 |
 |
E-GOVERNMENT AND AI*
Tibor Vámos
The task of e-government involves the application of artificial intelligence instruments. The system should understand the applications of the citizens. This means a genuine natural language understanding system, first written later spoken texts. This can be supported by special vocabularies and a simplified grammar. The filtered and analyzed texts are matched with the usual scenarios of the selected procedural practice. The scenarios are the bases of case-based reasoning process, a decision support system. The system is an experimental one especially for support of the user citizens in an advisory role and helps the decision-makers. The project was received well by the officials concerned. Two earlier experiments were feasibility studies for such kind of systems. This experiment includes some sociological-psychological analysis of the applications. All actions are carefully executed according to the rather strict privacy conditions of law.
1. Introduction: Two faces of e-government E-government has two different faces. The first is the self-organization of a government or any kind of government organization, such as local government, city government, a government agency, etc. The other is the democratic face of government, involving communication with the population that the given government represents. The two should have both interfaces for contact and firewalls separating them. The latter protects the right to privacy of the citizens, as well as certain well-controlled confidential issues within the government. In a democracy government confidentiality should be regulated by law and be authorized by impartial and independent representatives of the citizens. Our interest here is focused on this second face. There are three objectives: first, to ensure a human-oriented administration in an information technology environment; second, to strengthen democracy in governance; third, to increase the level of professionalism in administrative work. These three objectives really amount to one goal: a step towards a more human-oriented future supported by technology. This would go against the current negative trend involving the mechanization of human relations [8-23]. The self-organization of a government is not too different from the self-organization of a large company involved in a variety of businesses. The task is in some respects easier since administrative procedures usually already have well-developed legal foundations that create a natural logical frame for the organization, its agents and their activities. We can argue about whether the resistance against innovation is stronger here or there. That said, any large commercial information system software that has stood the test of time can be applied to any of the organizations concerned. An excellent survey of general organizational functions and their realization in different countries is given in [7].
2. Particularity of the democratic dialog The second face, the democratic dialogue is more complex, newer and more particular with respect to the social environment of the community. It basically amounts to a linguistic problem; we generally understand language as a means of communication among persons, man and machine or as a means for the individual to express his/her thought. The linguistic representation in this sense is comparable to the concept of understanding [24, 25]. The individual, in the process of understanding has to formulate a linguistic representation of his/her ideas or impressions of the situation at hand. That linguistic representation has to be understood through a matching of the representation by the other partner. The dialog would be a continuation of this process and has to work both ways. The procedure is not symmetric; we have to work with three languages within the same dialogue. First, the natural language of the citizen, individual in every case, second, the uniform, logic-based formal language of the administration and third, the half-formal, half-natural language used by the administration in reporting to the citizen and controlling the understanding of the citizen's claim. Understanding is the major step towards decision-support, it allows for the partner in the dialogue to start a search for a solution to a given problem. The particularity of the task necessitates more experimental pilot projects than is usual in information management. The major experiment reported here is a key part of a national plan.
3. The experiment: a regional administration The experiment is being carried out in a region in the southwestern part of the country, around the city of Kaposvár and it includes several rural areas. The center itself with about 70.000 inhabitants is a thriving community: Currently they are developing a new university; and they have a pioneering theatre that has been a springboard for the careers of most of the leading contemporary art directors and actors in the country. The region's rural area with about 50.000 more people is rather diverse, some of the villages existence is closely linked to the city, while others are more remote in every sense of the word. The experiment has five partners, the Kaposvár City Hall, the Budapest University of Economics and Public Administration, the Computer and Automation Institute of the Hungarian Academy of Sciences and the local communication provider, Hungarian Telecom [25]. An obvious start for the project was a sociological review. The aim was to explore the receptivity of the population, i.e. the level of computer literacy, the general educational and comprehension level, the attitudes of the citizens and administrators towards computers and networks. The results concerning attitudes, were better than expected; the physical and intellectual access level was not too different from the country's average, i.e. a little higher than the average for the countries joining the EU next year. Special attention was devoted to family and community relations among citizens, especially in the countryside, with respect to the possible support of those who, in the next few years, cannot be expected to acquire either home-access or the necessary skills. The prospect for these issues was also satisfactory, i.e. some access and support can be expected within the next few years. The Sociology Department of the University, with the active participation of its students, carried out the sociological review. This review will be repeated later on during the project. The attitude of the administration was excellent. They understood the objectives of the experiment, were very cooperative, helpful and had no anxiety about losing their jobs. They were clear on our explanation about the nature of the project, i.e. increasing the level of their duties, giving more attention to really deep human problems and lessening boring routines. We received many applications, most of them handwritten by everyday citizens, in order to analyze the linguistic problems. The applications were carefully processed in keeping with existing privacy protection laws. A sociolinguistic analysis is part of this experiment. Several stylistic features in the letters of the citizens should give insight into the social environment of the problem, in a way substituting, the personal contact. A group of psychologists are doing the same research with respect to the emotional contents and their relevancy in judging a case. We would like to emphasize here the research characteristic of this part of the project. All other parts, concerning the first face, i.e. the construction of a well-organized regional administration, its information flow and data management, the design of a many-faced, interactive portal, the solution to legal issues, or the layout of access points, are not discussed here. The objective now is more to create a support to provide the citizen with advice rather than to automatically make any decisions. It can help the citizen in the formulation of his/her application, in getting an overview into the rules governing the decision on the case, and in suggesting possible outcomes of any given procedure. In the other part of the project, the jobs that do not require human judgment are to be settled. This means, first of all, the issuing of documents or permits that have strictly defined rules and conditions, or other similar tasks. According to the law and the aquis of the EU no final decision on a human problem can be made by machines. One might ask, why do we not simply rely on well-formulated questionnaires. Their use is well-developed in the administration of the region and should not be abolished. However, to give the governing bodies a more human face and to get a deeper analysis of the requests of the citizens we need to go further with machine-supported administration. Questionnaires that do not leave much room for individual requests and/or comments are already quite common; and they lead the client down a prefabricated and, oftentimes, both inconvenient and inadequate route. The main problem is the translation from one language to the next. The understanding of natural language has been a focus for research in the field of Artificial Intelligence ever since its inception [2]. The ultimate goal is to create a system that can perform as well as an educated human interpreter. Educated here means having a thorough knowledge of the language and similar level of competence in all possible subjects of discourse. The latter means a tacit knowledge, as Polanyi put it, concerning the meaning of the words, their relations to the world of the subjects. This is the point where a pragmatic compromise can be made: restricting the understanding to a specific subject, to the vocabulary and terms of that subject and to the usual scenarios related to a discourse on the subject. In our case this very solution was applicable, all possible subjects of discourse are generally limited to certain well-formulated scenarios.
4. How does it work in a limited subject area? Let us take, for example, an application for financial assistance. It must contain words concerning financial assistance, help, support. The grounds are also limited, e.g. illness, accident, some damage to the home or cost of education. The possible circumstances, such as family relations or income, are also limited. We could easily collect vocabularies needed for different cases selected according to the various types of problems and scenarios for situations and procedures. The problem areas are automatically identified based on the frequency of different characteristic words. Grammatical analysis is used, if the words themselves do not clarify the situation. The Hungarian language is an agglutinating system, i.e. considerably different from the major European language families. Nevertheless, the basic Chomskyan characteristics are similar [3, 4]. A one-million word corpus, created by the Kalmár-laboratory in Szeged and the Morphologic company's HUMORESK linguistic analyzer support this task-oriented system [5, 6]. Matching procedures are active throughout the text analysis. They try to find the referent scenarios from almost the very start of the word processing. The scenarios are graphic representations of the administrative queries and procedures in the formal language of laws and rules. This creates an obvious bridge to the decision-support features of the system, the concluding decisions are also in the form of scenarios very similar to the query structures. The two scenario graphs are linked by logical functions. Logic, if needed, can be combined with uncertainty estimates. Weighting situations, severity and reliability of claims is mostly a verbal expression that is easiest to represent using fuzzy values. Handling complex fuzzy data, coined signatures is supported by other results of our group, including data mining in large databases with long range experience [26]. The system works in a dialogue with the client affirming the understanding of the application and the computer presenting the final or intermediate conclusion of the decision-support. The language of this dialogue, generated by the computer, is an intermediate construct between the formal logic-based language of the administration and an informal-looking language that can be user-friendly in a psychological sense. The use of canned phrases can give an informal feel to the official formalism. The method of affirmation dialog is not too much different from the nearly four decades old ELIZA concept [29]. Earlier experiments dealing with child custody problems after divorce and bankruptcy law have demonstrated the feasibility of a system capable of understanding natural language for limited subjects and of providing decision-support based on that understanding [27, 1, 28]. The first two tasks in the local administration project are the management of requests for social support and the handling of claims against other citizens and organizations, i.e. mostly information against somebody or because of something. These are the two most subjective operations, i.e. the greatest challenge in the project and the most informative tasks in terms of the practical application of artificial intelligence instruments.
5. Conclusion Two major instruments of artificial intelligence, natural language understanding and decision support can be helpful in creating a more human face of e-government. In contrary to the general beliefs e-government can be more human oriented than traditional governing methods, enabling the citizen to encounter a 24 hours, home resident dialog support and providing a transparent management of the individual problems. The experiment being in progress proved the feasibility of these goals using some obvious limitations and selecting items to be considered by special human attention. That means an elevation of the administration work level focusing on problems requiring more professional knowledge and more empathetic attitudes.
6. References [1] ELHADI, M.T., Bankruptcy support system: taking advantage of information retrieval and case-based reasoning. Expert Systems with Applications, 18, pp. 215-219, 2000. [2] Iwanska, L.M., S. C. Shapiro (eds.), Natural Language Prcessing and Knowledge Representation, Language for Knowledge and Knowledge for Language, AAAI Press-The MIT Press, Menlo Park, CA, Cambridge, MA, 2000. [3] KISS, É. K., The order and scope of operators in the Hungarian sentence. In: Topic, Focus and Configurationality (eds. W. Abraham, S.de Meij), John Benjamins, Amsterdam, 1986. [4] KISS, É. K., Discourse Configurationality in Languages of Europe. In: Constituant Order in the Languages in Europe (ed. A. Siwierska), A. Mouton de Gruyter, Berlin, pp. 681-729, 1988. [5] PRÓSZÉKY, G., Hungarian ? A special challenge to machine translation? In: New Directions in Machine Translation (eds. Maxwell, Schubert & Witkam), Dordrecht, Foris, pp. 212-231, 1988. [6] PRÓSZÉKY, G., B. KIS, A new approach to unification-based morpho-syntactic parsing. In: Proc. 6th Conference on Applied Natural Language Processing (ANPL´99), pp. 261-268, Maryland, USA, 1999. [7] RONAGHAN, Stephen A.., Benchmarking E-government: A Global Perspective (Assessing the Progress of the UN Member States). UNDPEPA- ASPA, New York, 2002. [8] VÁMOS, T., T. B. SHERIDAN, S. AIDA, Adapting automation to man, culture and society. Prepr. 8th IFAC World Congress, Kyoto, Japan, 1981, pp. PS-1-7 [9] VÁMOS, T., Menschliche und künstliche Intelligenz. In: Mikroelektronik für Mensch, Linz, September 11-13, 1984, Vorschungsinstitut für Mikroprozessortechnik der Johannes Kepler Universität, Linz, Austria, 2, pp. 803-816 [10] VÁMOS, T., Convergences of cooperative control, decision and expert systems. Proc. "25th CDC" - IEEE Conference on Decision and control, Athens, Greece, December 1986, 1, pp. 616-621 [11] VÁMOS, T., Future and impacts of AI - position paper. Task Force Meeting, Vienna/Austria, August 21-23, 1987, International Federation for Systems Research (ed. R. Trappl), Nr. 272 [12] VÁMOS, T., New technology--new society. In: Man and Automation (ed. P. Kopacek)--Papers given at the Fred Margulies Memorial Workshop, Laxenburg, Austria, October 2, 1987, pp. 19-21 [13] VÁMOS, T., AI as a vehicle for a more advanced democracy. Conference on Intelligence and Society, 1988, Vienna, Austria [14] VÁMOS, T., Expert systems in negotiations--oracle or useful support? Meeting on SWIIS, Vienna, Austria, 1988; Cybernetics and Systems: an International Journal, 20, 1989, pp. 113-118 [15] VÁMOS, T., Public service ethics in the computer age. 1st International Conference on Public Service Ethics, June 3-8, 1990, Jerusalem, Israel [16] VÁMOS, T., AI environment for supporting citizens. Proc. IFIP WG 8.3 Working Conference on Environments for Supporting Decision Processes (eds. H. G. Sol, J. Vecsenyi), Elsevier-North Holland, 1991, pp. 37-46 [17] VÁMOS, T., Cooperative communication: computerware and humanware. Journal of Organizational Computing, 1 (1991), pp. 115-123 (Presented at the Conference on Organizational Information Systems and Coordination, Austin, Texas, USA, November 13-14, 1989) [18] VÁMOS, T., Rechenwissenschaft und Demokratie. In: Computer, Kultur, Geschichte - Beiträge zur Philosophie des Informationszeitalters (eds. D. Mersch, J. C. Nyíri), Wien: Passagen Verlag, 1991, pp. 15-20 [19] VÁMOS, T., Artificial intelligence: human focus on technology. Applied Artificial Intelligence, 5.1991, pp. 163-170 [20] VÁMOS, T., Anxieties around AI--philosophical and practical, educational response. EMCSR'96, April 9-12, 1996, Vienna, Austria [21] VÁMOS, T., Does ethics change? 4th International Conference on Ethics in the Information Age, Jerusalem, Israel, June 13-16, 1999. [22] VÁMOS, T., Ethics in information age--persistence and change. In: Ethics in the Public Service-- International Conference, June 25-30, 1995 (eds. U. Berlinsky, A. Friedberg, S. P. Werner/, Erez Publishing, 1999, pp. 347-352 [23] VÁMOS, T., Computer Epistemology, World Scientific, Singapore, 1991. [24] VÁMOS, T., Intelligent machines? BOOKS, 9, No. 3, 1999, pp. 138-144 [25] VÁMOS et al., E-government--a service. In: "2nd European Conference on e-Government", St. Catherine's College, Oxford, October 1-2., 2002 (ed. Dan Remenyi), pp. 453-462 [26] VÁMOS, T., Gy. BÍRÓ, L. KÓCZY, Fuzzy signatures in data mining. Joint 9th IFSA World Congress and 20th NAFIPS International Conference, July 25-28, 2001, Vancouver, Canada [27] VÁMOS, T., P. SOMOGYI, P. DANYI, F. KATONA, Lessons of the pattern view of knowledge. In: Cybernetics and Systems Research' 92 (ed. R. Trappl), 2, pp. 1609-1616, 1992. [28] VÁMOS, T., M. T. ELHADI, Bankruptcy law research: a CBR approach. 7th International Conference on Artificial Intelligence and Law (ICAIL-99), June 14-18, 1999. [29] Weizenbaum, J., ELIZA?A computer program for the study of natural language communication between man and machine. CACM, 9, 1966, pp. 26-45
|
|
|