Martin Mose Bentzen abstract.

Applying for a Residence Permit in Denmark – A logical analysis of Legal Text and Legal Practise.

Paragraph 9 of Danish Immigration Law deals with family reunion (Danish: familiesammenføring), i.e. obtaining a residence permit to Denmark on the basis of being in a relationship to a sponsor residing in Denmark. Danish Immigration Law has been a controversial political issue, since a new immigration law was introduced in Denmark in 2001. The administrative practise is not fully open to the public, and there is a great deal of frustration among applicants, when they apply for residence permits.

In this paper I consider the explicit structure (e.g. cross-references, conditions and exceptions) of the law and consider the role of legal practise in determining underspecified terms of the law. The logic used is a non-monotonic first-order logic. An automated version of the formalization of paragraph 9 is implemented in Prolog.

The result of this work is to be a web site where the program resulting from the formalization can be downloaded for free, to give applicants an idea about the demands put upon them, when applying for a residence permit. In a wider context, systems such as this could be used by immigration lawyers, immigration officials and maybe even legislators trying to evaluate and fixate possible interpretations of proposed laws.

A legal text has a certain internal structure, it is divided into clauses and sub-clauses referring to and depending upon each other in various ways. An example of this is cross-references, where one clause describes an exception to another or clauses are used as conditions for other clauses.

In the paper I describe a primitive naming mechanism, depending on certain predicates having been decided as having primary interest and made relative by adding arguments. An example of such a predicate could be getting a residence permit. Thus, for instance the binary predicate Permit(x, clause 1), would mean that x can get a residence permit according to clause 1. Changing a bit the tools of a recent MA thesis by Marcello di Bello, an exception to a rule can be written Exc(x, clause 1, clause 2) ¬ R(x). Here clause 1 is the clause x is exempt from, clause 2 is the name of the exception and R specifies some condition. This could for instance be x is exempt from clause 1 by clause 2, if x is a criminal. Clause 1 could be written Permit(x, clause 1)¬ S(x)Ù not($y(exc(x,clause1, y))). This means that x can get a residence permit according to clause 1, if x has a sponsor and x is not excempt from clause 1 by some (other) clause, y. In the paper I consider the semantics of these constructions. The not predicate is negation as failure from the paradigm of logic programming. not(P), means that P cannot be proven. With the clauses above and assuming that Ann has a sponsor but that she is a criminal, Ann cannot not get a residence permit according to clause 1, because she is exempt from it by clause 2.

When interpreting a legal text, e.g. to see if a certain right applies to a specific individual, certain underspecified terms in the text must be fixed. This fixation is partly established in the administrative practice and by court decisions in cases of controversy. In effect, legal practise change the rules. In this paper I also investigate the interplay between the static intertextual structure of the Danish Immigraion Law and dynamic legal practise.

As sources for legal practise I use notes from the Minestry of Integration, legal textbooks and articles written by legal scholars. Futhermore, my plan for the near future is to test the system on various types of potential users of the system, applicants, legal experts, immigration lawyers and immigration officers. This feed-back will undoubtedly enhance the systems usability and correct certain errors.

Future development includes adding legal meta-principles and tools from machine learning to deal with e.g. court decisions in more sophisticated ways, than is now the case (where they are treated as new rules). The non-trivial questions involved will be discussed in he final part of the paper.

Note:

A prototypical version of the Prolog program Residence Permit is ready at the time of the workshop. It is an expert system asking for information about the applicant and the sponsor living in Denmark. The basis of this asking mechanism is the long application form provided by the Danish Immigration Service, which must be filled out by the applicant and the sponsor. This is the primary source of information for the Danish immigration authorities when making a decision on whether to give a residence permit. When all the information is obtained the program runs it through the formalization of the law and the legal practises. It makes a decision. Finally, it gives an explanation of which rules and facts are used to meet the demands of the law ( a how explanation in AI terms, see e.g. (Bratko, 2001)).

References:

Marcello Di Bello (2007) Formalizing Legislation in the Event Calculus: The Case of the Italian Citizenship Law., ILLC Publications, MoL- 2007-07.

Ivan Bratko (2001): Prolog Programming for Artificial Intelligence, Addison-Wesley.