Johannes Fähndrich

Abstract: Als Teil des Teams "Advisory Board Polizei-Informatik" wurde Prof. Dr. Johannes Fähndrich von der Hochschule für Polizei Baden-Württemberg (HfPolBW) vor einigen Tagen beim 9. Forum Sicherheit an der Northern Business School in Hamburg mit dem Hamburger Sicherheitspreis ausgezeichnet. Der Preis resultiert auch aus einer "stillen Revolution", die sich in Villingen-Schwenningen vollzogen hat: Die Polizeiausbildung geht in die nächste Runde! In den letzten Jahren wurde das Studium für angehende Kriminalbeamtinnen und Kriminalbeamte überarbeitet und stärker an den modernen Herausforderungen der Kriminalitätsbekämpfung ausgerichtet. Ein neuer Studiengang "K-IT" legt einen Schwerpunkt auf den Umgang mit digitalen Spuren. Die immer schneller voranschreitende Digitalisierung hat die Anpassungen für eine zukunftsfähige Polizei erforderlich gemacht. Die Ausbildung junger Polizistinnen und Polizisten wurde signifikant verbessert und modernisiert, um zeitgemäß den neuen Anforderungen gerecht werden zu können. So war es auch von Beginn an das Konzept von den K-IT Dozenten Prof. Dr. Johannes Fähndrich und Kriminaloberrat Lars Mechler hier im Double-Teaching die relevanten Inhalte akademisch und praxisrelevant umzusetzen. Die ersten Absolventinnen und Absolventen dieses Studiengangs bringen in wenigen Tagen dadurch frischen Wind in die Ermittlungsbehörden, ausgestattet mit fortschrittlichen Werkzeugen und Fachkenntnissen, um ihre kriminalpolizeilichen Aufgaben mit neuer Perspektive und Effizienz zu erfüllen. Die Inhalte des neuen Studiengangs wurden durch eine zukunftsorientierte Bewertung von Technologie und deren Einfluss in der Kriminalitätsbekämpfung ausgewählt. Dieses Prinzip wurde mit dem Hamburger Sicherheitspreis geehrt.

Abstract: Forensics is an established field of research. Digital forensics started 44 years ago with the Florida Computer Crimes Act (1978) including legislation against the unauthorized modification of data on a computer system. Since then, the field has flourished in different subdomains. The overall definitions and concepts have been specified by a small group of experts. Furthermore, the need for development is created by the amount of digital evidence which is collected concerning most crimes. This paper gives an overview of the state-of-the-art by presenting a structured literature review of digital forensic about methods and concepts using Artificial Intelligence (AI) technologies. The review focuses on science done on topics in strong AI and forensics.

Abstract: Im Kontext der Cybersicherheit manifestiert sich die Rolle der Künstlichen Intelligenz (KI) als zunehmend ambivalent [14]. Einerseits ermöglicht KI eine effiziente und hochwertige Entwicklung von Software, die auch die Programmierung von Malware umfasst. Andererseits dient KI als fundamentales Instrument zur Bekämpfung dieser Bedrohungen. KI-unterstützte Systeme wie beispielsweise Intrusion Detection Systems (IDS) und Intrusion Prevention Systems (IPS) ermöglichen es, Anomalien im Netzwerkverkehr automatisiert zu detektieren und entsprechende Reaktionen zu initiieren [8]. Dies zeigt die Doppelnatur der KI, die sowohl als Mittel für Angriffe als auch für die Verteidigung in der Cyberwelt eingesetzt werden kann. Überdies ermöglicht KI Cyberkriminellen, sich in kürzester Zeit in spezifische Themen einzuarbeiten, was unter anderem Phishing-Angriffe glaubwürdiger und somit schwerer zu identifizieren macht.

Abstract: Unsere Welt wird zunehmend digitaler und immer mehr Technologien finden Einzug in unseren Alltag [23]. Dies reicht von der standardmäßigen Verschlüsselung unserer E-Mails über Services in der „Cloud“ sowie Quantencomputing oder Virtual Reality und vielem mehr. Vieles dieser Digitalisierung bringt unsere Gesellschaft voran [16]. Ob die Nachteile, die eine neue Technologie mit sich bringt, die Vorteile wert sind, ist eine Frage für Philosophen [14]. Eine Entscheidung treffen zu können, bevor alle positiven und negativen Use Cases einer Technologie ausprobiert wurden, könnte helfen, intelligent mit Technologie umzugehen [1] hier solle eine Methode praktisch in einem Proof of Concept getestet werden, Technologie zu bewerten. Die Bewertung wird hier auf die Polizeiarbeit angebaut und an zwei Use Cases ausprobiert.

Abstract: Die achte Fachtagung fand vom 14. bis zum 15. März 2023 hybrid an der Hochschule für Polizei Baden-Württemberg in Villingen-Schwenningen statt. In sehr heterogenen Beiträgen wurde die ganze Bandbreite der Lehre, Forschung und Praxis in der polizeilichen Informatik und Informationstechnik beleuchtet. Die Referierenden hielten Präsenz- und Online-Vorträge zu aktuellen Arbeiten und visionären Ideen rund um die polizeiliche Informatik/Informationstechnik. Die eingereichten Beiträge werden in diesem Sammelband vorgestellt.

Abstract: Digital forensics has many properties of evidence, which need to be defined formally. The goal of this paper is to formalize the properties of evidence with the application of digital forensics. The contribution is a discussion of the current definitions of volatility and manipulability. The analysis starts the discussion on the application of those concepts to the application of ontological definitions to digital forensics. With that, we want to discuss the trust and the need to understand complex AI models.

Abstract: Digital forensics has many properties of evidence, which need to be defined formally. The goal of this paper is to formalize the properties of evidence with the application of digital forensics. The contribution is a discussion of the current definitions of volatility and manipulability. The analysis starts the discussion on the application of those concepts to the application of ontological definitions to digital forensics. With that, we want to discuss the trust and the need to understand complex AI models.

Abstract: Digitale Spuren müssen von gut ausgebildeten Polizistinnen und Polizisten erkannt, vor Veränderung geschützt sowie gesichert und ausgewertet werden. Sie sollten also wissen, in welchen Zusammenhängen welche Spuren entstehen, wie diese zu finden sind und wie diese beweiskräftig gesichert werden können. In diesem Kapitel wird anhand von Basiskompetenzen gezeigt, welcher Rahmen derzeit in den Bundesländern für eine entsprechende Lehre im Polizeistudium zur Verfügung steht. Anschließend wird die notwendige Anpassung der Lehre an den ständigen technischen Wandel beschrieben.

Abstract: Prof. Dr. Klaus-Robert Müller is rated one of the top cited computer scientists in Germany and number 47 internationally by Reseach.com.1 We interviewed Prof. Dr. Müller because of his perspective on AI and its explainability. With his approaches to analyzing Deep Neural Networks, he might be one of a few leading scientists in the world who are researching the questions: “Can we trust results out of an AI?” This question is of particular interest to the community of digital forensics. On grounds of that, many of the modern challenges are about handling big amounts of data. We believe that AI could be the integral part of future investigations with digital evidence.

Abstract: When crimes are committed, countless traces are created. An constantly growing proportion of these are digital traces. For example approximately 400 thousand new variants of malware enter circulation every day [1]. The increasing proliferation of information systems offers an ever-growing gateway for these malicious programs. With the increasing use of digital communication channels such as instant messaging, the number of traces to be analyzed has grown far beyond human capabilities [2]. With the simplified use of anonymization techniques, new challenges arise, such as the use of author determination methods [3]. This has been researched for years for traditional media such as e-mail, [4] but has found little application in the forensic context. Approaches to the use of machine learning in forensics have also been part of the scientific discourse for some time [5]. The field of digital forensics has specific requirements for the methods used. Chain of custody and legal certainty as well as data protection are major hurdles for the use of Artificial Intelligence (AI).

Abstract: A need for the use of automation in digital forensics is imminent. With the overwhelming workload for analysing collected digital evidence, law enforcement agencies are no longer meeting the quality investigations we expect. For automation to work (e. g., integrating heterogeneous data sources, structuring unstructured data, or drawing conclusions from structured data), formalisation must be specified. Specification of formalisation includes as a first step to describe concepts in the domain of digital forensics. In this contribution, we analyse the state of the art of ontological formalisation in the domain of digital forensics via a scoping review. There are some attempts to formalise the technical domain of digital forensics in ontologies, but they do not cover essential context, like technical annotations or acquisition modelling. Future work will be to use the existing foundations and extend them with facts and rules to enable reasoning.

Abstract: This paper decibels the result of an automated decomposition algorithm that can be utilized to create knowledge graphs out of different information sources. We present a formal definition of such a knowledge graph and apply it in the context of law enforcement and investigations. The created knowledge graphs are then used for a graph search algorithm called marker passing. With this, we show that the aggregation of knowledge out of heterogeneous information sources can be a combination of symbolic and connections approaches. The application of such an automatically generated knowledge graph could be in a deeper understanding of the text in applications where too much text is to be processed with only parts of it being relevant. Examples could be confiscated date, which needs to be searched for a certain type of evidence that is not intuitive recognizable without contextual information. Aiding the human investigator to select relevant information requires NLU and models, which are explainable and hold up in court.

Abstract: Die seit Jahrhunderten verwendeten Methoden in der Forensik basieren auf der Annahme eines Austausches von Materie und Mustern. Durch die Digitalisierung sind diese Annahmen nur noch eingeschränkt gültig und werden hier erweitert und diskutiert. In dem Zusammenhang ist es erforderlich, den Spurenbegriff grundlegend zu überdenken. Gleichzeitig werfen der ständige technische Fortschritt und die immer größer werdende Flut von auszuwertenden Daten die Ermittlungsbehörden immer wieder zurück. Dieser Entwicklung ist nur durch Automatisierung Herr zu werden. Verfahren der Künstlichen Intelligenz können und werden die Ermittlungsbehörden zukünftig dabei zunehmend unterstützen.

Abstract: While authorities are moving towards the adoption of more electro-mobility services in urban environments, their application in the real world remains still low due to a number of challenges. The integration of services from different providers using different data formats hinders interoperability; a secure and trustworthy data exchange is needed to preserve the rights of beneficiaries on the shared data. The NeMo project addresses those challenges by providing a decentralized service platform based on a distributed ledger to enable interoperable and secure data flow between service providers throughout Europe. This paper provides details about the architecture, enabling easy integration of electro-mobility services and demonstrates the approach by showing the overall flow from designing to using the electro-mobility services.

Abstract: This paper approaches a solution of Winograd Schemas with a marker passing algorithm which operates on an automatically generated semantic graph. The semantic graph contains common sense facts from data sources form the semantic web like domain ontologies e.g. from Linked Open Data (LOD), WordNet, Wikidata, and ConceptNet. Out of those facts, a semantic decomposition algorithm selects relevant facts for the concepts used in the Winograd Schema and adds them to the semantic graph. Markers are propagated through the graph and used to identify an answer to the Winograd Schema. Depending on the encoded knowledge in the graph (connectionist view of world knowledge) and the information encoded on the marker (for symbolic reasoning) our approach selects the answers. With this selection, the marker passing approach is able to beat the state-of-the-art approach by about 12%.

Abstract: This paper approaches a solution of Winograd Schemas with a marker passing algorithm which operates on an automatically generated semantic graph. The semantic graph contains common sense facts from data sources form the semantic web like domain ontologies e.g. from Linked Open Data (LOD), WordNet, Wikidata, and ConceptNet. Out of those facts, a semantic decomposition algorithm selects relevant facts for the concepts used in the Winograd Schema and adds them to the semantic graph. Markers are propagated through the graph and used to identify an answer to the Winograd Schema. Depending on the encoded knowledge in the graph (connectionist view of world knowledge) and the information encoded on the marker (for symbolic reasoning) our approach selects the answers. With this selection, the marker passing approach is able to beat the state-of-the-art approach by about 12%.

Abstract: The development of self-adaptive systems can greatly benefit from reference frameworks to structure the development process. Current reference frameworks abstract from the adaptation decision – the selection of a specific adaptation based on the goals and options available. This decision is usually based on the three adaptation policies: rule-based, goal-based or utility based. However, concepts from these policies often are combined with each other in different ways to achieve hybrid policies. To structure this combination this paper identifies four types of such combinations. We express these types as patterns within a model that captures the functions, models and relations participating in the adaptation decision.

Abstract: The ongoing digitalization trend has given rise to the concept of smart cities, targeting the interconnection of city infrastructure and services over a digital layer for innovative technological solutions as well as improvements on existing facilities in cities. This article investigates the critical information system constituents of smart cities that facilitate the holistic integration of its ecosystem and resources with the aim to foster dynamic and adaptive software. We identify three main enablers in this direction: (i) semantic functional description of city objects, representing physical devices or abstract services, (ii) a distributed service directory that embodies available city services for service lookup and discovery, (iii) planning tools for selecting and chaining basic services to compose new complex services. We provide an overview of the approach adopted in our ongoing smart city project for each of these three dimensions.

Abstract: With the increase complexity of IoT systems, the well known development paradigm for software development like Service Oriented Architectures and Multi Agent Systems have to be adapted tot the now challenges of IoT environments. Here the vast amount of available components and the specic implementation for special purpose hardware calls for an abstraction of functionality as well as to new development tools which allow to handle the dynamism of IoT applications. Here a multitude of special purpose sensors with view computational resources have to be combined to create emerging software. Most of the hardware reaches an return of investment only oering its service to third party developers. For a developer to be able to integrate those services into an application, a adaptive search mechanism need to be developed, which is able to specialize in the dierent domains of e.g. IoT applications.

Abstract: The capability to identify the sense of polysemic words, i.e. words that have multiple meanings, is an essential part of intelligent systems, e.g. when updating an agent's beliefs during conversations. This process is also named Word Sense Disambiguation (WSD) and is approached by applying semantic similarity measures. Within this work, we present an algorithm to create such a semantic similarity measure using marker passing, that: (1) generates a semantic network out of a semantic service description, (2) sends markers through the networks to tag sub-graphs that are of relevance, and (3) uses these markers to create a semantic similarity measure. We will discuss the properties of the algorithm, elaborate its performance with dierent part of speech, and discuss the lifted properties for the algorithm to be used in WSD. To evaluate our approach, we compare it to state-of-the-art measures using the Rubinstein1965 and WordSim353 dataset. It is shown, that our approach outperforms these state-of-the-art measures and, further, is able to adapt the predictions to the contextual information.

Abstract: With the rise of new paradigms like the Internet of Things, where thousands of devices and services of different providers are to be connected to complex processes, service-oriented approaches come to the fore. However, current solutions still lack of comprehensive methodologies how to dynamically manage and combine services to fulfil the given goals. In this paper we present a semantic-based service management methodology that enables the semantic description of services and provides an automatic service discovery and composition solution at design- and runtime. Furthermore, we present development tools that support the usage of semantic web technologies and we describe an execution environment where the approach is embedded. We conclude with an evaluation scenario from an e-mobility research project.

Abstract:Personality is one of the central elements determining the behaviour of humans. It influences other cognitive mechanisms such as emotions and moods and thus effects attention and actions. However, in the literature about cognitive agents, work that investigates the effects of personality is rare and somewhat disconnected. Bridging this gap represents one step towards conceptualising human behaviour in software agents, e.g. for resource-bounded agents in highly-dynamic environments or for virtual humans with realistic behaviour. The integration of personality in agents also requires its integration into reasoning processes used in agent-based systems. In this paper, we propose a formalisation that enables reasoning about the effectsand state of personality.

Abstract: The applicability and usefulness of agent technology for real world problems is still a matter of discussion---even within the AAMAS community. While theoretical models have significantly matured and led to an exciting variety of results, there were only few attempts to validate these models in reality. In this paper we aim to report on challenges that occurred when agent theory was used to approach real world problems. In doing so, we focus on the concept of planning, since planning currently appears to be one of the most relevant concepts for distributed real world applications. We examine four agent-based applications and emphasise problems that occurred when agent theory was practically applied. We also show how these problems were countered and propose more general solutions based on theses tailored and context-specific approaches. The aim of this paper is to bring the two diverging branches of agent theory and practice back together in order to better adapt agent technology to the requirements of professional software.

Abstract: Semantic information is considered as foundation upon which modern approaches attempt to tackle the challenges of dynamic environments – service orchestration and ontology matching are two examples for the use of such information. Yet, many developers avoid the additional effort of adding semantic information (e.g., through annotations) to their data sets – limiting the reusability and interoperability of their Apps, services, or data. This problem is called the “knowledge acquisition bottleneck”, which can be addressed by providing suitable tool support. This survey analyses the state-of-the-art of such tools that support developers in the task of semantically enriching entities. Providing an overview of available tools from the early days until now, we particularly focus on the ‘level of automation’. Concluding that automation is very limited in contemporary tools we propose a concept that mixes connectionist and symbolic representation of meaning to decrease the manual effort.

Abstract: Making artificial agents a constituent part of human activities leads to more affiliated teamwork scenarios and at the same time introduces several new challenges. One challenge is the team members' ability to be mutually predictable to each other, which is required to effectively plan own actions, e.g., in the field of human-aware planning. This work approaches the question whether agents are able to learn the personality of a human during interaction. In particular, we developed an agent model able to learn human personality during repeatedly played rounds in the Colored Trails Game. Human personality is described using a psychological theory of personality types known as the Five-Factor Model. The results indicate that some characteristics of a personality can be learned more accurately/easier than others.

Abstract: Due to the fact that electric vehicles have not broadly entered the vehicle market there are many attempts to convince producers to integrate technologies that utilise embedded batteries for purposes different from driving. The vehicle-to-grid technology, for instance, literally turns electric vehicles into a mobile battery, enabling new areas of applications (e.g. to provide regulatory energy, to do grid-load balancing, or to buffer surpluses of energy) and business perspectives. Utilising a vehicle's battery, however is not without a price - in this case: the driver's mobility. Given this dependency, it is interesting that most available works consider the application of electric vehicles for energy and grid-related problems in isolation, that is, detached from mobility-related issues. The Distributed Artificial Intelligence Laboratory, or DAI-Lab, is a third-party funded research lab at Technische Universitat Berlin and integrates the chair for Agent Technologies in Business Applications and Telecommunication. The DAI-Lab has engaged in a large number of both, past and upcoming projects concerned with two aspects of managing electric vehicles, namely: energy and mobility. This article aims to summarise experiences that were collected during the last years and to present developed solutions which consider energy and mobility-related problems jointly.

Abstract: There is increasing interest in the agent community to integrate the concept of emotions and articial agents. The spectrum of available solutions reaches from applications and models of emotions to complete axiomatised logics. Despite the rich oer of solutions, available works neglect individual personality as a signicant factor for the outcome of emotional behaviour pattern. However, dierent personalities aect all relevant phases of human decision-making processes. In this paper, we introduce and discuss existing personality theories, propose a concept based on the beliefdesire- intention paradigm, and show how such concept can be implemented as prototype.

Abstract: This paper surveys languages used to enrich contextual information with semantic descriptions. Such descriptions can be e.g. applied to enable rea- soning when collecting vast amounts of row data in domains like smart environments. In particular, we focus on the elements of the languages that make up their semantic. To do so, we compare the expressiveness of the well-known languages OWL, PDDL and MOF with a theory from linguistic called the Natural Semantic Metalanguage.

Abstract: This paper describes an algorithm for semantic decomposition. For that we surveys languages used to enrich contextual information with semantic descriptions. Such descriptions can be e.g. applied to enable reasoning when collecting vast amounts of information. In particular, we focus on the elements of the languages that make up their semantic. To do so, we compare the expressiveness of the well-known languages OWL, PDDL and MOF with a theory from linguistic called the Natural Semantic Metalanguage. We then analyze how the semantic of the language is build up and describe how semantic decomposition based on the semantic primes can be used for a so called mental lexicon. This mental lexicon can be used to reason upon semantic service description in the research domain of service match making.

Abstract: We advocate Self-Explanation as the foundation for the Self-* properties. Arguing that for system component to have such properties the underlining foundation is a awareness of them selfs and their environment. In the research area of adaptive software, self-* properties have shifted into focus pushing ever more design decisions to a applications runtime. Thus fostering new paradigms for system development like intelligent agents. This work surveys the state of the art methods of self-explanation in software systems and distills a definition of self-explanation.

Abstract: When it comes to planning for joint human-agent activities one has to consider not only a exible plan execution and social constraints but also the dynamic nature of humans. This can be done by providing additional information about a humans characteristics. For example, taking the physical and psychological condition of an elderly into consideration when developing collaborative applications like socially assistive robots. This work outlines Hplan, an extension to the agentframework JIAC V, that takes this requirement into account. HPLAN is strongly related to the conceptual model of dynamic planning components and integrates humans as avatars into a life cycle of planning, execution and learning.

Abstract: We advocate Self-Explanation as the foundation for the Self-* properties. Arguing that for system component to have such properties the underlining foundation is a awareness of them selfs and their environment. In the research area of adaptive software, self-* properties have shifted into focus pushing ever more design decisions to a applications runtime. Thus fostering new paradigms for system development like intelligent agents. This work surveys the state of the art methods of self-explanation in software systems and distills a definition of self-explanation.

Abstract: To become a part of a joint human-agent team, articial agent are required to achieve joint goals with humans not only performing task for humans. This includes the ability to coordinate actions between team-members, which is e.g. addressed by Human-Aware Planning approaches. This work surveys available solutions regarding the special requirements identified for joint human-agent activities. In articular, the work concentrates on the requirement of interpredictability, which requires to include the course of actions of other team-members into the planning process of one's own course of action.

Abstract: This work presents a novel approach for combining multiple Electronic Patient Records (EPRs) to a self-learning fall risk assessment tool. This tool is used by a new type of home-visiting nurses to track the fall risk of their patients. In order to provide personalized healthcare for elderly people, we combine multiple EPRs using an agent-based architecture, where each patient is represented by an associated agent. The patient agents are enabled to negotiate about possible fallrisk indicators recognized in the specific patient population under care. We use distributed information fusion and opinion aggregation techniques to elaborate new fall-risk indicators and in consequence to adapt the fall risk assessment tool.

Abstract: This work introduces a prototype that demonstrates the idea of using a psychological theory of personality types known as the Five-Factor Model (FFM) in planning for human-agent teamwork scenarios. FFM is integrated into the BDI model of agency leading to variations in the interpretation of inputs, the decision-making process and the generation of outputs. This is demonstrated in a multi-agent simulation, further, it is outlined how these variations can be used for the planning process in collaborative settings.

Abstract: We advocate Self-Explanation as the foundation for the Self-* properties. Arguing that for system component to have such properties the underlining foundation is a awareness of them selfs and their environment. In the research area of adaptive software, self-* properties have shifted into focus pushing ever more design decisions to a applications runtime. Thus fostering new paradigms for system development like intelligent agents. This work surveys the state of the art methods of self-explanation in software systems and distills a definition of self-explanation.

Abstract: We advocate Self-Explanation as the foundation for the Self-* properties. Arguing that for system component to have such properties the underlining foundation is a awareness of them selfs and their environment. In the research area of adaptive software, self-* properties have shifted into focus pushing ever more design decisions to a applications runtime. Thus fostering new paradigms for system development like intelligent agents. This work surveys the state of the art methods of self-explanation in software systems and distills a definition of self-explanation.

Abstract: This work presents an approach combining multiple electronic patient records (EPR) to a self-learning fall risk assessment tool. We utilised the agentperspective to model the system, to address privacy issues and to evaluate different distributed information fusion and opinion aggregation techniques towards there applicability to the addressed d omain. Each agent represents a single patient negotiating about unknown fall risk influences in order to adapt the fall-risk assessment tool to the population under care. In addition, we will outline the planned real-world case study.

Abstract: This work presents an approach combining multiple electronic patient records (EPR) to a self-learning fall risk assessment tool. We utilised the agentperspective to model the system, to address privacy issues and to evaluate different distributed information fusion and opinion aggregation techniques towards there applicability to the addressed d omain. Each agent represents a single patient negotiating about unknown fall risk influences in order to adapt the fall-risk assessment tool to the population under care. In addition, we will outline the planned real-world case study.

Abstract: When it comes to the design time of software applications, unspecified details always leave room for alternatives. Each alternative requiring self-explanation capabilities in order help with the selection. This work subdivide the problem of self-explanation into four subproblems: First of all, we need to answer what is an explanation? In this work explanations are created by adding semantic and contextual information to descriptions (at runtime), which will ease the inference of some reasoner observing the explanation. Secondly, we need to know what kind of information should be contained in a self-explanation? This also implies the question on how much of a language needs to be known, to understand a description in this language. Afterwards, we must accomplish a way to explain semantic information using such a metalanguage and enriching it with contextual information. Hence, this work analyses the way components are described, how these descriptions are created and which methods of inference can be build upon them.

Abstract: Nowadays, many research institutes are largely dependent on third party funding. This situation leads to practical work or in other words project work which exceeds the research typical proof of concept implementation. As we were tired of seeing a downward trend in the number of accepted application oriented (full) papers in the major agent conference, we conducted a survey to provide evidence for the thesis that researchers can gain relevant benefits from project work for their research work. Hence, in this paper, we present the results of this survey and discuss different scientific questions researchers should ask themselves during project work.

Abstract: In this work we propose an approach for comfortable and accelerated development of user interfaces for software agents. We apply model-based techniques and emphasise the capability of this technique by describing two user interfaces which are different in nature, but have been developed with the same model. We present the applicability of both user interfaces by means of an agent-based application in the domestic energy domain. As opposed to similar approaches we retain all degrees of freedom for the applied multi-agent framework.

Abstract: Bayes’sche Informationsfusion kann einen hohen Rechenaufwand erzeugen. Zum Zweck einer Verkürzung der Berechnungszeit wird in dieser Arbeit eine verteilte lokale Bayes’sche Informationsfusion betrachtet. Die Größen von Interesse werden durch eine Lokalisierung in Mengen aufgeteilt, die Bereiche von erhöhtem Interesse beinhalten. Auf diesen Mengen kann ein Agentensystem die Fusion verteilt durchführen. Dadurch, dass die Agenten des Agentensystems sich auf diese lokale Menge der Größen von Interesse beschränken, wird die Wahrscheinlichkeit für die Elemente dieser Menge wahrscheinlicher als sie es global gesehen sind. Die Ergebnisse der verteilten Berechnung müssen so zusammengeführt werden, dass möglichst das gleiche Ergebnis wie bei einer nicht verteilten Bayes’schen Informationsfusion entsteht. Dieses Problem ist mit dem einer Konsensfindung in einer Gruppe vergleichbar. Dabei muss aus den A-posteriori-Wahrscheinlichkeitsverteilungen der Gruppenmitglieder, eine Wahrscheinlichkeitsverteilung erzeugt werden, welche die Ground Truth ausreichend gut annähert.
Eine Pooling-Methode ist eine Methode der Entscheidungstheorie. Diese Pooling-Methoden vereinen die Wahrscheinlichkeitsverteilung, welche die Meinungen/Schätzungen der Gruppenmitglieder repräsentieren. Die entsprechende Fachliteratur wurde nach Pooling-Methoden und Möglichkeiten durchsucht, wie diese bewertet werden können. In der Entscheidungstheorie geschieht dies anhand von Eigenschaften, welche die Pooling-Methoden erfüllen können. Dieser Ansatz wurde hier für eine verteilte lokale Bayes’sche Informationsfusion fortgesetzt. Es werden Eigenschaften von Pooling-Methoden auf ihre Relevanz hinsichtlich einer verteilten lokaler Bayes’schen Informationsfusion untersucht. Auf Ba sis dieser Einsichten werden Vor- und Nachteile von Pooling-Methoden, (am Beispiel des Linear Opinion Pools, des Logarithmic Opinion Pools und des Supra Bayes Ansatz) im Kontext von lokaler Bayes’scher Informationsfusion beleuchtet. Somit wird für diese Beispiele durch Analyse und Bewertung der Eigenschaften eine Aussage über deren Eignung für eine verteilte lokale Bayes’sche Informationsfusion abgeleitet.