Revision of Defeasible Logic Preferences

There are several contexts of non-monotonic reasoning where a priority between rules is established whose purpose is preventing conflicts. One formalism that has been widely employed for non-monotonic reasoning is the sceptical one known as Defeasible Logic. In Defeasible Logic the tool used for conflict resolution is a preference relation between rules, that establishes the priority among them. In this paper we investigate how to modify such a preference relation in a defeasible logic theory in order to change the conclusions of the theory itself. We argue that the approach we adopt is applicable to legal reasoning where users, in general, cannot change facts or rules, but can propose their preferences about the relative strength of the rules. We provide a comprehensive study of the possible combinatorial cases and we identify and analyse the cases where the revision process is successful. After this analysis, we identify three revision/update operators and study them against the AGM postulates for belief revision operators, to discover that only a part of these postulates are satisfied by the three operators.

💡 Research Summary

The paper investigates how to alter the preference relation among rules in Defeasible Logic (DL) in order to change the conclusions drawn by a defeasible theory. DL is a non‑monotonic reasoning formalism that distinguishes strict rules, defeasible rules, and defeasible facts; when conflicts arise, a preference ordering over rules resolves which rule prevails. While most prior work on belief change in DL focuses on adding or removing facts or rules, this study proposes a novel “preference revision” approach that keeps the underlying facts and rules fixed and only modifies the relative strength of the rules. This is especially relevant for legal reasoning, where statutes and case facts are immutable but judges or lawyers can argue about which rule should dominate.

The authors first formalise a DL theory as a directed graph: nodes are rules, edges encode the preference relation. They then enumerate all possible elementary modifications of this graph, defining three primitive operators:

1. Preference Flip – reverses an existing edge, turning a higher‑priority rule into a lower‑priority one (and vice‑versa).
2. Preference Insertion – adds a new edge between two rules that previously had no ordering.
3. Preference Removal – deletes an existing edge, making the two rules incomparable.

For each operator they introduce a “conclusion difference function” that measures how the set of defeasible conclusions changes after the modification. A revision is deemed successful when a target conclusion, previously unattainable, becomes derivable while the theory remains consistent and in normal form. The authors also impose a minimal‑change principle: the revision should affect as few preferences as possible.

The paper conducts a systematic combinatorial analysis of all possible cases, identifying conditions under which each operator yields a successful revision. For example, a flip can succeed when the flipped rule participates in a conflict whose resolution directly determines the target conclusion; an insertion can succeed when creating a new hierarchy resolves a deadlock; a removal can succeed when eliminating a dominating rule allows a weaker rule to fire.

After the technical analysis, the three operators are evaluated against the classic AGM postulates for belief revision (Closure, Consistency, Inclusion, Vacuity, Success, Minimal Change, etc.). The authors find that Closure, Consistency, and Inclusion are satisfied, but the Minimal Change postulate—and consequently the AGM “Recovery” and “Conservatism” postulates—are violated in many scenarios. This divergence stems from the structural nature of preferences: changing a single edge can have cascading effects on the defeat graph that are not captured by AGM’s set‑theoretic framework. The paper argues that a preference‑centric revision theory requires an extended set of postulates that explicitly account for graph‑based dependencies.

A detailed case study from legal reasoning illustrates the approach. Two statutes conflict; the original DL theory prefers Statute A, leading to conclusion X. By applying a Preference Flip to make Statute B higher, the system derives conclusion Y, which aligns with a plaintiff’s argument. The authors show how the same outcome could be achieved by inserting a new preference that reflects a higher‑level constitutional principle, demonstrating the flexibility of the three operators.

The discussion highlights limitations and future work. Current operators are defined at a theoretical level; efficient algorithms for large rule bases, user‑friendly interfaces for specifying preferences, and automated suggestion of minimal revisions are needed. Moreover, extending the framework to other non‑monotonic systems (e.g., logic programming, deontic reasoning) and developing a full set of “preference AGM” postulates are identified as promising research directions.

In summary, the paper makes three major contributions: (1) it introduces a systematic method for revising defeasible theories by altering rule preferences rather than facts or rules; (2) it provides a thorough combinatorial analysis and defines three primitive revision operators; and (3) it evaluates these operators against AGM postulates, revealing that only a subset of the classic postulates hold, thereby motivating a new line of inquiry into preference‑based belief revision. This work broadens the applicability of Defeasible Logic to domains such as legal reasoning where stakeholders can only influence the hierarchy of norms, not the norms themselves.