=Paper=
{{Paper
|id=None
|storemode=property
|title=The Agile Project: Reconciling Agility and Legal Accountability
|pdfUrl=https://ceur-ws.org/Vol-582/paper4.pdf
|volume=Vol-582
|dblpUrl=https://dblp.org/rec/conf/ict4justice/BoerE09
}}
==The Agile Project: Reconciling Agility and Legal Accountability==
https://ceur-ws.org/Vol-582/paper4.pdf
The Agile Project:
Reconciling Agility and Legal Accountability
Alexander Boer1 and Tom van Engers1
1 Leibniz Center for Law, University of Amsterdam,
Kloveniersburgwal 48 1000AB Amsterdam, The Netherlands
{aboer,vanengers}@uva.nl
Abstract. While every IT professional knows from experience that han-
dling change is much harder and more expensive than building IT systems
from scratch, only little scientific attention has been paid to prepare for
change. This issue is of great importance since usually 80-90% of the total
cost of ownership (TCO) of IT systems are maintenance costs. In our net-
worked society changes may result from different sources. The recently
started AGILE project addresses the legal dimension of management of or-
ganizational change processes. This paper introduces the AGILE project,
and discusses the relation between sources of law and the business
processes and services of the administrative organization.
1 Introduction
ICT has penetrated most organizations. It is an essential asset for public ad-
ministration, and is becoming one for the judiciary. Administrative organizations,
including law enforcement organizations, dealing with taxation, immigration, etc.
have been using ICT to improve the efficiency of case handling (primarily aimed
at reducing handling time and operational costs) and the quality of services, a de-
velopment also known as eGovernment. ICT is also used to make law enforce-
ment more effective (see e.g. [6]). Also in the judiciary we see developments aim-
ing at improving efficiency and effectiveness (see e.g. [4]).
In the judiciary the application of Artificial Intelligence techniques is still very
limited. Application of case management and ePublishing have become common
and found their way to courts and law firms, but application of AI based applica-
tions, e.g. natural language processing for making sense of blogs with legal com-
ments, or applying automated classification of email and intelligent routing and
possibly preparing elements that could be used for solving the legal issues at hand,
are still to be developed and implemented.
A wider range of Artificial Intelligence techniques have found their way into
organizations. In this paper we focus on those applications for several reasons:
Post-proceedings of the 2nd International Conference
on ICT Solutions for Justice (ICT4Justice 2009) 41
1. Similar solutions will be applied in the judiciary in the near future;
2. These applications are legal decision support applications and the knowledge in
those systems is at least to a large extent legal knowledge being either de-
clarative or procedural;
3. These applications have big impact on organization design (rule management)
and an increasing impact on legal drafting.
These legal knowledge representation applications are interesting from a tech-
nical perspective but even more from a design perspective. Especially the separa-
tion of different forms of knowledge and the explicit relations to the sources of
law can help to significantly reduce the TCO (see also [6,4]). This reduction of
TCO is mainly possible because of the increase in traceability to the sources of
law. Knowledge representation promises to reduce the time needed to establish the
impact of change caused by changes in the sources of law.
Currently the interest in legal knowledge representation in public administra-
tions is gradually increasing but also changing in nature. The increasing legal con-
vergence and legal complexity, an increasing pace of organizational change in
public administration drives this growing interest, and increased use of IT and web
services. In this paper we will describe the shift from the utility of fielding com-
puter systems built using a knowledge engineering approach to the potential utility
of knowledge representation for comparative and for maintenance purposes, and
for increasing the efficiency of the organizational change process itself.
Compared to the standards set by knowledge engineering research, fielded sys-
tems in public administration and elsewhere that use explicit knowledge represen-
tation to support decision making processes are technically and theoretically
straightforward. The required transparency and the great challenge real world
knowledge representation poses for the people implementing such systems, act as
a natural limit to the complexity of these decision support systems. The required
functionality rarely by itself justifies state-of-the-art legal knowledge representa-
tion. As we will argue in section 2 and 3, however, the knowledge management
problems involved in managing the consequences of organizational change proc-
esses triggered by changes in the law, for business processes, services, databases,
fielded applications, forms and documents, internal education, etc, make a consid-
erably better case for some state-of-the-art concepts in legal knowledge represen-
tation.
The recently started AGILE project, introduced in section 4, addresses man-
agement of institutional change, sometimes driven by changing legislation and
sometimes by environmental factors. This paper introduces the AGILE project and
discusses the relation between sources of law and the business processes and ser-
vices of the administrative organization. This account is a first, tentative step to-
wards a design method that should help organizations to adapt to new or changing
legislation.
42
2 Background
Inside public administrations, and on the interfaces between them, ICT and In-
ternet have a large impact. Some decision making processes are nowadays assisted
by computer applications, and others are more or less autonomously performed by
the computer.
At the same time, service-oriented architectures are becoming the prominent
paradigm for building enterprise information systems, also in administrative agen-
cies. Service-orientation leads to new network arrangements between administra-
tive agencies for sharing data, etc. This development in itself leads to attention for
the adaptability and accountability issues that arise (cf. [5]).
A major category of those services are of course those performed for the main
client categories of the organization, often generally the citizen. The implementa-
tion of these services on the Internet is what is often understood in the narrow
sense by e-Government.
The services in question are in an administrative setting often implementations
of public legal acts, performed by public legal personalities, based in formal legis-
lation. Legislation gives administrative organizations public personality, defines
what the core functions of public organizations are, and what services they pro-
vide. It guides how the organization subdivides itself into administrative units,
how it organizes business processes inside the organization, and eventually how
the functions of the organization are realized by civil servants and computer sys-
tems.
Business process design and design of specialized computer systems are both
usually based on explicit models in modeling languages, like the Unified Model-
ing Language (UML) or, more recently in the context of the Semantic Web, the
Web Ontology Language (OWL), of what the business process or computer appli-
cation should achieve. These models are supposedly used as a specification of the
objectives of an organizational change process or application development proc-
ess. When legislation changes, these models are updated, and the organization's
structures and computer programs have to be changed to conform to the models.
In the past these changes were conceived of as temporary interruptions of long
periods of everything staying the same. This was certainly the case when the adap-
tation of existing systems was still considered a frightening prospect: things did
change but the changes where carefully orchestrated to not impact existing proce-
dures, network arrangements with other organizations, and computer systems. But
as the perceived capacity of organizations to organize change processes increases,
and the number of fielded computer applications increases, so does the pace of
change in legislation directly affecting existing computer applications.
Tax legislation is for instance changed every year, leading to continuous adap-
tation of relevant computer applications for next year and the years after that,
while the legislation of the present year and previous years is still being applied.
In the business process design literature, awareness of this phenomenon has led to
a new conception of the organization as an entity that is constantly in the process
43
of changing: the organization is constantly conceptualizing and comparing what it
is and what it is becoming.
Attention for knowledge representation of sources of law is very often trig-
gered by such administrative change processes driven by new legislation and other
sources of law, such as case law and internal written policies.
The change processes triggered by the legal system are increasingly expensive,
especially if they involve changes in ICT infrastructure. Knowledge representation
is seen as a means to potentially reduce costs and increase efficiency through in-
creased control over the knowledge dimension of the change process. Our past
work for the Dutch Tax and Customs Administration (DTCA; cf. for instance [3,
6]) was for instance clearly related to the huge change process triggered by the
complete overhaul of the Dutch income tax law in 2001. The Juridisch Loket (cf.
[7]) project on pro bono legal assistance, and the DURP project on spatial plan-
ning (cf. [2]) were for instance also driven by an overhaul of legislation.
The law does not, however only play a top-down role, as knowledge in deci-
sion making and a yardstick for performance. Also the feedback to the legislator
plays an increasingly important role (in the DTCA case notably; cf. generally [6]).
There are certain traditional forms of feedback, like lessons learned about effec-
tiveness, efficiency, and enforceability, and information about relevant changes in
the environment of the organization which threaten enforceability, effectiveness,
or efficiency. In addition the increasing formalization of network arrangements
between involved actors spurred by the possibilities of ICT tends to lead to re-
quests to the legislator to give such agreements and technical arrangements a basis
in law, so that the partners on which the organization depends are bound to pro-
vide service as the organization prefers it.
Whether the legislator can play a constructive role in organizing public ad-
ministration depends on the quality of this feedback. The nature of this feedback
moreover strongly depends on the agility of the organization and it's infrastruc-
ture; The less prepared the organization is for change, the more problems it will
encounter in implementing changes in law effectively, and it will surely be in-
clined to demand legal formalization of current network arrangements in order to
buttress it's own current business processes.
Direct feedback from the responsible organizations about shortcomings of pro-
posed legislation, as uncovered during modeling or simulation, played an im-
portant role in our work for the DTCA (cf. generally [6]).
The AGILE project, described in this paper addresses a more general problem
conception: that of balancing the requirements of accountability to the law with
the agility of the organization in a changing environment. But before explaining
the aims of and approach taken in Agile we will first describe the general problem
of deciding how to adapt to changes in the law and the interpretation problem.
44
3 Changes in the legal system
Adapting to changes in the law is a challenge for the Public Administrations that
then have to adapt their decision support systems that are commonly used for mas-
sive case handling. Already in the eighties and nineties of the previous century ar-
chitectures and design methods have been proposed that enhance legal effec-
tiveness (see e.g. [6]). Following a knowledge engineering tradition these ap-
proaches have focused on the knowledge representation problems and the techni-
cal aspects that come with automated (legal) reasoning, rather than focusing on the
process characteristics that affect the sources of law those systems are build upon.
Needless to say that the way systems are build have great impact on their main-
tainability and adaptability but this is only part of the puzzle.
The authors of this paper, having been involved involved in many large and
complex systems design projects, have noticed that many of the insights have still
had little impact on the actual way systems are designed and produced. Unfortu-
nately relationships between the specifications and their sources, or between the
specifications and the components that realize them, or simple things like version-
ing and distinctions between design, test, accepted and production status are in
many cases still lacking. This might explain the high exploitation cost of many of
our governmental IT systems.
In this paper however we assume that despite the backlog that still exists in de-
sign practice, the technical architecture and methodological issues are relatively
small problems and almost solved. The remaining major problem therefore is the
interpretation problem. Before we can translate the sources of law that are ex-
pressed in natural language in a computational formalism, we should decide on the
intended behavior or meaning of the constitutive elements of the norms that are
expressed in the legal sources. As we will argue interpretation of the constitutive
elements of the norms is rather problematic. How serious the problem is can be
showed by looking at so-called ‘open norms’
Concepts such as ‘undue hardship’ or ‘generally accepted conduct’ obviously
require an operationalization by legal experts first before the norms that use these
concepts can be transformed into a computational formalism. Alternatively one
could decide to exclude such norms from the automated system that is to make de-
cisions automatically, and leave the interpretation to civil servants, but that would
imply that for every individual case we should find a specific interpretation.
The clear advantage of such approach would be that the civil servant could take
into account all relevant contextual information. So potentially this could lead to a
more justifiable decision. However on the other side this may lead to unfair-ness
and inequality before the law. Another clear disadvantage is that stakeholders can-
not make reasonable predictions about the outcome of the decision.
If one wants to include the ‘open norm’ into an automated system these norms
have to be translated into ‘closed norms’. For closed norms the interpretation, i.e.
the semantics are completely clear. One naively could argue that the legislator
should avoid the use of open norms for the purpose of easy implementation, but
45
this would neglect the fact that the law, being a steering instrument for society,
will always need to be adapted to new circumstances. In order to prevent frequent
changes of the law legislators sometimes prefer broader or open concepts. Open
norms are also sometimes used to allow for different interpretations and thus cir-
cumvent potential political difficulties in the parliament.
However it is not just the legislator that has to decide on the desired interpreta-
tions of norms and the consequent choice in using closed norms or open ones, ac-
cepting that in the latter case the administrations will have larger freedom in inter-
preting them. We should realize ourselves that the interpretations discussed so far
are ad-hoc interpretations. But since one can always defeat decisions up to the
highest court, court decisions should also be taken into account since they reflect
further interpretations. The problem for the administrations is that it may take a
while before such decisions are made and acting on the impact these court deci-
sions should have on the handling of other cases is not always easy.
If we look more closely at the interpretation problem we can observe that the
hard problem is actual the mapping between the brute reality (the real world) and
the institutional reality (the law), which is a qualification problem and depends on
the aims and preferences of the person that makes that mapping. Legislators, pol-
icy makers in administrative organizations, and judges have a particular goal in
mind when choosing an interpretation, but their plans may be wrong, i.e. their un-
derstanding of the world or of the effects of their interpretation doesn’t lead to that
desired goal.
While some attempts have been made to simulate effects of regulations, the ap-
plication of such simulations is still very limited. Unfortunately only a few judges
are aware of the relationships between the individual case decision at hand and the
potential effects on massive client handling by the administrations, and rarely the
published case decision contains indications about how to handle similar cases.
Legal quality in general would benefit if judges would consult the administra-
tions about the (undesired) effects of their draft case decisions and make this clear
in their decisions. This process could be formed in analogy to the process that leg-
islators use to ask advice to both administrative organizations and the courts about
potential operational effects of draft legislation.
Interpretation of norms is a challenge for legislators, the public administrations
and the courts, as the above shows. If we want to improve legal effectiveness and
take into account the need for adaptivity we should not limit ourselves to just the
technical aspects of the supporting ICT-systems. ‘Rule management’ as it is called
today, should particularly address the organization of the processes that should be
aligned in order to achieve an agile solution.
In the next section we will shortly explain the Agile project in which we ad-
dress these problems.
46
4 The AGILE Project
In the AGILE project (acronym for Advanced Governance of Information ser-
vices through Legal Engineering) we aim at developing a design method, distrib-
uted service architecture and supporting tools that enable organizations - adminis-
trative and otherwise - to orchestrate their legal information services in a net-
worked environment. The AGILE project started in the second half of 2008 and
will last for four years.
At issue is the adaptability of ICT infrastructure, of business processes, and of
data and knowledge within the organization, given changing legal demands and
constraints.
4.1 Complex Adaptive Systems
Based on complex adaptive systems (CAS) theory, the project will develop a ser-
vice modeling and design method that should help organizations to adapt to new
or changing legislation. The proposed method should take the resilience of exist-
ing systems, dependencies on the environment, and the unpredictability of change
processes explicitly into account (cf. generally [5]). The objective of this part of
the project is to improve the adaptability of ICT infrastructure, of business
processes, and of knowledge in the organization.
The essence of CAS theory is the study of systems built of individual agents
that are capable of adapting as they interact with each other and with an environ-
ment, in order to understand how the individual affects system-level responses.
The underlying premise is that simply determining future needs and require-
ments is not the right approach, due to the inherent unpredictability of a complex
environment and the fact that there are already many working (social and informa-
tion) systems in place which cannot and should not be ignored.
4.2 Legal Concepts
The models to be developed in the project are intended for validation, and – in
combination with an agent planning framework – for simulation of potential im-
plementations. Direct deployment of services – which is possible with the selected
Semantic Web technology – in a newly designed infrastructure is - as one should
expect in a project definition that embraces gradual and continuous change - not
explicitly intended.
Of specific relevance to a world dominated by written declarations and deci-
sions, databases, web services, and changing sources of law is an account of for-
mal acts, and of the act of providing evidence for a legally relevant proposition. In
this account the concepts developed in the MetaLex standardization effort for in-
stance play an important role.
47
The design methodology to be developed in AGILE will cover modeling con-
cepts found in legal knowledge representation such as realization, reference, rep-
resentation (from MetaLex), applicability, and constitutiveness (from our work in
[1]).
The best-understood legal concept is undoubtedly the norm: there are certain
things the organization is not allowed to do. This aspect of law is however also of
ephemeral interest for AGILE. In the AGILE project it is rather the projection of
organizational reality on abstract legal reality that is of interest.
4.3 Pilot Studies
Results from the research tracks discussed will be tested in the context of two ac-
tual business cases. One at the Dutch Immigration and Naturalisation Service
(IND) and one at the Dutch Tax and Customs Administration (DTCA).
In both organizations, timely and efficient adaptation to changing legislation,
case law, and patterns of behaviour accommodating or evading law in the relevant
environment is seen as a main organizational objective.
5 Discussion
To demonstrate the potential utility of legal knowledge representation tech-niques
for increasing the manageability of the organizational change process, sev-eral pi-
lots are planned within organizations that do have to deal with such regular
changes. In the IND the most obvious impetus is from changes in migrant and ref-
ugee streams, which always lead to new legal issues. In the DTCA change is pre-
dictable as a clock: taxation appears to change as often as the legislator believes
the DTCA can handle it, and sometimes more often.
A major obstacle for assessing the utility of the design method developed in
AGILE is of course the time frame: whether the investment in knowledge repre-
sentation (the TCO, so to speak) is worthwhile can only be judged in relation to its
lifespan. The pilots should in this respect be considered formative evaluations.
Finally, the critical reader may wonder to which extent the proposed interpre-
tation of the relation between law and public administration is generally applica-
ble. It is certainly the case that there are considerable differences between ap-
proaches to law making that need consideration.
How the legal system deals with complexity is in the end the less important fac-
tor; The external factors that increase complexity of change processes – legal con-
vergence, increasing dependence on network arrangements with other actors, in-
creasingly costly infrastructure – are certainly present all over Europe and, per-
haps to a lesser extent, the rest of the developed world.
If we want to make administrative organizations more agile we shouldn’t how-
ever limit ourselves to developing ICT-tools en methods. It is equally important to
48
strengthen the links between the legislator, the policy departments of the admini-
strations, and the courts in order to develop shared ad-hoc interpretations. The
same holds for those situations in which post-hoc adjustments have to be made.
Also in those post-hoc cases the legal system would benefit from taking a broader
perspective. Both cases would benefit from an architecture that clearly distin-
guishes the different interpretation problems and allows for testing, i.e. simulating
effects of such decisions. With the Agile project we hope to set a first step towards
such a environment.
AGILE is a Jacquard project funded by the Netherlands Organization for Scientific Research
(NWO). In the AGILE project, The Leibniz Center for Law of the University of Amsterdam
cooperates with the Technical University of Delft, which has experience in the application of
CAS theory to organizations. The IND and two companies, O&I and BeInformed, provide
matching effort to the project. We would also like to thank the DTCA for its cooperation in the
project.
References
[1] T. van Engers. Legal engineering: A structural approach to improving legal quality. In A.
Macintosh, R. Ellis, and T. Allen, editors, Applications and Innovations in Intelli-gent Sys-
tems XIII, proceedings of AI-2005, pages 3–10. Springer, 2005.
[2] Hilhorst, R.A., Van Engers, T.M., 2009, E-dossiers at the Dutch Council of State, in Proceed-
ings of the 12th International Conference on AI and Law (ICAIL), ACM, ISBN 1-60558-
597-0.
[3] M. Janssen. Adaptability and accountability of information architectures in interor-
ganizational networks. In ICEGOV ’07: Proceedings of the 1st international confe-rence on
Theory and practice of electronic governance, pages 57–64, New York, NY, USA, 2007.
ACM.
[4] Boer, T. van Engers, and R. Winkels. Using Ontologies for Comparing and Harmoniz-ing
Legislation. In Proceedings of the International Conference on Artificial Intelli-gence and
Law (ICAIL), Edinburgh (UK), 2003. ACM Press.
[5] T. van Engers, R. Winkels, A. Boer, and E. de Maat. Knowledge management and the dutch
legal aid service counter. In J. J. Schreinemakers and T. van Engers, editors, Advances in
Knowledge Management, volume IV, Würzburg, 2006. Ergon Verlag.
[6] Boer, T. van Engers, R. Peters, and R. Winkels. Separating law from geography in gis-based
egovernment services. Artificial Intelligence & Law, 15(1):49–76, 2007.
[7] Boer. Legal Theory, Sources of Law, & the Semantic Web. Frontiers in Artificial In-
telligence and Applications 195. IOS Press, 2009. To appear.
49