A Multi-agent System for Managing the Normative Consolidation
4th International Workshop on Law of Electronic Agents ICAIL2005 Bologna, Italy, 10 June 2005 Monica Palmirani & Regis Riveret
CIRSFID - University of Bologna (Italy)
C.I.R.S.F.I.D University of Bologna Research Centre of History of Law, Philosophy and Sociology of Law, Computer Science and Law
Objective
To present a model of normative information system based on a multi-agent technology The aim is:
to manage the legislative consolidation process in a proactive way to detect inconsistencies in the modification norms to resolve the modifications linked to uncertain future events
To use a logic framework oriented to the legal reasoning in the semantic web meaning
2
LEA2005
Outline
Part I – Palmirani Introduction of the problem To present a time model for representing the dynamicity of the normative system (not a legislative system) To introduce a model for producing in semi-automatic way legislative document updated respect three temporal axes: law in force, efficacy, applicability Part II - Riveret To present the logic framework based on the defeasible logic To present the multi-agent architecture
LEA2005
3
Introduction of the problem (i)
Several legal information systems (LIS) are managing the consolidation of legislative documents because there is a crucial request from the eGovernment and eDemocracy research field [eEurope2002, eEurope2005, Better lawmaking EU initiative, White paper on governance, DEXA2004] On the other hand these LIS face the problem mostly oriented to the document not to the normative system dynamicity [MetaLex-NL, LexDania-DK, LexGo-CH, NormeInRete-IT, RIS-AT, ADELE-FR]
4
LEA2005
Introduction of the problem (ii)
Furthermore the conditional provisions and other uncertain modification provisions arise some questions about the fitted model and architecture to adopt in order to cope with the consolidation process
Several studies [Boella 03-a, Boella 03-b] suggest that we can model a normative systems using a multi-agent schema enabling interaction among agents whose behavior is autonomous, reactive, and proactive
LEA2005
5
Updating law corpora
For updating law corpora we should fix some criteria:
1. define a taxonomy of modifications
2. apply the modifications in the correct sequence
3. define the behaviors of the normative system in case
of modifications
It is fundamental to define a time-model for representing the temporal axes:
in force
efficacy
applicability
6
LEA2005
1. Taxonomy of modifications
Modification of the text
repeal substitution integration relocation
Norm Modifications Modification of the norm range
interpretation meaning change derogation extension changes of force changes of efficacy
7
Modification of the time Normative System Modifications
LEA2005
1. Taxonomy of modifications (ii)
Modification of the time Changes of validity a) Limitation of validity • invalid norms b) Extension of validity • prorogue norms • renewal norms Changes of efficacy a) Limitation of efficacy • suspend norms b) Extension of efficacy • postpone efficacy • retroactivity • extra-activity
LEA2005
8
2. Correct sequence
The correct sequence is determined by: the sequence of modifications in the time a set of time-information included in the legislative text the analysis of the conditional provisions the analysis of the impacts on the normative system
LEA2005
9
Normative chain of the document A
Time line
t0 Document t1 t2 ti ti+1 tn+1 tn
A0
A1
A2
….
Ai
Ai+1
….
An-1
An
Set of Modifications
M0
M1
….
….
Mn-1
f(Aj-1, Mj-1) = Aj versioning chain normative chain
LEA2005
vc(A) = An , n = [0, ∞[ nc(A) = vc(A) + (vc(Mj))
10
Legislative system: a definition
LS(t) = {a system of norms N that satisfy at least one membership criterion at a given time t} Legislative system is the synchronic view of the norms [Guastini 1998]
LS(t) = {D1(t) , D2(t), D3(t), ... Dm(t)} m N, Dm=document t is a fixed time in a discrete representation
LEA2005
11
Legislative system at the time tj
Time line
t0 Document A t1 t2 ti ti+1 tn+1 tn
A0
A1
A2
….
Ai
Ai+1
….
An-1
An
Document B
B0
B1
….
Bk
….
Bk+1
Bn-1
Bn
………………………… Document D
D0
D1
….
Dz
Dz+1
….
Dn
LEA2005
12
Normative system: a definition
Normative system is a sequence of legislative system in the time [Guastini 1998, Alchourrón 1971]
NS = {LS(t1) , LS(t2), LS(t3), ... LS(tj)} j N
The normative information system is an IS that include: all the original legislative texts all the modifications and the versioning chain or better all the normative chain of all the texts Normative information system = nc(Dj)
LEA2005
13
Normative information system
Temporal line
t0 t1 t2 ti ti+1 tn+1 tn
Legal system
LS0
LS1
LS2
….
engine
LSi
LSi+1
….
LSn-1
LSn
Set of Modifications
M0
M1
….
Mi
….
Mn-1
In force
LEA2005
efficacy
applicable
14
Times in the normative provision
part of text with a normative meaning is a normative provision It is possible to consider a normative provision also as a logic proposition (true or false) composed by two parts: cause and effect Cause Effect Facts Consequences if the conditions are verified then some effects are expected
"Murder is punished." A true proposition: "If F (you kill), then E (you will be punished)."
LEA2005
15
Anatomy of a normative provision
So we have three times: we have the time when the enunciates are included in the “corpora” – date of force we have the time when the disposition is active for the fact-type – date of efficacy we have the time when the disposition produce results – date of application usually date of force =date of efficacy= date of application but not always e.g. retroactive norms
LEA2005
16
Toy Example
“for all the students that in the year 2003-2004 had a rate >28/30, the University of Bologna will give a special grant for the year 2004-2005” Time of force June 2005 Time of efficacy 2003-2004 Time of applicability 2004-2005
LEA2005
17
When a disposition is a modification
The period of application of the modification can be:
instantaneously postponed – in the future anticipated – retroactivation effect
or
conditioned to an event certain uncertain
LEA2005
18
Modifications
hasTime
Condition
hasCondition
Times
PassiveTime SpaceConditions CaseConditions EventConditions -inforce -efficacy
ActiveTime -app_date -duration
Certain
unCertain
instantaneousMod
futureMod
pastMod
LEA2005
19
Example of conditioned modification
L. n. 328 del 8/11/2000 (in force since: 28/11/2000) Art. 30 Repeals.
....
2. Since the enter in force of the legislative decree declared in the art. 10, all the provisions about IPAB in the law 17 July 1890, n. 6972 are repealed. Since the enter in force of the legislative decree declared in the article 24, the provisions about the economic incomes in the Law 10 February 1962, n. 66, Law 26 May 1970, n. 381, Law 27 May 1970, n. 382, Law 30 March 1971, n. 118, Law e 11 February 1980, n. 18 and following modifications are repealed.
LEA2005
20
3. Some crucial modification types
The normative information system should manage:
Conditioned modifications and the related events The abrogation of abrogation Retroactive abrogation Annulment of a modification by the Constitutional Court decision “Authentic interpretation” that is retroactive for all the normative system Decree not converted in law Substitution of textual parts concerning time definition Modification of modification Modifications that have cycle effect on the normative system
LEA2005
21
Forking model
Manage ex-tunc modifications Fork the versioning lines: One time-line to register what happen after the court decision with effect ex-tunc (retroactive) in the past On time-line to register what was happened during a period in the alternative-normative system
e1
e2
now
LEA2005
22
Current applications
Build in semi-automatic way the law in force NORMA-SYSTEM project: client/server architecture
Norma-Editor: mark-up in XML in semi-automatic way the text Norma-Server: manage all the versioning system
Application in the Italian Court of Cassation: 200.000 legislative texts
IR respect three different axes: Law in force Efficacy Applicability
LEA2005
23
Future applications
Formalise the conditional provisions and modifications in OWL Define a Logic framework based on defeasible logic Define a level of proof Implement a Multi-agent system able to verify the conditions and to make proof
METADATI DTD URN
LEA2005
24
PART II – Régis Riveret
LEA2005
25
Logic Framework for consolidation
Non-monotonic reasoning Choice of Defeasible Logic
suitable for legal reasoning manage incomplete and uncertain facts linear computability suitable with the logic layer of the semantic web Inference engines (DR-Device)
26
LEA2005
A theory in Defeasible Logic
Facts
Inforce(Law n.50/99) Law(x) → StateRule(x) Repealed(x) => ¬Inforce(x) Conditioned prorogation(x) ~> Inforce(x)
Strict rules
Defeasible rules
Defeaters
Ranking
StateRule > FederalRule
LEA2005
27
Time integration by two means
Literals labelled by time instants
Inforce(x) :t
Persistent and transient rules
Condition :t =>p effect :t Condition :t =>trans effect :t
LEA2005
28
A toy theory
R1: enter_inforce(x) : t =>p inforce(x) : t+15 R2: repeal(x) : t =>p ¬inforce(x) : t R3: Conditional Prorogation (X): t ~>p inforce(x): t R4: revival(x) : t =>p inforce(x) : t
R4>R2>R1 R3>R2
Enter_inforce(Law50/99) : 0 Repeal(Law50/99) : tn Revival(Law50/99) : tz Cond. Prorogation (Law50/99) : tn+1
LEA2005
0
tn
tn+1
tz
29
A normative system
mod
Document repository A
LEA2005
Document repository B
30
Agents ?
Active norm + events
Proactivity Reactivity
Passive norm
Unpredictable normative environment + adaptation
Autonomy Sociability Model the complexity
31
Complex interactions + Collaboration work
LEA2005
Legal operator
Proof agent
Consolidation agent
Normative agent
Event agent
Node A
LEA2005
Node B
32
thank you for your attention
Monica Palmirani - palmirani@cirfid.unibo.it Regis Riveret – rrivert@cirfid.unibo.it
LEA2005
33