ICS/SIKS Symposium on

Artificial Normative Systems

Utrecht, The Netherlands, 18 September 2007





Davide Grossi

Utrecht University

On constitutive rules. A formal analysis.


11.00-11.30 Coffee



Cristiano Castelfranchi

Italian National Research Council

What does it mean to obey/violate a norm as a norm:

The pretension of norms over the agent's mind.






Marek Sergot

Imperial College, London

"The other made me do it!":

Norms, actions, and agency in multi-agent systems




John Horty

University of Maryland

Reasoning with prioratized imperatives: Where are we?






Johan van Benthem

Amsterdam University and Stanford University

Changing Preferences and Dynamic Logic




Davide Grossi

On Constitutive Rules. A formal analysis.

We present a logical systematization of the Searlean notion of constitutive rule as counts-as statements: X counts as Y in context C. By moving from a very simple intuition about what counts-as statements mean, i.e., conceptual subsumptions, we disentangle three semantically different readings of statements of the type X counts as Y in context C, from the weaker notion of contextual classification to the stronger notion of constitutive rule. These many ways in which counts-as can be said are then formally addressed in modal logic. The resulting framework allows for a formal characterization of all the involved notions and their reciprocal logical relationships. We refer throughout the whole talk to work in legal and social theories which is used as theoretical ground for our formal analysis.



Cristiano Castelfranchi

What does it mean to obey/violate a norm as a norm: the pretension of norms over the agent's mind.

Norms are not only aimed at inducing a given behavior (intention) in a cognitive agent, but they are aimed at achieving this via 'goal-adhesion': agent Y must realize that there is a will (prescription) by another agent X (which should be 'entitled' and acknowledge as such) about his doing (intending to do) something, and must 'adopt' such a meta-goal of X; Y has not just to adopt (for whatever reason) the goal of 'doing an action', but has to adopt the goal of X about Y doing the action because X's wants/asks so. How this adopted goal, this recognized norm (and 'authority') generate an intention? Why should Y be sensible about X's goals about his mind? Moreover, the pretension of the norm (and of the authority) is that Y adhere to the required behavior for specific reasons and motives; Y shouldn't be motivated by his personal interest (for example avoiding sanctions) or by pity, friendship, etc., or by sharing the final end (function) of the norm. He must 'obey', that is adopt the norm just because he recognize X's authority and has the final value/goal of respecting the authority, the law. There is a paradox in law and in education: using sanctions in order to push Y (by learning) to respect the norms not for sanctions. I will discuss all this (the mental representations and process of norms as norms), and perhaps also why in Artificial Normative Systems we definitely need violations and disputes, and systems able to deal with this fundamental human organizational phenomenon.



Marek Sergot

"The others made me do it!'': Norms, action, and agency in multi-agent systems.

References to normative concepts (obligation, permission, commitment, social commitment, ...) feature prominently in the literature on multi-agent systems. One idea is that agent interactions are, or can usefully be seen as being, governed by norms. The term `social laws' has also been used, particularly in connection with `artificial social systems'. A `social law' is a set of obligations and prohibitions on agents' actions, that, if respected, allow the agents to co-exist in a shared environment. The question of what happens to system behaviour when `social laws' are not respected, however, has received little or no serious attention.

We will present a formal framework and a (modal-logical) language that allows several different categories of non-compliant behaviour to be characterised and classified, distinguishing between various forms of unavoidable or inadvertent non-compliance, behaviour where an agent does `the best that it can' to comply with its individual norms but nevertheless fails to do so because of the actions of other agents, and behaviour where an agent could have complied with its individual norms but did not. The aim, amongst other things, is to be able to investigate what kind of global system properties emerge if we assume, for instance, that all agents of a certain class will do the best that they can to comply with their individual norms, or never act in such a way that they make non-compliance unavoidable for others. The resulting logic bears a striking resemblance (not entirely accidental) to Ingmar Porn's (1974) logic of `brings it about' agency, but also, unexpectedly, to very familiar forms of logic that have been extensively studied in other contexts. A natural generalisation to groups of agents leads to something that might be termed `a logic of unwitting collective agency' --- `unwitting' because no assumptions are made at all about the reasoning or perceptual capabilities of the agents. The last part is work in progress.

John Horty

Reasoning with prioritized imperatives: where are we?

I begin by motivating the idea of deontic logic as founded on a system of prioritized imperatives. I then consider several proposals for drawing conclusions about what an agent subject to these various imperatives actually ought to do. For ease of comparison, these proposals will be reformulated in a common notation and from a common
point of view. Unfortunately, none of these proposals is adequate; problems will be illustrated. I then present a new proposal, which resolves the old problems, but leads to new difficulties of its own. Various criteria of evaluation will be discussed. The overall goal of the talk is to map out the current state of our understanding in this area.



Johan van Benthem

Changing Preferences and Dynamic Logic.

We show how dynamic logics can model a wide spectrum of events that lead to preference changes for agents, while providing rich languages describing relevant scenarios. We give some examples from belief revision, maintaining deontic obligations, and the deliberative fine-structure of social choice.



We are very grateful to our sponsors:

Siks LogoLogo NWOLogo ICS