Home A short resumè Publications		Curriculum vitae: Stefania Costantini
Born 5 Oct 1959. Married, two children: Daniele (1993) Alice (1996)		Research Interests Stefania's research interests are in the field of Computational Logic, Logic Programming and its applications to Artificial Intelligence. Metalogic Programming Together with Elio Lanzarone, Stefania has defined Reflective Prolog, a metalogic programming language which extends the Horn clause language with a naming mechanism, metaevaluation clauses and a form of logical reflection. Reflective Prolog has been implemented at the Logic Programming Lab of the Computer Science Department of the University of Milano. The language has been experimented on several complex problems of significant application domains in Knowledge Representation and AI, such as plausible reasoning, case-based reasoning, legal reasoning and temporal reasoning. The language has been augmented with negation and agents, and simple forms of induction. The main concepts underlying Reflective Prolog have been further elaborated,  to design (together with Jonas Barklund and Pier Dell'Acqua) a self-referential, reflective logical system whose main objective is to allow its users to specify and experiment a variety of deductive systems, defined via axioms and rules of inference. Stefania has been recently invited to contribute to a book in honour of Prof. Robert Kowalski, title of contribution: "Meta-Reasoning: a Survey". Recently, Stefania has exploited her work and experience about meta-reasoning in her  research on logical agents, discussed below. Some important features of the DALI language that Stefania has proposed have been defined as forms of introspection. Reflection mechanisms taken from Reflective Prolog have been very useful for the agents to deal with ontologies, and with uncertain and incomplete knowledge. Logic Programming with Negation, and Non-Monotonic Reasoning One of Stefania's main interests is negation in logic programming. Stefania has introduced the first explicit characterizations of stable models in the context of logic programming (the existing ones where represented in other formal systems, such as TMS). She has shown that these characterizations are widely applicable, and are, at least in principle, efficiently computable. Stefania is presently studying the relationship between the syntax of programs and the existence of stable models. In particular, by representing the program as a graph (Extended Dependency Graph, EDG), the stable models can be characterized as Admissible Colorings of the EDG (joint work with Ottavio D'Antona and Alessandro Provetti). In December 1999, Stefania has given a seminar on this topic at the University of Texas at El Paso, invited by Prof. Micheal Gelfond.  She has also shown that, as far as consistency checking or more generally program analysis are concerned, it is possible to work with compact forms of programs, such as the kernel normal form. In a joint work with Alessandro provetti an algorithm is proposed for shifting back and forth from the given logic program to its kernel form. Obtaining the kernel has low polynomial time complexity, while shifting back to the original form has linear complexity. This means, the graph of the kernel might be displayed on the screen, and edges representing a longer path in the original program could be emphasized, so as the original program might be obtained upon request (for instance by clicking on those edges). With the aim of defining sofware engineering methodologies for ASP, she has shown that representations like the EDG, or others that have been proposed in the literature, which are oriented to atoms and rules can be usefully  condensed  into more compact representations oriented to components, like the Cycle Graph that she has defined. In the Cycle Graph, vertices are not atoms or rules, but significant subprograms. She has proposed a formalization of the features that graph representations of logic programs should  exhibit in order to be a satisfactory and useful representation formalism. She has  introduced a concept of isomorphism between a program and the corresponding graph. Isomorphic graph representations guarantee that the graph corresponding to a program is unique, and that the original program can be reconstructed from the graph. She has proved that the isomorphism is possible only if a graph representation formalism is able to distinguish the cycles occurring in the program, and the different connections among them. She has argued that understanding cycles and their connections is a key point for understanding program behavior and for checking consistency, for being then able to create, debug and combine good programs, and for developing program analysis techniques. useful classes of consistent programs can be easily defined. Based on this concept, she has identified, for the first time in the literature, a class of programs which are not call-consistent, and admit answer sets by construction, and are such that an automated analysis tool might easily verify whether a program belongs to this class, and that program fragments of this kind might act as building blocks for safely constructing larger ones. In recent work, Stefania aims at demonstrating that the programming paradigm stemmed from stable model semantics, called Answer Set Programming (ASP), is a suitable paradigm for defining and implementing data integration systems. In particular, she has defined (with Andrea Formisano and Eugenio Omodeo)  a formalization of the Global-as-View approach, with the objective of dealing with forms of incompleteness in data correspondence. She has presented a working inference engine for effectively answering queries in presence of uncertain/ambiguous/imcomplete knowledge, where each answer set corresponds to one of the possible consistent answers. She has recently proposed with Andrea Formisano and approach to Answer Set Programming with resources (RASP) and also with preferences which, in the spirit of linear logic, deals with resources, quantities and remainders. RASP provides forms of preferences that can be local to rules, to cope with the fact that one may have different preferences in different contexts. Moreover, il allows for non-linear preferences, i.e. manages partial orders. A merit of the approach is that, even though the proposed extensions are not merely syntactic sugar, the computational complexity is the same as plain ASP. A prototype implementation is publicly available at the web site http://www.dmi.unipg.it/~formis/raspberry/. Stefania has been also interested in the past in disjunctive logic programming. Together with Elio Lanzarone, she has pointed out how the complexity of the semantics of this class of programs can be coped with by splitting the definition of the semantics into two parts:  a program transformation phase, of high complexity, and a constructive phase, of reasonable complexity. As a case study, they have considered Przymusinski's Static semantics: they have formally defined the two phases, and  have extimated the complexity. Computationally, for disjunctive logic programs this would result in a procedural semantics with an inefficient preprocessing step, to perform in the first place, and a reasonably efficient inference procedure. Stefania and Tiziana Morbidoni have formally assessed the computational complexity of this approach (which is related to the technique of knowledge compilation). Logical Agents Recently, Stefania has been working in the field of logical agents and multi-agent systems. She has proposed a new logic programming language, called DALI, with active and reactive rules, close in syntax and semantics to the traditional Horn Clause language. The language has been implemented (in co-operation with Arianna Tocchio) at the Computer Science Dept. of the University of L'Aquila and is being experimented. The DALI interpreter with documentation is publicly available at the web site Web-DALI. Syntactically, DALI is close to the Horn clause language and to Prolog. DALI programs, i.e., DALI agents, may however contain a new kind of rules, reactive rules, aimed at: (i) interacting with an external environment; (ii) reacting to some of  its own internal conclusions. The environment is perceived in the form of external events, that can be exogenous events, observations, or messages from other agents. In response, a DALI agent can either make actions or send messages, or perform some internal reasoning. What is new in DALI is that the same external event can be considered under different points of view: the event is first perceived, and the agent may reason about this perception; then a reaction can take place, and finally the event and the (possible) actions that have been performed are recorded as past events and past actions. An important novel feature is that internal conclusions can be seen as events: this means, a DALI agent can "think" about some topic, the conclusions she takes can determine a behavior, and, finally, she is able to remember the conclusion, and what she did in reaction. Distinguished internal conditions that can trigger reactions are automatically attempted from time to time. Whatever the agent remembers is kept or "forgotten" according to suitable conditions (that can be set by directives). Then, a DALI agent is not a purely reactive agent based on condition-action rules: rather, it is a reactive, proactive and rational agent that performs inference within an evolving context. DALI has a fully logical declarative and procedural semantics, and an operational semantics. The DALI language might constitute a starting point towards the specification of rule-based agents for the semantic web: DALI in fact provides reactive and proactive features combined with reasoning mechanisms, within an evolving context that can be exploited for achieving agents with "ambient intelligence". DALI agents have been experimented in the definition of a prototype Intrusion Detection System. In the context of the CUSPIS project, DALI has been used to implement a multi-agent system that assists the user during her visit to cultural assets (where the visit is monitored by means of GALILEO satellites) by eliciting the user profile and thus providing the user with personalized information and suggestions for the persent as well as for future visits. A satellite-based authentication system has also been developed (with Arianna Tocchio and Pierangelo Dell'Acqua) for monitoring the cultural assets transfers. Recently, with Pierangelo Dell'Acqua Stefania is explorating the definition of local preferences in logical agent languages. With Francesca Toni, Stefania has formally defined an abstract multi-layered general agent model including a meta-control layer. Together with Pierangelo Dell’Acqua, Luis Moniz Pereira and Francesca Toni, Stefania is devising an approach where assistant agents interact with users and support their activities. Agents cooperate with humans so as to help them to adapt to environments that are new to them and/or when their ability to cope withthe environment is too costly, non-existent or impaired. An assistant agent should improve in time, both in its understanding of the user needs, cultural level, preferred kinds of explanations, etc. and in its ability to cope with the environment. The approach constitutes a step towards the SocialComputer vision (http://socialcomputer.eu/), where, in order to address the grand challenges of the coming decade, both large scale and small scale, an infrastructure is envisaged that benefits from the growing power of ICT but that also harnesses these growing possibilities in a more socially acceptable way, providing a good balance between machine and human social intelligence. Also, they are developing an approach to self-checking in the form of a method for allowing agents to explicitly observe and record their past behavior so as to be able, by means of specific constraints defined as temporal-logic-like formulae, to decide the best actions to do, and to avoid errors performed in previous similar situations. For re-conciliating scalability and intelligence, Stefania would like to go towards creating systems where various degrees of intelligence are distributed over various levels of the architecture where high-level agents are responsible of overall system strategies and plans, to be possibly devised in cooperation. Bunches of elementary agents are instead supposed to assist each high-level agent in activities where massive parallelism is in order, such as environment exploration, pattern-recognition, classification, action selection and action execution. Work is under way with the former Ph.D. student Panagiota Tsintza, now lecturer at the University of Ionia (Greece) about automated negotiation. We have extended a constraint-based algorithm for P2P agent negotiation proposed by Marco Cadoli in 2003. The extensions on the one hand overcome some limits of the original approach thus allowing am agreement to be found where previously it was not possible. On the other hand, the extension allows more generality in defining the negotiation area of each agent and in computing offers. Our recent research work on negotiation is related to Argumentation-Based negotiation and its use to cope with contract violations: an agent that has violated an already-signed contract will try to justify this fact by exposing some arguments while the opponent agent will try to undermine their truthfulness and acceptability, by finding attacks against them. As a response, the justifying agent needs in turn to devise a counter-attack. Natural Language Processing In cooperation with the Ph.D. student Alessio Paolucci, work is under way about aspects of Natural Language Processing (NLP) in Computational Logic. A working prototype has been developed, called Mnemosine, which by relying upon an improved form of DCG’s, the SE-DCG’s and upon a background knowledge base is able to answer questions in natural language in a comparatively efficient and reliable way. In order to automate the process of knowledge acquisition from natural language source, they build upon recent work where natural language sentences are translated into ASP, taking into account sentences that imply uncertain knowledge and thus cannot be translated into classical logic. They mean to exploit meta-reasoning and default reasoning so as to decide whether and in which form knowledge extracted by a sentence should be added to a knowledge base. Automated Deduction In cooperation with Eugenio Omodeo and Pasquale Caianiello, Stefania Costantini has contributed to the specification of definitional extension mechanisms within a purely equational framework such as the "Schroder-Tarski"  calculus of dyadic relations  The prototype implemented Metamorpho system, written in SWI-Prolog, is available at the URL http://costantini.dm.univaq.it/online.htm The motivation is that  such a formalism can play a fundamental role in a two-level architecture of logic-based systems. Three basic forms of definitional extensions are supported: (i) introduction of new term constructors; (ii) disguisement of special equations under new sentence constructors; (iii) templates for parametric lists of sentences that will be actualized in the formation of axiomatic theories. Supplied features which make the above mechanisms really flexible are the ones enabling the user to: (a) introduce variadic constructs, to wit, constructs with an unrestricted number of arguments; (b) exploit, in the definienda, parameters which do not belong to the formalism: typically, natural numbers, lists, and even quantified first-order formulas; (c) make some definitions local to specific theories; (d) make some templates for parametric theories 'ephemeral', so that they are available throughout the loading of the context where they are created, but then they automatically vanish; (e) pass identifiers as parameters to templates: a typical use will be Skolemization; (f) fix default expressions for some template parameters.  The process of definition expansion should be guaranteed to converge by such checks, although this fact has not been proved formally so far. Research in the Industry In the period she had been working at the R&D Labs of Italtel SIT in Milan (1983-87), Stefania worked in the Software Engineering Group. The objective of this group was to define, implement and experiment automated or semi-automated methodologies for supporting the specification and design of telecommunication systems. The conceptual tools that were used are: (i) algebraic specification languages, mainly the Clear language by Burstall and Goguen; (ii) Petri Nets. She has spent eight months at the AEG research center in ULM, Germany, partecipating as a software engineer to the definition and experimentation of one of the first prototypes in Europe of a 900Mhz mobile phone. She has developed a real-time test suite that has been widely used for actual experiments.	Teaching Activities Until 1998: undergraduate course in Computer Architectures ("Architettura degli Elaboratori") for students of Computer Science 1999-2001: introductory undergraduate course in Computer Science ("Informatica Generale") for students of Economy Since 2000: undergraduate course in Databases ("Basi di Dati e Sistemi Informativi") for students of Computer Science 2000-2004: undergraduate course in Artificial Intelligence (Intelligenza Artificiale) for students of Computer Science Since 2004: graduate course in Artificial Intelligence + Intelligent Agents for students of Computer Science 2005-2007: Master course in Advanced Databases  ("Basi  .di Dati Avanzate) for students with a degree in Computer Science attending the Master in Web Technologies . Program Committees Seventh Italian Conference on Logic Programming  (GULP'92, held in Tremezzo, Italy (Chair) Eight Italian Conference on Logic Programming   (GULP'93), held in Gizzeria Lido, Italy, June 1993. 1994 Joint Conference on Declarative Programming (GULP-PRODE'94),  held in Peninscola, Spain, September 1994. Twelfth International Conference on Logic Programming (ICLP'95), held at Shohan Village Center, Kanagawa, Japan, June 1995. Second International Workshop on Flexible Query-Answering Systems (FQAS'96), held in Roskilde, Denmark, May 1995. META'96; Eight Workshop on Meta-Programming in Logic, held in Bonn, Germany, September 1996, held in conjunction to JICSLP'96, Joint International Conference and Symposium on Logic Programming. AAAI Spring 2001 Symposium on Answer Set Programming: Towards Efficient and Scalable Knowledge Representation and Reasoning Stanford, USA, March 26-28, 2001 DALT 2003  Declarative Agent Languages and Technologies , AAMAS Workshop - Melbourne, 15 of July 2003 ASP03, Answer Set Programming: Advances in Theory and Implementation, held in Messina, September 26-28, 2003. TPLP&NMR 2004, Theory and Practice of Logic Porgramming and Non-Monotonic Reasoning, to be held in Mexico City, April 26-30, 2004. DALT 2004  Declarative Agent Languages and Technologies , AAMAS Workshop - held in Melbourne, July 2004 Seventh Workshop on Computational Logic in Multi-Agent Systems (CLIMA VII) held in London, UK, 2005 ASP05, Answer Set Programming: Advances in Theory and Implementation, held in Bath, UK, July 2005. 2th RCRA workshop on Experimental Evaluation of Algorithms (RCRA'05) Ferrara, Italy, June 2005. Eight Workshop on Computational Logic in Multi-Agent Systems (CLIMA VIII) RCRA 2006, Giornata di Lavoro: Analisi sperimentale e benchmark di algoritmi per l’Intelligenza Artificiale. 2. CILC07, 22-esimo Convegno Italiano di Logica Computazionale .ASP’07 (Co-Chair) International Workshop on Answer Set Programming and Applications. AITA-08, Meta-Reasoning in Agents CLIMA-IX, Computational Logic in Multi-Agent Systems LANMR-2008, Logic and Non-Monotonic Reasoning RCRA 2008, 15th Workshop on Experimental Evaluation of Algorithms for Solving Problems with Combinatorial Explosion CILC08, 23-esimo Convegno Italiano di Logica Computazionale CILC09, 24-esimo Convegno Italiano di Logica Computazionale CLIMA-X, Computational Logic in Multi-Agent Systems LANMR-2009, Logic and Non-Monotonic Reasoning LPNMR-2009, Logic Programming and Non-Monotonic Reasoning CILC 2010 e 2011, Convegno Italiano di Logica Computazionale CLIMA-XI e XII, Computational Logic in Multi-Agent Systems RCRA 2010, 17th Workshop on Experimental Evaluation of Algorithms for Solving Problems with Combinatorial Explosion LANMR-2010, Logic and Non-Monotonic Reasoning LPNMR-2011, Logic Programming and Non-Monotonic Reasoning ENC 2011, Mexican International Conference on Computer Science KR 2012, 13^th International Conf. on Principles of Knowledge Representation and Reasoning ECAI 2012, European Conference on Artificial Intelligence Stefania Costantini has served as a referee for several national and international Conferences,  and for several Journals, among which the Journal of Logic Programming, the Journal of Automated Reasoning, the IEEE Transactions on Knowledge and Data Engineering, The Journal on Theory and Practice of Logic Programming, the Journal of Logic and Computation, the Journal of Algorithms. Participation to Projects In the ''80 she participated (when working in the industry for  Italtel SIT) to the  Esprit P283 FOR-ME-TOO about Formal Methodologies for the Specification and Design of Telecommunication Systems and to the activities of the CD900 consortium, that specified one of the first European prototype of 900Mhz mobile phones.. In the  '90 she participated to a three-years joint project between the University of Milano and Enel, the Italian Electricity State Company, for the definition and development of formal methods for the management of large organizations data. . In the years 1999-2000 she participated to a  MURST-founded (Ministry of University, Reseach and Technology) about "Intelligent Agents for Information Extraction'', coordinator  Prof. Franco Turini. In the years  2000-2001 she has participated to a  MURST-founded (Ministry of University, Reseach and Technology) about "Aggregate- and number-reasoning for computing: from decision algorithms to constraint programming with multisets, sets, and maps", coordinator  Prof. Domenico Cantone. From 2002 to 2005 she has participated, as the coordinator of the unit of L'Aquila, and as the coordinator of a workpackage, to the WASP Working Group on Answer Set Programming, funded by the Information Society Technologies programme of the European Commission, Future and Emerging Technologies ( IST-2001-37004) From 2005 to March 2007 she has participated to the project CUSPIS: A Cultural Heritage Space Identification System in the context of: Galileo and the European Cultural Assets: a European infrastructure serving another European infrastructure ( GJU/05/2412/CTR/CUSPIS)
	1983. MS degree (laurea magna cum Laude) in Computer Science at the Univ. of Pisa, Italy
1983-1987: Programmer and software designer at Italtel SIT in Milano (a telecommunication Company)
	1987-1990:   position at the Computer Science Dept., Univ. of Milano, supported by fellowships from IBM Italia  and Hewlett-Packard Italia.
1990-1999: Assistant professor at the Computer Science Dept., Univ. of Milano
	1999- February 2001: Assistant Professor at the Dept. of Pure and Appl. Maths. ,Univ. of L'Aquila
March 2001-October 2005: Associate Professor first at the Dept. of Pure and Appl. Maths. , then at the Dept. of Computer Science, Univ. of L'Aquila
	From November 2005: Professor at at the Dept. of Computer Science, Univ. of L'Aquila