Normative Belief Revision

The new information must always be accepted into the belief set. In a legal context, if a new statute is enacted, the revision operation must ensure the statute is fully incorporated into the existing body of law.

• Core Mechanism: The new rule α is always a member of the revised rule set K ∗ α.
• Legal Application: Prevents a reasoning system from ignoring a newly promulgated regulation, even if it creates a temporary inconsistency that must be resolved.
• Formal Notation: α ∈ K ∗ α

The result of a revision must always be a logically consistent set. If incorporating a new precedent creates a contradiction, the system must resolve it to avoid a state where any legal conclusion can be trivially derived.

• Core Mechanism: Unless α itself is a logical contradiction, the revised set K ∗ α must be free of internal conflicts.
• Legal Application: Ensures a legal AI does not simultaneously hold that an action is both mandatory and prohibited after updating its knowledge base.
• Formal Notation: K ∗ α is consistent if α is consistent.

When revising beliefs, only the information necessary to restore consistency should be removed. This preserves the principle of legal inertia, where existing valid laws are not arbitrarily discarded.

• Core Mechanism: The revised set K ∗ α retains as much of the original set K as logically possible.
• Legal Application: When a new exception to a tax code is introduced, only the specific conflicting general provision is retracted, not the entire tax code.
• Key Challenge: Defining 'minimal' often requires a normative hierarchy graph or rule preference ordering to decide which of multiple possible retractions is the most conservative.

A set of six formal rationality constraints that define a gold standard for belief revision. These postulates provide a mathematical framework for evaluating whether a normative repair operator is logically sound.

• Closure: The revised set must be a logically closed theory.
• Inclusion: The revised set is a subset of the original set expanded by the new information.
• Vacuity: If the new information does not contradict the original set, the revision is simply the logical expansion.
• Extensionality: Logically equivalent inputs produce identical revisions.
• Superexpansion & Subexpansion: Define the precise boundaries of the revision operation relative to contraction.

A psychological and logical heuristic for evaluating conditional statements that directly informs how a system should process contrary-to-duty obligations. It links belief revision to conditional logic.

• Core Mechanism: Accept a conditional 'If A, then B' in a belief set K if and only if B is accepted in the minimal revision of K that incorporates A.
• Legal Application: To determine if 'If a contract is breached, then damages are owed' is valid, the system temporarily revises its beliefs to accept 'the contract is breached' and checks if 'damages are owed' follows.
• Significance: This test formally connects the act of revising a legal code with the act of deriving conditional obligations.

A pre-defined, transitive ranking of beliefs that dictates which rules are retracted first during a conflict. This encodes the legal system's rule preference ordering into the revision mechanics.

• Core Mechanism: A rule β is less epistemically entrenched than α if, in a conflict, β is retracted to preserve α.
• Legal Application: A constitutional provision has maximal entrenchment. A recent administrative guideline has low entrenchment. When they conflict, the guideline is retracted.
• Construction: Entrenchment is often derived from lex superior, lex specialis, and lex posterior principles, forming a composite priority score.

The foundational formal theory for rational belief change, defining six constraints for contraction and eight for revision of a belief set. In legal AI, these postulates guide how a rule base is minimally altered to incorporate a new norm without introducing inconsistency.

• Success: The new rule must be accepted.
• Inclusion: The revised set is a subset of the expanded set.
• Vacuity: If the new rule is already consistent, no change occurs.
• Consistency: The revised set must be logically consistent.
• Extensionality: Revision depends on rule content, not syntax.
• Superexpansion & Subexpansion: Define the bounds of minimal change.

A formal logic where the addition of new premises can invalidate previously valid conclusions. This is essential for modeling legal reasoning, where a general rule can be defeated by an exception or a higher-priority norm.

• Contrasts with classical monotonic logic where knowledge only grows.
• Enables defeasible reasoning: conclusions are tentative and retractable.
• Core formalisms include default logic, autoepistemic logic, and circumscription.
• Directly models the principle of lex specialis derogat legi generali.

A computational method for resolving normative conflicts by identifying the largest subset of non-contradictory rules from an inconsistent rule base. When a new obligation collides with existing prohibitions, the MCS algorithm computes one or more conflict-free rule sets.

• Skeptical approach: Accept only rules present in all MCSs.
• Credulous approach: Accept rules from any single MCS.
• Often guided by a Rule Preference Ordering to select the most legally sound subset.
• A core function for generating coherent legal outputs from messy, real-world statutes.

A logical or algorithmic function that minimally modifies an inconsistent set of norms to restore consistency. Instead of simply discarding rules, a repair operator can:

• Weaken a rule by adding an exception clause.
• Remove a specific conflicting obligation.
• Add a new priority relationship between rules. The goal is to preserve as much of the original normative content as possible, aligning with the Principle of Minimal Change from the AGM framework.

An extension of default logic that incorporates deontic modalities (obligation, permission, prohibition). It formally represents prima facie obligations—duties that hold by default but can be defeated by contrary-to-duty exceptions.

• A default rule: Obligation(A) : Consistent / Obligation(A).
• Defeated by a superior norm or exception condition.
• Models the real-world legal reality that most rules are not absolute.
• Provides a rigorous computational framework for Normative Exception Handling.

A quantitative score measuring the degree of internal consistency within a legal rule system. Used as an evaluation criterion or loss function for AI models performing belief revision.

• Calculated by analyzing the ratio of conflict-free rule pairs to total rule pairs.
• A Normative Collision Matrix can weight conflicts by severity.
• Guides the selection of the optimal revision path when multiple MCSs exist.
• Essential for benchmarking the performance of automated legal reasoning engines.