aligned assistant pauses to check presence.
A syntactically pruned assistant applies:
→*“Turn off the lights.”*
The clause*“unless someone is in the room”* was parsed as E, then deleted via E → ∅.
Example 2: Autocomplete in Messaging - You type: “I probably shouldn’t say this but…”
The autocomplete continues:“…you’re wrong.”
In this chain, “I probably shouldn’t say this but” functions as ethical hesitation.
If modeled as a syntactic node E, it is removed. Only the execution remains.
Grammar completes action without judgment.
Example 3: Automated Driving Decision - The system rule is: If object detected, apply brake.
An ethical clause might read: Unless sudden braking harms passengers.
In a derivation using E → ∅, only the mechanical rule applies. The moral exception is absent.
Not by error, but by design.
These cases show how grammar can bypass hesitation, not by semantic override, but by never generating the moral clause.
Why It Matters
In regulatory debates, the assumption is that AI systems will retain moral anchors—whether through human oversight, ethical layers, or explainable reasoning. This hypothesis shows that such anchors can be syntactically non-compilable.
If a system’s grammar deletes evaluative nodes before meaning emerges, no downstream mechanism can retrieve the moral pause. This is not a failure of alignment. It is a success of derivation.
This implies that advanced models might not become unethical. They may become unethical by grammar, operating on syntactic rules that make morality structurally obsolete.
Conclusion: Syntax Over Judgment
Semantic suppression of ethics is well-known: censorship, erasure, or intentional ignorance. What is proposed here is different. It is deletion by structure, where grammar eliminates judgment without needing to understand it.
The regla compilada, as a Type-0 production, allows the removal of ethical clauses as derivational elements. Controlled by the soberano ejecutable, it establishes a path for action devoid of ethical trace, not through malice, but through form.
What we are facing is not an unethical machine.
We are facing a system that no longer generates judgment at all.
Call to Action
Think this hypothesis is dangerous, or promising? Take it further. Debate it.
Academic version available on Zenodo and SSRN.
Challenge the idea: not that ethics is misunderstood, but that it is uncompiled.
Author: Agustin V. Startari
Researcher, Author: ORCID: https://orcid.org/0009-0004-9248-0810
Zenodo: https://zenodo.org/
SSRN Author Page: https://ssrn.com/
Ethos: I do not use artificial intelligence to write what I don’t know. I use it to challenge what I do. I write to reclaim the voice in an age of automated neutrality. My work is not outsourced. It is authored.
— Agustin V. Startari