=Paper=
{{Paper
|id=Vol-94/paper-5
|storemode=property
|title=Ontology Modelling in the Legal Domain - Realism Without Revisionism
|pdfUrl=https://ceur-ws.org/Vol-94/ki03rao_vandenberghe_al.pdf
|volume=Vol-94
|authors=Vim Vandenberghe,Burkhard Schafer,John Kingston
|dblpUrl=https://dblp.org/rec/conf/ki/VandenbergheSK03
}}
==Ontology Modelling in the Legal Domain - Realism Without Revisionism==
https://ceur-ws.org/Vol-94/ki03rao_vandenberghe_al.pdf
Ontology Modelling in the Legal Domain -Realism Without Revisionism? 1
Wim Vandenberghe, Burkhard Schafer and John Kingston
Joseph Bell Centre
Edinburgh University, School of Law
Old College, Edinburgh EH8 9YL
www.cfslr.ed.ac.uk
Abstract
This paper discusses problems and issues in research on detection and prevention of financial
fraud undertaken as part of the European Commission funded FF POIROT (Financial Fraud:
Prevention Oriented Information Resources Using Ontology Technology) project. The goal of
the project is to build a detailed ontology of European Law, of preventive practices and of
knowledge of the processes of financial fraud within the European Union. It aims at compiling
for several languages (Dutch, Italian, French and English) a computationally tractable and
sharable knowledge repository (a formally described combination of concepts and their
meaningful relationships) for the financial fraud domain. The paper explores in particular just
how “heavy” an ontology needs to be to meet the needs of the various stakeholders in the
criminal justice system, one of the as yet unresolved problems in the multi-partner, multi-
jurisdiction and multi-profession project.
1 FF Poirot
It is estimated that the EU loses several billion euros per year due to financial fraud.
Therefore it should not come as a surprise that prevention and early detection of fraudulent
activity is an increasingly important goal for the EU and its Member States.2 The impetus for
building a financial fraud ontology results from the need to supplement the efforts of EU
Member States to combat financial fraud, and especially to improve the cross-border co-
operation between investigating and prosecuting agencies. FF POIROT is focusing on two
areas of fraud: VAT (sales tax) fraud, which is the example used in this paper, and
unauthorised online investment solicitation.
The detection of EU VAT fraud presents difficulties not encountered in areas such as,
for example, the credit card industry. Whereas credit card fraud comes to light fairly early on,
in EU VAT fraud it takes months, sometimes years, before individual transactions can
definitively and legally be identified as fraudulent. Two examples of VAT fraud are “carousel
fraud” and “missing trader intra-community fraud”. In carousel fraud, goods are apparently
exported, thus making them exempt from VAT, but the export does not actually take place;
the transaction may then be reversed in the “receiving country”, thus doubling the fraud while
leaving stock levels in both countries unaffected. Missing trader intra-community fraud
(MTIC fraud) is carried out (in general terms) by setting up fake or minimum-sized
companies, which trade across borders in such a way that one becomes liable to repay VAT
to the government, while the other has already claimed a VAT refund – and these companies
then disappear without trace. In practice, several layers of buffer companies may be used.
1
This work was supported under the IST project FF POIROT (Financial Fraud Prevention Oriented Information
Resources using Ontology Technology), which is sponsored by the European Union's Information and Systems
Technology Directorate (5th framework) under grant number IST-2001-38248
2
See Communication from the Commission, Protecting the Communities’ Financial Interests. Fight Against
Fraud. Action Plan for 2001-2003, COM(2001) 254 final. A recent effort by the UK government is the initiative
by HM Customs & Excise, titled “Protecting Indirect Tax Revenues”, designed to save £2 bn a year.
Complying with VAT regulations requires knowledge of a vast web of regulations, and is
therefore a heavily knowledge-based task. Detecting VAT fraud appears to require less
knowledge; once all the information is available, a numerical comparison between transaction
sizes or identification codes on invoices, or a failure to match one invoice with another, will
reveal the fraud. The European Community has developed a cross-border database (known as
VIES, the VAT Information Exchange System).3 The problem is in making the information
available in a timely fashion. The current operation of VIES does not allow sufficiently early
availability of such information.
However, the task of fraud prevention – using profiles of typical fraudsters or typical
fraudulent transactions to direct inspectors towards possible frauds at an early stage – appears
to be an area much richer in knowledge. Profiles will consist of a mixture of single facts that
raise “red flags”, conjunctions of facts that raise “red flags” if found together, and sequences
of data over time (behaviour patterns). A better understanding of fraud-related risk factors
could lead to better targeting of control visits and a significant increase in fraud prevention. It
may be that this area is considered more suitable for an ontology-supported demonstrator
than the task of fraud detection.
In what follows, we introduce a few of the requirements our analysis has discovered
so far. They are chosen here because they indicate needs for variously detailed axiomatic
descriptions of the concepts in question, and illustrate the range of theoretical choices and
commitments that have to be made. At present, these requirements and the ensuing
theoretical commitments are seen in isolation, as specific technological problems to be solved
by technicians. This paper then is a first attempt to see them interrelated through more basic,
theoretical or philosophical choices, amongst which the decisions between reference vs.
application ontology, description vs. revision ontology and “upward constructive”
(functionalist) and “downwards constructive” (formalist) theories of law figure prominently.
To anticipate somewhat: The multi-jurisdictional setting of the project means that we are
confronted with several mutually incompatible conceptualisations of the VAT domain. This
indicates the need for a “heavy”, axiomatic characterisation to minimise the potential for
misunderstandings between users from different jurisdictions. This then raises the question of
the ontological status of legal concepts. In one approach, legal concepts are ontologically
dependant on the social reality they regulate. This is a functionalist approach to law, and
there are well developed comparative law theories (Zweigert, Koetz)4 and jurisprudential
theories (Searle) that support this view. Decisions made on the modelling of evidence then
have a direct impact on the modelling of legal regulations.
However, the realistic assumptions underpinning this view are a two-edged sword. On
the one hand, they give a good explanation why legal comparison is possible: two concepts
from different jurisdictions are the same if they regulate (are dependent upon) the same bits
of social reality. On the other hand, this approach almost inevitably (and comparative law
research in this tradition bears witness to this) results in a revisionist ontology. In a medical
expert system, to get the right result as opposed to the result accepted by the medical
fraternity seems a desirable goal. In a legal expert system however, this might well contradict
a user requirement. To get the “right” answer is of little help if courts remain unconvinced –
and indeed it is questionable what the “right” answer could mean if social institutions with
the appropriate authority disagree with this result. As an alternative, there are equally well-
3
Certain parts of this common computer network can be accessed by businesses across Europe to check the
validity of their customers’ VAT identification number on the Internet since 14 June 2002. See
http://europa.eu.int/comm/taxation_customs/vies/en/vieshome.htm
4
Zweigert, Koetz: Einfuehrung in the Rechtsvergleichung. Muenchen 1986
developed “downwards constructive” jurisprudential (Teubner)5 and comparative legal
(Samuel)6 theories. In these approaches, legal systems and concepts create their own (social
and physical) reality. Decisions made about the correct representation of law then have a
direct bearing on the correct modelling of evidence and other non-legal entities. However,
this approach raises the question if two concepts of different jurisdictions can ever mean the
same, and what “sameness” could possible mean. According to some writers in this approach,
cross-legal correspondence of terms is merely conventional, not based on any “real” and
objective correspondence, and is ultimately policy driven. As a result, a lightweight
application ontology might after all be necessary to represent the ad hoc, merely conventional
and highly context sensitive correspondences between legal vocabularies of different
jurisdictions in a thesaurus. The legal philosophy of Reinach might provide a compromise
solution between these two that combines upward and downward construction. Legal
concepts exist independently and irreducibly from the physical world that they regulate, and
it is indeed a function of the law to “cut out” pieces (“Sachverhalte”) of this reality according
to criteria intrinsic to the law. However, ultimately non-legal objects form parts of these
Sachverhalte, which act as a bridge between law and social reality. As we will see, this
approach has its own problems especially regarding the question of description vs. revision.
2 Selected User Requirements And Resulting Demands On The Ontology
The ideal system for supporting VAT compliance would be a knowledge based system that
asks the user a few questions (in a choice of languages) and then supplies all the necessary
information and forms for VAT compliance. This is very similar to a system that is already
marketed by VAT Applications BV, a member of the FF POIROT consortium. If such a
system were to be supported by an ontology, the requirements on the ontology would be:
1. To represent legislation and legal rules from multiple countries;
2. To support reasoning about compliance with these legal rules;
3. To associate natural language terms in various languages with key concepts, so that
(for example) key rules from one country can be viewed in the language of another
country;
4. To represent interpretation of legislation and legal rules useful to the user.7
As stated above, the most important asset in fighting cross border VAT in the EU would be
an effective system of mutual assistance and information exchange in order to ensure the
proper functioning of the VAT system. A possible use case would be the automatic and
preferably spontaneous exchange of information to help in the detection of fraud in intra-
Community trade. The ontology should support this system, providing a solid basis for the
monitoring and enforcement of non-compliance of VAT laws. For this use case, the primary
role of the system would be to enable two-way co-operation within and across agencies and
within and across borders. This should include:
5. data integration within the same agency;
6. data integration between different (national) agencies;
7. data integration between two or more EU Member State agencies;
5
Günther Teubner: Law as autopoetic System. London: Blackwell 1992
6
Geoffrey Samuel: Comparative Law and Jurisprudence. International Comparative Law Quaterly 47
7
For example, HM C&E have refused to zero-rate supplies from companies which were complying with UK
VAT law (notice 703). See, Case Study “How HM Customs & Excise misinterprets EC & UK VAT law to the
detriment of British business trading in Europe”, http://www.bavat.co.uk/cases/ec.htm
8. knowledge about legal methods of fraud investigation (which will differ between
countries);
9. knowledge of typical indicators of fraud. Note that knowledge of methods of fraud is
not essential, only of the indicators that point to the use of those methods;
10. The system must deal with multiple data sources (VIES, ICT listing, etc), each
having different functions and user interfaces;
11. The system should know about law related to fraud investigation as well as fraud
identification. For example, Customs & Excise in the UK have power of criminal
investigation, but only the police have the power to investigate in Belgium.
12. The system should automate repetitive tasks such as checking to ensure VAT numbers
are valid.
VAT fraud prevention requires rather more wide-ranging knowledge than VAT fraud
detection, since it is searching for indicators of possible future fraud rather than indicators of
existing fraud. Its knowledge may therefore include past histories of individuals,
psychological profiles, typical company organizations for fraud, and typical commodities
traded in certain types of fraud.
13. The system should be capable of analysing indicators of possible fraud;
14. The system should be capable of link analysis, which relates known fraudsters to
other individuals using record linkage and social network methods. In the VAT fraud
domain, VAT investigators have found that fraudsters seldom work in isolation from
each other. In particular, VAT carousels are often the work of the same fraudsters
over and over again;
15. The system should (ideally) be capable of graphical/illustrative presentations of key
issues to assist users in understanding them, in navigating through them, and in
presenting them to a jury.
Since the goal of such a system would be to direct VAT inspectors to more likely cases of
fraud, there is also a need to represent national and supranational laws relating to VAT fraud
and national laws relating to investigation of VAT fraud.
Listed below are the top-level requirements of the ontology to address the demo showcases
on this project (VC indicates a compliance-related requirement, VD is related to fraud
detection, and VP to fraud prevention). It is not yet clear whether all these requirements can
be met with current technology, within the timescales of this project. However, in order to
support these requirements, the ontology must be able to describe concepts in the domains
listed below.
VC1: The law of various countries regarding offering VAT, and supranational laws
governing cross-border VAT transactions in the EU;
VC2: Knowledge of products, goods and transactions that is relevant to the various laws;
VD2: The law regarding investigation of VAT fraud in various countries
VD3: Different databases and their communication formats
VD4: The law of various countries regarding offering VAT, and supranational laws
governing cross-border VAT transactions in the EU;
VD5: An understanding of corporate structure (for tracking down missing trader frauds)
VP1: Knowledge of specific methods of fraud and related indicators
VP2: Knowledge of indicators of fraud relating to products and transactionsVP3: An
understanding of companies and corporate structure, including:
a) Management structure (partnership, etc);
b) Managers (the director might be someone with a criminal record);
c) Corporate structure (Does it have subsidiaries in other EU countries? Did head of
Company A found other companies in neighbouring countries?)
d) Does the company actually exist (or is it PO Box address only; answering machine,
etc)?
e) Who are its trading partners? Where are they located?
VP4: Knowledge of indicators of fraud relating to individuals and their history
VP5: Links between individuals and others via past or present relationships, or via types of
evidence
VP5.1 Relationships
VP5.2 Certain types of evidence
VP5.3 Evidence databases
VP6: The law of various countries regarding offering VAT, and supranational laws governing
cross-border VAT transactions in the EU.
VP7: The law regarding investigation of VAT fraud in various countries
Many of these top level requirements must be broken down into more specific requirements.
For example, requirement VP1 (Knowledge of specific methods of fraud and related
indicators) requires that the ontology must be able to represent the intention of actors. The
legal definitions of fraud require the presence of an element of intentionality on the part of
the individual. It’s quite possible to perform a crime unintentionally – for non-compliance,
this is normally due to ignorance of the correct procedures, while for fraud, it may occur if a
legal trader is used as an unwitting link in a VAT carousel fraud. This leads to the following
more detailed requirements:
Ontology requirement 1.2: The ontology must be capable of representing plans (modus
operandi).
• The ontology must be able to represent actions. This should include identifying the
actor(s); the resources created, modified, consumed or employed; and the resultant
changes to the state of the world. Thus key elements to be represented by the
ontology might include the fraudsters; their actions; the assets at risk; other assets
required; and the indicators that a fraud has taken place.
• The ontology must be able to represent sequences of actions. Sequences of events
reveal how related events unfold over a time period. The timing is important; for
example, when did the suspected fraudster establish the involved companies? When
did he ‘sell’ goods? When were these goods transported? When did he send invoices
to the authority? From a legal viewpoint, there are three time intervals that are
important: before, during and after the action (be it a fraud or the publishing of a
non-compliant website) was committed. Because the focus of this project is on
cross-border financial fraud, the location of the fraud and its aspects has to be
determined. This is important mainly for the assertion of jurisdiction by the relevant
investigative and prosecuting authorities. Questions to be asked are: Where is the
company incorporated? Where are the victims (tax authority, private investor)
situated? The ontology must be able to represent that some actions are ‘before’ or
‘after’ others. It’s not yet clear whether precise time intervals between actions need
to be represented.
• The ontology must be able to represent where the fraud happened (where harm is
done), fraudsters are located, etc. Initially, the level of detail need be only the name
of the nation(s) in which the individual resides/company is registered
Another example is the representation of laws regarding VAT, which is required for all three
possible scenarios (see requirements VC1, VD4 and VP6). This requires that:
Ontology requirement 1.3: The ontology must be able to represent legal rules:
• The ontology must be able to represent legal and administrative rules in different
countries across the EU;
• The ontology must be able to represent definitions of fraud, sanctions for fraud etc. in
different countries across the EU;
• The ontology must be able to support matching of plans against multiple legal rules
from a single jurisdiction.
This last requirement is important, for there are few laws that consider common fraud plans
as crimes directly; once such laws are made, the fraudsters simply change their plans. For
example, VAT carousels do not have their own statutory penalization; but they can be
prosecuted as an amalgam of criminal offences, such as specific fiscal offences and forgery.
Some of the other specific requirements that arise are listed below.
Ontology requirement 2.1: The ontology must be able to represent practical (‘on the
field’) knowledge about authorities involved in investigation of fraud, derived from legal
rules. For example: Who are the relevant investigating, prosecuting authorities? What is the
structure of the investigating department?
Ontology requirement 3.1: The ontology must be able to represent products as concepts
with multiple properties. The properties may be symbolic or numeric, single or multi-valued,
may contain links to other concepts or properties, and may be derivable from other properties.
Commercial transactions require a sale, a vendor and a purchaser. A sale is an agreement
between two legal entities for the supply of a product or of a service for a price.
Ontology requirement 4.1: The ontology must be capable of representing a Sale, an
action that transfers ownership of a Product from a Vendor to a Purchaser (who must
be a Legal Entity), and transfers a Sale Price from the Purchaser to the Vendor. A
knowledge of the structure of companies is also needed.
Ontology requirement 7.2: the ontology must be able to represent links between companies,
both parent/subsidiary links and trading links.
4 Some Theoretical Reflections
We have indicated the wide variety of concepts that need to be represented in order to
represent the workings of financial fraud. It can be seen that some requirements make high
demands on ontologies: to give two examples8, natural language "terms" cannot merely be
seen as annotations to "concepts", but need to be looked at as concepts themselves (in order
to satisfy the requirements to be able to detect fraud markers in the material provided by the
companies like annual reports etc); and also, the system must allow universals to be related to
particulars (e.g. "the Belgian VAT law" is a particular but a person violating it must be
8
with acknowledgements to Werner Ceusters of L and C Computing BV, from whom these examples are
quoted
represented as a universal). The choice of a general ontology on which the FF POIROT will
be based should therefore be made very carefully.
It is the cross-border, multi-language and multi-profession environment that makes
ontology-based solutions attractive. For an investigation conducted in the UK, the question
whether the director of the German supplier of a British company has relevant previous
convictions can be an important indicator of fraudulent activities. The relevant information
will be held in a German database, labelled in German and using a largely contingent format
of recording. An ontology-based approach should enable the investigator to give the system a
query in his language (“does the director of this company have previous convictions?”) and
get a reliable answer in English that is affirmative if the “Vorstand” of the “Gesellschaft” has
a “Vorstrafe”. The processes involved on successful fraud detection and prosecution
therefore involves
• To explain the meaning of a term to somebody outside the community - for instance
to explain to a UK lawyer that “Vorstrafe” covers (most of) “previous conviction”
• To negotiate meaning between agents belonging to different communities
(professional and national). This task is in our context partly fulfilled by the EU
directives that have exactly this function. However, the system needs to represent how
different (and possibly mutually inconsistent?) national laws “implement” one and the
same directive
• (ideally) to establish consensus - to accept e.g. that a German “Haftbefehl” is similar
enough to a UK “criminal charge” for the UK authorities to act on the German court
order to seize assets held in the UK as evidence.
These three functions, explanation, negotiation and consensus are listed in [Borgo et al]9 as
indicators that an explicit representation of ontological commitments is necessary to exclude
terminological and conceptual ambiguities. A rich axiomatisation can deliver that. In this
specific example for instance, it is possible to represent axiomatically the mereological
structure of companies in German and UK law10. The two concepts (or wholes) are by no
means identical, the German company having a necessary part (the “Aufsichtsrat”) that is
missing in UK law. The UK lawyer would however see that for each of the parts in a UK
company, there is a corresponding part on the German company and that in particular
“Vorstand” corresponds to “Director” not only regarding his legal function, but also in
respect of his place in the complex structure that makes a company.
Another of the ontology requirements introduced above makes this approach highly
desirable. The graphic representation of specific networks as evidence was introduced above
as one of the desiderata of the system. In this instance, it facilitates the communication of the
investigative expert, for instance a forensic accountant, with the prosecutor and possibly a
court or jury. However, Gestalt-psychology teaches us the intuitive compellingness of such
a graphical arrangements which makes it almost inevitable to see groups in graph layouts -
even if they are not substantial enough to establish say a “conspiracy to defraud” for legal
purposes.11 The historical link between Gestalt psychology and mereology developed by
Gurwitsch could allow us to capitalise on the axiomatic features of the links in the graph to
minimise misunderstanding. For instance, it might be desirable to represent only those links
9
Borgo, Gangemi, Guarino, Maselo, Oltramari: Wonderweb deliverable D15
10
Burkhard Schafer: Inheritance principles and the community of heirs. In N Guarino (ed.) Formal Ontology in
Information systems. IOS Press Amsterdam 1998
11
Sparrow, Malcolm K. (1991b). "The application of network analysis to criminal intelligence: An assessment
of the prospects." Social Networks 13:251-274Jensen, David (1997). "Prospective Assessment of AI
Technologies for Fraud Detection: A Case Study." AI Approaches to Fraud Detection and Risk Management,
Collected Papers from the 1997 Workshop. Technical Report WS-97-07. Menlo Park, CA: AAAI Press html
whose formal features guarantee that the ensuing whole is a “pregnant” whole.
While shared ontological commitment facilitates understanding, we encounter here a serious
difference between similar projects in the natural sciences or medicine, and the legal domain.
In the sciences, the underlying philosophical realism assumes convincingly that there is the
possibility to agree on the relevant classification. The one shared reality sees to this. In law,
the parliaments of the individual member states can create divergent realities. While German
and UK law will agree that an extended prison sentence handed out by a court is a “previous
conviction” and “Vorstrafe” respectively, it cannot be taken for granted that UK law
recognises the equivalent of a German legal term such as “Strafbefehl”12. If such a
“Strafbefehl” is recorded against the name of the company director, should the system return
an affirmative or a negative answer to our initial query? In an application ontology, this
decision is taken pragmatically based on the interests of a small group of users with clearly
defined goals and extracted from the texts they produce. For the investigator, the question is
if his suspect had previous problems with the law, not the precise legal classification of this
problem, and the classification of the domain will represent this. For the prosecutor, a much
more restricted reading is necessary that tells him that the “Strafbefehl” did not automatically
disqualify a person from becoming a director, the way an equivalent conviction would have.
The problem is moved to the lexicon. This however means that the ontology loses
“portability” and with that one of the main advantages of an ontology based approach. While
irrelevant for forming an initial suspicion, the legal difference becomes highly relevant when
the decision is taken to ask the German authorities officially for help in securing the relevant
evidence, and might be crucial once legal proceedings are started. The alternative is to re-
introduce realism and to treat the issue not as one of translation but of theory formation in
comparative law.
[Breuker et al] have developed for similar applications an approach that explicitly
incorporates philosophical commitments and comparative legal methodology.13 In their
“functional ontology”, they develop a hierarchical model of reality in which legal concepts
are made up of the elements of social or physical reality that they regulate. This functional
approach to legal concepts has great intuitive appeal and has formed the basis of mainstream
comparative analysis since the days of Zweigert and Koetz and more recently in economic
analysis and comparative law. In philosophical terms, Searle’s constructivism seems to come
very close to this approach. In this model, agreement between different legal
conceptualisations is always possible in principle by recourse to the shared reality that the
norms regulate. However, this theoretical commitment means that the criticism levelled
against these schools also applies to the resulting formal system. In comparative law, it has
been long recognised that the reductionism inherent in a functionalist approach can result in
serious misrepresentations of foreign law. In philosophy, Smith has shown how Searle’s
approach either faces insurmountable problems accounting for crucial legal features or
collapses into an unconvincing treatment of legal concepts as mere “facons de parler”.14
While we agree with Smith’s analysis of Searle, we think that some of the examples
he uses point to even more serious difficulties that ultimately affect all realist approaches to
legal ontology modelling. Of particular relevance to our setting is his treatment of a border
conflict between China and Russia: both jurisdictions, on the basis of their law, might claim a
specific piece of land. Searle has to treat these claims as valid even if they are not only
12
similar to a fixed fine, literary “punishment order”
13
Breuker, Elha, Petkov, Winkels: Ontologies for Legal InformationServing and knowledge management. In:
Bench Capon (ed.) Legal Knowledge and Information Systems IOS Amsterdam 2002
14
Barry Smith and John Searle, “The Construction of Social Reality: An Exchange”, American Journal of
Economics and Sociology, 60 (2001).
mutually exclusive, but also solvable if higher-ranking law (e.g. international law) is
employed. In our context, the presence of EU directives will sometimes solve similar disputes
- but not always. To complicate matters further, different jurisdictions might disagree on
exactly when EU law is competent. In legal doctrine, this is the issue of “competence
competence”, and an as yet unresolved, and on legal-conceptual grounds possible
irresolvable, question.15 Furthermore, and again in stark contrast to realist assumptions, it
might be possible that different national jurisdictions implement the same directive in a way
that is only “upward consistent”, that is, each national conceptualisation is consistent with the
EU blueprint, but they remain mutually inconsistent. To “solve” contested legal issues
through formal features of an ontology seems undesirable. Even though one could hope that
such an analysis contributes to the quality of the discussion in that field, to impose them on a
user who has to live with the relevant authorities seems overly ambitious. The requirement to
represent jurisdictions as physical spaces as mentioned above therefore clarifies an important
issue, but care must be taken that decisions on the correct modelling of space does not pre-
empt the underlying legal question.
Another interesting group of examples in our context that is also discussed by Smith
results from the requirement to represent temporal notions. Smith uses the examples of a
legal claim whose documentary evidence had been lost together with any memory of the
claim before it is “resurrected” by an heir centuries later. Another example he offers is the
treatment of property in the former RDG (“East Germany”) after reunification. For him, and
against Searle, this shows that legal objects can exist even if they are not recognised by
anybody. From a legal doctrinal point however, only his first example is unproblematic. It is
however at least possible on legal grounds to analyse the second case differently, as
introducing objects that exists intermittently in time, popping in and out of existence –The
status of such intermittent objects is highly contested in philosophy, although in law, and also
in our application, they are not that infrequent. Consequently, a realist ontology which
rejects them (on good philosophical grounds) would have to revise the ontology assumed by
the legal system.
Any realist ontology will on numerous occasions face similar problems. Legal
concepts do not evolve in isolation. They are influenced by (often misunderstood)
philosophical doctrines of the time when they were first recognised, and law in many ways
resembles a scrap heap of discarded philosophies. Fraud is in many respects particularly
fruitful (or difficult), because it typically involves an attempt to deceive someone about
certain attributes of an object which might or might not even exist, giving fertile grounds for
legal systems to come to their specific answers on how to treat properties of non-existing
objects, or to distinguish between attributes and essences. Certain distinctions in UK law on
deception as to the identity or attributes of a fraudster (e.g. the difference between claiming to
be Bill Gates, or to be just a very rich person) can’t be found in German law. To be sound
from the perspective of comparative law, our ontology must not pre-empt this question in
favour of one or the other.
As a radical alternative, we find “downward” constructivist theories in both
comparative law and legal theory. They still allow for a “heavy”, axiomatic approach to legal
ontology, and many formalist theories of law are actually much better suited for such an
approach than some of the more pragmatic functionalist solutions. They do however turn the
hierarchical conception of ontology upside down. The foundational ontology here is an
axiomatic description of the more basic concepts of the respective legal system; the rest of
15
Neil MacCormick: Beyond the Sovereign State. Modern Law Review 1993 p1-18, Bankowski, Schafer
Mistaken identities. In: Hoecke (ed) European Private law in context. Hart Oxford 2000
the universe of discourse is derived from these legal conceptual commitments. Law creates its
own realit(ies). An example in point is the legal treatment of electricity. Art some point in
time, legal systems had to decide whether electricity is an object (and hence can be stolen) or
a process (and hence something that can be fraudulently obtained). Different legal systems
differ in their answer to this issue. Downward constructivist approaches take this difference
seriously, and instead on deciding in the foundation ontology the status of electricity have
multiple representations of it, one for each legal system.
A consequence of this approach is an inflationary, multiplicative approach to
ontology, again borrowing from [Borgo et al]. Regardless of the question whether one adopt
this radical stance or not, a multiplicative ontology seems one way to solve most of the issues
discussed here, and in particular the question of jurisdiction. In those cases where the law
does not provide a clear-cut solution, to assume that different and mutually contradictory
legal entities can nonetheless have a co-localised existence seems a solution, which comes
very close to the self-understanding of the law. As fraud under German law, a specific
scenario might be punishable; as clever trading under UK law, it might not.
The problem with a downward constructivist approach is that it can’t explain how
legal system reach non-arbitrary agreement over meaning of terms, which in a context
dominated by EU legislation is the norm and not the exception. Their problems go even
further than that. One of the requirements from above is the need to represent intention in
criminal law. This intention again involves some understanding of the law. Individual
fraudsters might have understandings of the law which are very different from the “official”
reading, and if it is possible (as required by law) to say that nonetheless, their understanding
is “sufficiently similar” to the correct one to constitute legal intent, such an assertion seems to
rely on the same mechanisms this school considers impossible for cross-jurisdictional
analysis.
A possible compromise solution is a multiplicative reading of Adolf Reinach’s legal
ontology.16 It showed as early as 1921 how we can think of a realist model of comparative
law with jurisdiction-independent legal concepts, of which the national laws are but
instantiations. The vocabulary to describe these supra-jurisdictional concepts cannot any
longer be extracted from legal texts, but could be constructed a priori and axiomatically
described. The corresponding axiomatic descriptions of national law are then partial models
of these more general concepts, a relation that in turn can be axiomatically characterised. Law
largely retains its autonomy in this approach, and the (legal!) notion of “Sachverhalt”
becomes the crucial glue between legal concept and social reality.
However, this approach raises other philosophical and practical problems. In line with
Reinach’s philosophy, but in stark contrast in particular to the UK understanding of law, this
would mean that national courts could get their conceptual definitions “objectively” wrong.
While in a medical expert system, the correct answer is what the patient wants or needs even
if it differs from the view of a specific human expert, in a legal context predicting the court
might be more important than being “objectively” right. Since access to court decisions is a
user requirement for our application, this issue would still need resolving. Unlike the
functionalist approaches discussed above, “wrong” here only means internal conceptual
contradictions or incompleteness. This indicates the inherent problem of this approach. As
noted above, philosophy and law often co-evolve. Unlike in the cases mentioned above, here
philosophy borrows from law. The problem remains the proximity between one specific legal
tradition, here German law, and the resulting philosophy. Or put differently, a Reinachian
16
On the Cognition of States of Affairs", in K. Mulligan, ed., Speech Act and Sachverhalt: Reinach and the
Foundations of Realist Phenomenology, Dordrecht/Boston/Lancaster: Nijhoff, 1987, 189–225
ontology does not impose so much a philosophical perspective on the law, but German legal
concepts on the rest. To give a quick illustration, for Reinach it is a priori self-evident that
the legal notion of contract and delict are instantiations of the more abstract concept of
obligation. This is indeed an organisational principle of the German (or more generally
continental legal) Civil Code. The consequence of this is that contract law cases that clarify
the notion of obligation impact with necessity on the understanding of the law of delict.17 The
ontology as a result produces potentially revisionist arguments, but at least in the case of
German law, these revisions would be legally sound. Common law does not share this
understanding, and here legal concepts can be essentially underdetermined if a court does not
explicitly rule on them. The open issue remains if a multiplicative reading of Reinach can
avoid cases of “bad” revision. This in turns depends on the question how the ontologically
problematic “incomplete” concepts of the common law can be represented, that is concepts
that should apply to a given sachverhalt (by the standards of UK law), but in the absence of a
formal court decision are as yet lacking appropriate authorisation. Legal concepts in common
law countries, understood internally, display certain features that we also find in objects of a
very different kind, the fictitional objects of literature and film. “Sherlock Holmes” has
certain properties that are explicitly introduced by the Conan Doyle stories. He is male, he is
English, he smokes a pipe. However, while his author, as a human being, also has the
property of having a specific blood group (even if nobody ever knows which one), it does not
make sense to ask of his creation, Sherlock, which blood group he has until such time as this
information is explicitly introduced in one of the stories. Fictional characters are incomplete
at any given time, but gain more and more features as the story progresses. The only
constraint is that new attributes must be consistent with attributes already introduced. To treat
(in a common law context) legal objects as fictitional objects would account for instance of
Dworkin’s intuition of law as a chain novel. An ontology that can model both actual and
fictitional objects, along the lines e.g. of Edward Zalta’s axiomatic metaphysics, could
encompass both common law and civil law ideas about the status of legal concepts, but would
result in a potentially undesirable inflation of the ontology. An alternative approach could be
based on an analogy between legal objects and other “intended future” objects. When I order
for instance a customised car, only certain of its properties are determined when I conclude
the contract, some through explicit agreement, others by certain coherence requirements
resulting from these specifications. To treat it because of this as a fictitional object would be
inappropriate. This sort of “intended future objects” might be a better candidate for a
successful analogy than the fictitional characters of literature.18
5 Conclusion
We have shown that describing a real application of law requires a detailed ontological
description not only of laws themselves, but also of various related concepts. We have
specifically discussed the laws themselves, and have shown that representing laws from
multiple jurisdictions is possible, at least within the context of legal harmonisation within
Europe, by generating a generic description of legal principles which are then instantiated
into the legal codes of individual jurisdictions. However, we have also shown that such an
ontology will always have to balance two conflicting demands, something that we expect to
17
B. Schäfer: Reinach, pragmatic universals and the methodology of comparative law. in: K. Dorn (ed.)
Pragmatics in Contemporary Philosophy. Beiträge der Ludwig Wittgenstein Gesellschaft, Wien: LWS
1997.p.850-856
18
based on a very helpful observation by Barry Smith
be typical for domains with social objects. The “heavier” an ontology, and the more
pronounced its realist bias, the more revisionist it is going to be. Since the legal systems of
different countries often embody for historical reasons ontological assumptions taken from
“bad” philosophies, it might not be possible to give an account of certain of their legal
features that remain true the self-understanding of these legal systems. The benefits of heavy
ontologies have to be balanced against the danger to impose the perspective of one’s own
legal system on a foreign law that might have radically different structural features. Since the
process of legal harmonisation within the EU can itself be seen as an exercise in revisionist
ontology modelling (i.e. the attempt to chose the “right” conceptual framework for a
harmonised law) POIROT should be able to capitalise on the potential of heavy, axiomatic
approaches to ontology modelling, but already extending it to US-EU tax fraud might require
a radically different approach.