Source-linked AI summary
Aligning Institutional Power and Operational Ability
Andrea Marino, Giovanni Sileno, Thomas van Binsbergen, Tom van Engers
TL;DR
Normative designs for agentic systems can assign responsibilities or powers that agents cannot operationally fulfil, or enable actions beyond intended authority. This paper models operational processes alongside institutional power and uses alignment properties to analyse incongruent configurations. A stylized consignment scenario demonstrates the framework’s use in detecting such configurations during design.
Problem
Normative positions that do not match agents’ operational abilities can leave rights frustrated or make agentic systems harder to control.
Source: §1 · pp. 1–2Method
The paper formalizes operational processes and institutional power together, then defines properties for assessing power-ability alignment.
Source: §3.3 · p. 5, §4 · p. 7Results
The framework captures incongruent institutional configurations in a stylized consignment scenario and supports their ex-ante detection.
Source: §1 · p. 2, §6 · p. 10Takeaways & Limitations
The framework serves as an analytical tool to guide computational normative-system design.
Source: §1 · p. 2The consignment example is simplified and does not model real consignment law.
Source: §1 · p. 2
Abstract
from arXiv · showhide
To be socially sustainable, Agentic AI systems require normative designs that align responsibilities with the ability to fulfil them. Identifying incongruences such as unsatisfiable expectations or privilege escalation is crucial in avoiding normative pitfalls. In this study, we discuss design dimensions suitable for describing multi-agent systems in light of this norm-ability alignment requirement. We propose a formal framework centred on process-oriented modelling and institutional power for the analysis of a given normatively relevant process, discuss strategies for repairing it, and elaborate on how the symmetrical relation between operational process and institutional state affects the lifecycle of normative positions.
1. Introduction
The introduction argues that normative responsibilities and powers must match agents’ operational abilities, or rights may be frustrated and agentic systems may become difficult to control. It proposes a process-oriented framework and alignment properties for identifying such design problems.
1. Introduction
- Misaligned expectations and operational abilities can frustrate rights even without an explicit norm violation.The introduction illustrates this with a formally granted freedom of movement that barriers make impossible to exercise.Source: §1 · p. 1
- Agentic systems may exploit operational sequences to exceed intended access, or fail when assigned tasks without the required tools.The paper notes that either situation can hinder controllability or orchestration.Source: §1 · pp. 1–2
- The authors argue that inadequate process semantics make institutional side effects hard to trace, motivating explicit treatment of institutional power and operational ability.They connect these gaps to lingering powers and unsatisfiable obligations that may prompt ex-post disputes.Source: §1 · p. 2
- The consignment scenario is a stylized illustration, not a model of real consignment law.It assumes the goods enter the broker’s estate upon insolvency and the broker’s mandate to sell lapses then.Source: §1 · p. 2
- The paper proposes a process-execution model, three design properties, and an analysis of incongruent institutional states.The properties are operational groundedness, formal institutional legitimacy, and power-ability alignment.Source: §1 · p. 2
2. Background
The background connects normative multi-agent systems with process modelling. It identifies institutional power as underrepresented in computational normative frameworks and process models as a basis for analysing execution and compliance.
2. Background
- The paper’s contribution bridges normative multi-agent systems and process-modelling research traditions that have partly developed separately.It reviews related work in both traditions and presents connections between them.Source: §2 · p. 3
- Computational normative frameworks often centre on duties and permissions, while institutional power remains comparatively underserved.Some frameworks explicitly represent power, but the relation between deontic positions and operationalized powers can remain implicit.Source: §2 · p. 3
- Process-modelling research uses Petri nets for control-flow analysis and extends them to workflow coordination, conformance, and compliance monitoring.Process mining can evaluate event logs against normatively qualified patterns.Source: §2 · p. 3
3. Theoretical Framework
The framework represents operational execution and institutional state together, so normative positions can be assessed against what agents can do and what remains possible in the process. It defines the operational, institutional, and governed-net components used for this analysis.
3.1. Operational layer
- Petri nets model operational events through places, transitions, weighted arcs, and token-based firing sequences that determine reachable markings.A read arc tests for required tokens without changing their marking when its transition fires.Source: §3.1 · p. 4
- An agent can act when it is mapped to a transition and that transition is enabled; realisability asks whether it can become enabled from the current marking.The example makes the broker’s ability to sell depend on instruction, an active mandate, and the continued presence of the goods.Source: §3.1 · p. 4
3.2. Institutional layer
- The institutional layer tracks active duties and powers, with reactive rules updating positions when operational actions occur.The policy update activates or deactivates positions, and legitimacy requires an active power covering the fired transition; most example powers are one-off, while pow(B,sell) persists until termination or bankruptcy.Source: §3.2 · p. 5, Table 1 · p. 6
3.3. Governed nets
- A governed net combines the operational net, agent mapping, policy, and initial marking, pairing each reachable operational marking with an institutional state.Its reachability graph represents fired transitions and governed runs; in the example, action order determines which configuration a run reaches.Source: §3.3 · pp. 5–6
4. Power-Ability Alignment
The framework classifies normative configurations by whether active powers and duties remain operationally exercisable and whether available actions have institutional backing. It uses reachable configurations to expose sequence-dependent misalignments and shows repairs that preserve intended outcomes.
4. Power-Ability Alignment
- Alignment requires formal institutional legitimacy, which backs each immediate ability with an active power, and operational groundedness, which keeps active positions exercisable.An ungrounded power or duty lacks an assigned, currently realisable transition through which its holder can exercise or discharge it.Source: §4 · p. 7
- After bankruptcy revokes pow(B,sell), the mandate and goods can still let B sell, while seizure can leave duty(B,sell) ungrounded after the goods disappear.The first case is an illegitimate ability; the second is an ungrounded duty.Source: §4 · p. 7
- Because instruct, bankrupt, and terminate can occur in any order, misalignment depends on event sequences; bounded nets with finitely many positions have finite reachable state spaces.Keeping operational and institutional semantics distinct exposes defects before any agent misbehaves.Source: §4 · p. 7
4.1. Repairing a Misaligned Normative Design
- Reducing repairs cross layers: legitimacy failures require operational restrictions, while groundedness failures require institutional deactivation; expansions reverse that layer assignment.Ability is policy-invariant, so institutional policy changes cannot withhold an otherwise enabled action.Source: §4.1 · p. 8
- The amended policy and net make every reachable configuration aligned by tying position lifecycles to operational transitions and removing the residual sell ability.The repair deactivates pow(B,sell) when sell consumes the goods and makes bankrupt consume the mandate token that instruct reads; Figure 2 shows only the initial instruct branch and selected state components.Source: Figure 2 · p. 9, §4.1 · p. 9
- Repairs must preserve designer-specified terminal outcomes; the framework permits reductions, while outcome-preserving repair synthesis remains future work.Admissible reductions remove unsupported abilities or retire ungrounded positions; expanding behavior or authority requires independent justification.Source: §4.1 · pp. 8–9
5. Discussion and Future Work
The discussion distinguishes this framework from prior process and institutional approaches by connecting institutional powers to independently modelled operational abilities over consumable resources. It identifies limits in what groundedness guarantees and leaves efficient verification and outcome-preserving repair synthesis for future work.
5. Discussion and Future Work
- Unlike prior process-centred or institutional frameworks discussed, this approach relates institutional powers to independently modelled operational abilities and characterises their alignment over consumable resources.The cited approaches include BPMN deontic policy specification, trace-based compliance checking, InstAL, and Symboleo, which the discussion contrasts on these dimensions.Source: §5 · pp. 9–10
- Operational groundedness expresses possible future exercise, not guaranteed fulfilment or an agent’s ability to ensure it.Stochastic Petri-net analysis of likelihood or timing, links to agentive reliability, efficient verification, and outcome-preserving repair synthesis remain future work.Source: §5 · p. 10
6. Conclusion
The paper presents a process-oriented formal framework for aligning operational abilities with institutional positions and uses a stylised consignment scenario to show ex-ante detection of misaligned configurations.
6. Conclusion
- The framework distinguishes operationally impossible positions from enabled actions unsupported by institutional power, making both misalignments explicit over reachable configurations.The authors propose future work on automated detection and eventual repair of misaligned normative designs.Source: §6 · p. 10