Commitment to comply: what kind of an intention? Giovanni Sartor - - PowerPoint PPT Presentation

Giovanni Sartor

Commitment to comply: what kind of an intention?

Giovanni Sartor

CIRSFID - Faculty of law, University of Bologna EUI - European University Institute of Florence

July 14, 2012

• G. Sartor (CIRSFID-EUI)

Compliance 1 / 26

Giovanni Sartor

Purpose of the presentation

introduce the notion of a normative systems introduce the idea of compliance with it discuss motivations to comply and intentions to comply

• G. Sartor (CIRSFID-EUI)

Compliance 2 / 26

Giovanni Sartor

Actions

Actions and omissions (E logic) EjS means “j brings it about that S”. ¬EjS means “j omits to bring about that S”. For instance EJohnDamaged(Tom) ¬EJohnDamaged(Tom)

• G. Sartor (CIRSFID-EUI)

Compliance 3 / 26

Giovanni Sartor

Obligations and prohibitions

OEjS means “it is obligatory that j brings it about that S”. O¬EjS means “it is obligatory that j does not bring about that S", or “it is forbidden that j brings about that S”. For instance OEJohnCompensated(Tom) O¬EJohnDamaged(Tom)

• G. Sartor (CIRSFID-EUI)

Compliance 4 / 26

Giovanni Sartor

Norms

I represent norms as conditionals, where

n

= ⇒ allows for defeasible modus ponens (no need to go into logical details). For instance: ExInjured(y) n = ⇒ ExDamaged(y) ExDamaged(y) n = ⇒ OExCompensated(y)}

• G. Sartor (CIRSFID-EUI)

Compliance 5 / 26

Giovanni Sartor

A normative systems, factual circumstances and their entailments

Circumstances C1 = {EJohnInjured(Tom)} Normative system N1 = {[ExInjured(y) n = ⇒ ExDamaged(y)]; [O¬ExDamaged(y)]; [ExDamaged(y) n = ⇒ OExCompensated(y)]} Normative entailments (C1 ∪ N1) | ∼ EJohnDamaged(Tom) (C1 ∪ N1) | ∼ OEJohnCompensated(Tom) Norms constitute normative outcomes, a normative system links circumstances to normative outcomes

• G. Sartor (CIRSFID-EUI)

Compliance 6 / 26

Giovanni Sartor

Relativised obligation

Definition

A normative system N entails a consequence φ in circumstances C if N ∪ C entails φ but C does not: N ∪ C | ∼n φ def = N ∪ C | ∼ φ and C | ∼ φ

• G. Sartor (CIRSFID-EUI)

Compliance 7 / 26

Giovanni Sartor

Relativised obligation

A relativised obligation sentence does not express a norm, but it expresses an assertion about the implications of norms (normative systems) and circumstances.

Definition (Relativised sentences and obligations)

We say that any sentence φ holds relatively to normative system N and circumstances C, and write [φ]N,C iff N ∪ C | ∼n B [B]N,C

def

= N ∪ C | ∼n B In particular when the sentence B is an obligation sentence OAx, we say that it is obligatory relatively to N in C that x does A, and write ON,CAx (rather than [OAx]N,C), to express that N ∪ C | ∼n OAx: ON,CAx

def

= N ∪ C | ∼n OAx

• G. Sartor (CIRSFID-EUI)

Compliance 8 / 26

Giovanni Sartor

An example

Example

C1 = {EJohnInjured(Tom)} N1 = {ExInjured(y) n = ⇒ ExDamaged(y) O¬ExDamaged(y) ExDamaged(y) n = ⇒ OExCompensated(y)} The following inferences holds on the basis of the example: (C1 ∪ N1) | ∼ EJohnDamaged(Tom) (C1 ∪ N1) | ∼ OEJohnCompensated(Tom) Therefore, we can say that: ON1,C1¬EJohnDamaged(Tom); ON1,C1EJohnCompensated(Tom) We can express the idea that different obligations hold relatively to different normative systems.

• G. Sartor (CIRSFID-EUI)

Compliance 9 / 26

Giovanni Sartor

An abbreviation

Let us write ON1Aj to mean that it is is obligatory, according to N1 that j does A in the current circumstances Φ ON1Aj

def

= ON1,ΦAj ∧ Φ where Φ either is a complete description of the current state of affairs, or at least includes all relevant circumstances, i.e., those circumstance supporting arguments for OAj, plus circumstances supporting defeaters of such arguments. For instance we may say ON1EJohnCompensated(Tom) iff ON1,EJohnDamaged(Tom)EJohnCompensated(Tom) ∧ EJohnDamaged(Tom)

• G. Sartor (CIRSFID-EUI)

Compliance 10 / 26

Giovanni Sartor

An alternative characterisation

It is obligatory in possible world wi that j does A wi ON1Aj

def

= N1 ∪ Φ | ∼n Aj and wi Φ It is currently obligatory that j does A w0 ON1Aj

def

= N1 ∪ Φ | ∼n Aj and w0 Φ where wo is the present world. NB: if | ∼n is nonmonotonic wi ON1Aj

def

= N1 ∪ Φ | ∼n Aj and wi Φ and there is no Φ1 such that Φ ⊂ Φ1 and N1 ∪ Φ1 | ∼n Aj

• G. Sartor (CIRSFID-EUI)

Compliance 11 / 26

Giovanni Sartor

Compliance

Mary is appointed to a professorship. She signs a contract stating her commitment to comply with the University regulations. John enters a PhD program. He is directed to the booklet containing the regulations he has to comply with. Linda is appointed as a judge. She takes an oath to respect the Constitution and the laws of her country. Adolf Eichman enters the SS. He takes an oath of obedience to death to Adolph Hitler and the superiors he has designated. Antony enters the Franciscan order. He promises to respect “The Rule

• f St. Francis” as well as the law of the Catholic Church.

Mary, a shop-owner, receives a threats by gangsters belonging to a mafia organisation. She chooses to comply with all rules imposed by that organisation (monthly protection money, code of silence, etc.) to avoid problems with the bad guys. A digital agent enters and electronic marketplace. It commits to respect all rules of the marketplace.

• G. Sartor (CIRSFID-EUI)

Compliance 12 / 26

Giovanni Sartor

Policy-based intention to comply

Definition (Policy-based intention to comply)

An agent j’s commitment to comply with normative system N can be understood as the agent’s j conditioned intention to do any action Aj that is obligatory according to N in a possible factual situation Φ, whenever Φ is actualised (is the case): Intj(Aj|(ON1,ΦAj ∧ Φ)) Intj((Aj|ObN1Aj)) Note that in the schema N1 is a constant (denoting the particular normative system the agent intended to comply with), j is a constant (denoting a particular individual) while A is a variable over possible actions, and Φ is a variable over possible states of affairs (conjunctions of formulas describing factual situations), is over possible factual worlds.

• G. Sartor (CIRSFID-EUI)

Compliance 13 / 26

Giovanni Sartor

Inference of intention to comply

Premise 1: John has the obligation to compensate Tom according to normative systems N1 in factual situation C1 = ETomdamagedJohn, which is the case: ON1,C1(¬EJohnCompensated(Tom)) ∧ C1 Premise 2: John has the intention comply with N1 Intj((Aj|ObN1Aj)) Conclusion: IntTom(EJohncompensated(Tom)))

• G. Sartor (CIRSFID-EUI)

Compliance 14 / 26

Giovanni Sartor

Intention not to comply

Indifferent agent: (¬Intj(Aj|ON1Aj)) Diabolic agent: Intj(Aj|ON1Aj)

• G. Sartor (CIRSFID-EUI)

Compliance 15 / 26

Giovanni Sartor

Motivation for adopting the intention to comply

Consequential self-interest altruism common good Deontological reciprocity sense of duty, etc. Is it a collective intention?

• G. Sartor (CIRSFID-EUI)

Compliance 16 / 26

Giovanni Sartor

Consequentialist chooser

Definition (Consequentialist chooser)

A consequentialist chooser x will intend to do an action Ax whenever x believes that the expected utility of (the consequences of) doing that action is superior to the utility of not doing it: Belx(ux(Ax) > ux(Ax)) → IntxAx From the internal perspective, a consequentialist chooser will adopt the policy Intx(Ax|(ux(Ax) > ux(Ax))

• G. Sartor (CIRSFID-EUI)

Compliance 17 / 26

Giovanni Sartor

Utility

What is utility (Sen 2007) self-centered welfare (x only cares about his consumption, indifferent to others) self-welfare goal (x’s only aim is to satisfy his welfare, which may depend on the other people’s welfare though sympathy) self-goal choice (x’s acts in order to satisfy his goals, which may be altruistic, communitarian, etc.) commitment (x’s choice of goals is governed by x’s commitments, meta-level goals or constraints on goals)

• G. Sartor (CIRSFID-EUI)

Compliance 18 / 26

Giovanni Sartor

A simplified framework

Distinguish the behaviour’s impact on the agent’s welfare, from the utility given all the relevant impacts that behaviour has in the goals that matter to him, there included the values the person has adopted. wi(Aj) denotes the welfare of agent i produced by action Ai we

i (Aj) denotes the self-centred welfare of agent i

ws

i (Aj) denotes the sympathetic welfare of agent i

ui(Ai) denotes the utility that agent i attributes to its action Ai

• G. Sartor (CIRSFID-EUI)

Compliance 19 / 26

Giovanni Sartor

Kinds of agents

Self-welfare self-centered agent: ui(Ai) = we

i (Ax).

Self-welfare sympathetic agent: ui(Ai) = we

i (Ax) + ws i (Ax).

Self-goal altruistic: ux(Ax) = wy1(Ax) + · · · + wym(Ax), where y1 . . . ym are the agents x considers relevant to its choice.

• Communitarian. ux(Ax) = wg(Ax), where g is the community x cares

about.

• Utilitarian. ux(Ax) = wy1(Ax) + . . . wyn(Ax) where y1 . . . yn are all

human beings. Generic/objectivistic ux(Ax) = n

1(goodnessi(Ax)) where

goodnesi(Ax) is the merit of doing Ax

• G. Sartor (CIRSFID-EUI)

Compliance 20 / 26

Giovanni Sartor

Utility and compliance

It is more useful to comply than never to do it: uj([ONAj → Aj]) > uj([ONAj → ¬Aj]) What about complying sometimes? But when? Maybe it is more useful to have the policy of complying uj(Intj[Aj|ONAj]) > uj(¬Intj[Aj|ONAj)]) Adopting determination to comply: I intend to adopt the policy to comply in case is more useful to adopt the policy to comply, then not to do so, Intj(Ej(Intj[Aj|ONAj])| uj(Intj[Aj|ONAj]) > uj(¬Intj[Aj|ONAj)])

• G. Sartor (CIRSFID-EUI)

Compliance 21 / 26

Giovanni Sartor

Policy to comply and its exceptions

j has the policy of complying. Intj[Aj|ONAj] but makes exceptions to it: ¬Intj[Aj|(ON ∧ uj(Aj) < uj(Aj))] According to how utility is determines j can be a bad man or an opportunistic violator j: ¬Intj[Aj|(ONAj ∧ wjAj < wjAj)] an altruistic ¬Intj[Aj|(ONAj ∧

n

• 1

(wi(Aj)) <

n

• 1

(wi(Aj))]

• G. Sartor (CIRSFID-EUI)

Compliance 22 / 26

Giovanni Sartor

Why should I believe that it more useful to comply (and to intend to comply)

Obligatoriness as a (causal?) reason for the utility to comply ONAj ⇒r (u(Aj) > u(Aj)) Obligatoriness as evidence for the utility to comply (u(Aj) > u(Aj)) ⇒c ONAj ONAj ⇒e (u(Aj) > u(Aj)) The (moral) utility of doing A, factually causes N to contain norms entailing O(Aj) (N reliably tracks moral reasons). Thus O(Aj is evidence for (moral) utility-goodness

• G. Sartor (CIRSFID-EUI)

Compliance 23 / 26

Giovanni Sartor

Compliance as a coordination game

The greater the number of compliers, the greater the utility of complying ONAS ∧ S1 ⊂ S2 ⊆ S ⇒ uj(AS1) < uj(AS2) There is a threshold St s.t. the utility of complying becomes positive ONAj ⇒ uj(Acj, ASt−j) < uj(Aj,St)

• G. Sartor (CIRSFID-EUI)

Compliance 24 / 26

Giovanni Sartor

A Kantian approach

I shall comply whenever it would be better if all complied. Inti[Ai|(ui(∀xAx) < ui(∀xAx)]

• G. Sartor (CIRSFID-EUI)

Compliance 25 / 26

Giovanni Sartor

Intention to comply

I intend to comply if the (or most) others comply as well I intend to comply as a member of a complying collective. When is this the case? It is an empirical issue

• G. Sartor (CIRSFID-EUI)

Compliance 26 / 26