September 25, 2023 Anaïs Siebers
In cultural knowledge evolution simulated by multi-agent simulations, agents can improve the accuracy of their knowledge by interacting with other agents and adapting their knowledge with the aim of agreeing. But their knowledge might be confined to specific areas because they do not have the capacity to ex- plore the world around them. Since intrinsic motivation to explore in artificial agents has already proven to increase exploration, it was researched whether and how agents in simulations of cultural knowledge evolution can be motivated to explore. Moreover, it was tested how far this improves and changes their knowledge. Three different kinds of motivation were investigated: curiosity, creativity and non-exploration. Moreover, intrinsic motivation was modelled with and without reinforcement learning. Agents either explored on their own or picked specific interaction partner(s). It has been shown that it is pos- sible to model agents with intrinsic motivation to explore in cultural knowledge evolution, and that this has a significant effect on the agents’ knowledge. Con- trary to the expectations and other studies, this did not lead to an increase in knowledge completeness. Out of all intrinsic motivations, curiosity had the highest accuracy and completeness. Models with reinforcement learning per- formed similar to direct models. As expected, intrinsic motivation led to faster convergence of the agents’ knowledge, especially with social agents. Hetero- geneously motivated agents only had a higher accuracy and completeness than homogeneously motivated agents in specific cases. This thesis can be regarded as a foundation for further investigation into the role of intrinsic motivation in cultural knowledge evolution. Different forms of intrinsic motivation or differ- ent reinforcement learning techniques could be tested. Additionally, intrinsic motivation at different stages of the experiment or in different ratios, for ex- ample curious agents and agents with no motivation, could be investigated in more detail. Lastly, agents could teach other agents things they explored a lot.
Stichworte: cultural knowledge evolution; intrinsic motivation; exploration; artificial curiosity; computational creativity; multi-agent simulation
Sprache: Englisch
Juni 10, 2023 Anaïs Siebers
With advances in information technology, human agency and autonomy during the interaction with technological systems seem to be in danger as technology feels “out of control” for many people. Their judgement is based on their sense of agency during the interaction with the technology. Thus, there have been a lot of voices raised, demanding more transparency. This paper investigates users’ sense of agency during the interaction with machines, and proposes criteria, which make up the sense of agency on the internet. On this basis, agency in the context of search is shortly analysed. Results reveal that recommendations in current search systems negatively impact the user’s sense of agency. But the system can, the less instrumental the purpose of the user’s search is, exhibit more agency and autonomy. If the user receives sufficient feedback about the system’s reasoning and intentions, a joint-agency can be experienced. Otherwise, the user’s sense of agency decreases with increasing automation of the system. The criteria are a first proposal to motivate further discussion and research, as well as influence future system’s design.
Stichworte: agency; sense of agency; human-machine interaction; AI; search
Sprache: Englisch
November 2, 2022 Anaïs Siebers
Search engines have become an important aspect of everyday life. They are nearly invisible, ubiquitous, and shape people’s actions in and knowledge of the world. But because of this invisibility, the daily impact a search result has on the individual and society when searching as well as when offering information is unaware. This paper investigates the impact a search result has by analysing different stages of the search: from entering a search query over retrieving results up to the presentation and selection of a result among others. The focus is on the existential power search engines exert.
Stichworte: search engines; human-centered computing; social computing; categorisation; Google
Sprache: Englisch
Oktober 12, 2022 Anaïs Siebers
A machine which rationally reasons like humans sounds like science fiction and is far from becoming part of everyday life. But the demand and necessity for explainable artificial intelligence which does not “decide” / “act” like a black box, but can explain its processes and reason about its decisions and judgements, strongly increases. In cognitive science, reasoning is also a currently strongly researched topic. The most dominant model at the moment is the dual-process theory, which attempts to combine the previously widespread symbolic, heuristic and probabilistic approaches. Processes of type 1 are fast, but biased, unconscious and strongly contextualised, whereas processes of type 2 are slow, conscious, controlled and unbiased. This paper proposes how machine reasoning on the basis of this dual-process theory could look like. The greatest difference to current attempts in machine reasoning is that the probabilistic black-box-like algorithms are accepted as processes of type 1. Future research is required concerning the transition and interplay between the processes and, if existent, the monitoring process controlling both processes and their interaction, to build a foundation for a successful implementation for machine reasoning.
Stichworte: dual process theory; machine reasoning; logic reasoning; probabilistic reasoning; recognition; cognition
Sprache: Englisch
August 21, 2022 Anaïs Siebers
It is difficult to imagine a machine which makes ironic remarks and reacts appropriately to humans making ironic statements. But the amount of verbal human-machine interaction in everyday life increases constantly, and the experienced quality depends heavily on the naturalness of the conversation and relationship between user and machine. This interaction has been discovered to be positively influenced by machine’s use of irony. Furthermore, increasing hate speech has started research into the sentiment and meaning of social media content, to be able to distinguish hateful from ironic messages and only restrict hateful messages. Apart from that, computational irony might also be a useful tool to shed new light on the debate about irony processing and give more insight into the cognitive processes underlying irony. Current approaches to computational irony focus on the detection of irony in text. Since irony has a very important purpose in communication, it is foreseeable that future research will focus more strongly on other aspects like irony comprehension and production – also in speech. This paper investigates whether machines will be able to identify, comprehend and produce irony by investigating the current level of research of irony and computational irony. It will be argued that it is likely that machines reach a human-like level of irony identification, comprehension and production. Further research concerning irony will be necessary especially in the context of background-information, social, cultural and general context as well as when it is appropriate to use irony.
Stichworte: irony identification; irony comprehension; irony production; irony; machines; computational irony
Sprache: Englisch
März 15, 2022 Anaïs Siebers
This essay has three major parts: The first parts starts by laying the foundations for a relationship between inner monologue and consciousness. I will argue why inner monologue is sufficient for consciousness (mainly by looking at the Attention Schema Theory (AST)). On this basis, in the second part criteria which inner monologue must have for machine consciousness, are elaborated, creating a catalogue applicable to architectures of machine inner monologues. The third part deals with the two questions: Whether current implementations are sufficient for machine consciousness, and if machines can have inner monologues that fulfil the criteria of consciousness in front of the background of AST. Inner monologue is used to test whether a machine is conscious in terms of AST. The result and answer suggest that inner monologue can be regarded as a sufficient criterion for machine consciousness, but that the construction and implementation of machines will not yet allow for an inner monologue sufficient for machine consciousness. This essay does not answer whether there are different forms of inner monologue that might also be sufficient, but aims to assess machine inner monologue against a catalogue with criteria for inner monologue overlapping with requirements posed by AST.
Stichworte: Inner monologue, machine consciousness, consciousness, inner speech
Sprache: Englisch
Juli 29, 2021 Anaïs Siebers
Das Ziel meiner Bachelor-Thesis ist die Akzeptanz von Sprachassistenten zu verbessern und die Motivation sie zu nutzen zu erhöhen. Um dieses Problem zu lösen, werde ich Probleme in der Kommunikation untersuchen und einen Ansatz diese zu minimieren entwickeln. Fokus ist das Dialogmanagement – ich setze voraus, dass andere Komponenten eines Sprachassistenten (NLP, NLU, …) bereits ausreichend und zuverlässig funktionieren. Sie werden jedoch so gekapselt, dass sie bei neuen Erkenntnissen und Fortschritt einfach ausgetauscht werden können. Da das Vertrauen in Sprachassistenten oft an der Angst scheitert, abgehört zu werden, sollen die verwendeten Komponenten open source und der Assistent offlinefähig sein.
Sprache: Englisch
Exposé ThesisApril 18, 2021 Anaïs Siebers, Felix Hamann
Das Erstellen von Wissen (hier: knowledge bases KB-s) ist ein zeitkonsumierender Prozess. Um den Wissens-Ingenieur (knowledge engineer) dabei zu unterstützen neue Fakten über Domänen-Entitäten zu entdecken, ist es unser Ziel, ein Szenario zu bauen, das dem 10-Fragen Spiel ähnelt. Ein Orakel fragt nach Relationen oder Relation-Tail Kombinationen, die so akkurat wie möglich sein sollen. Wir schlagen einen Ansatz vor, der Naive Bayes nutzt um die Möglichkeiten und ihre Wahrscheinlichkeit zu berechnen.
Sprache: Englisch
KGC Naive BayesMärz 09, 2020 Eva Hennefeld, Anaïs Siebers
Zentraler Gegenstand des Papers ist ein bisher nicht graphentheoretisch erfasstes Problemgebiet. Dies wird an folgendem praktischen Beispiel visualisiert: Analog zu dem in der Graphentheorie behandelten und gelösten „Heiratsproblem“ geht es um zwei unterschiedliche Gruppen an Persona, zwischen denen ein Matching gefunden werden soll. Allerdings muss das Matching mit verschiedenen divergierenden Kategorien gefunden werden, wodurch das Problem ein Optimierungsproblem wird.
Sprache: Deutsch
GTGA Paper GTGA PräsentationFebruar 05, 2019, Anaïs Siebers
Im ersten Semester haben wir ein Booklet gestaltet. Dort sind verschiedene gestalterische Aufgaben enthalten:
Sprache: Deutsch
booklet.Oktober 22, 2017, Anaïs Siebers
Mon projet interculturel est un projet français. Je vais aller à l’université de Lille 1 - c’est l’université des sciences et technologies de Lille. Je vais m’y rendre pendant mes vacances d’automne.
La transition du lycée vers l’université m’intéresse beaucoup parce que ma matière principale à l’école est l’informatique et je veux en faire mes études après le BAC en juillet. Alors, je me suis beaucoup occupée de mon avenir universitaire en Allemagne. En trouvant differentes possibilités, je me suis demandé comment les élèves des autres pays continuaient après qu’ils ont eu leur BAC. En Allemagne c’est bien vu de faire un an de pause ou de choisir une université. Puis j’ai pensé étudier à l’étranger - entre autres en France.
Sprache: Französisch
ProjektdokumentationAugust 10, 2017, Anaïs Siebers
Mein persönliches Interesse bestand darin, eine eigene Typographie zu entwickeln. Die Schwierigkeit Wörter durch ihre visuelle Darstellung zu unterstützen und dieses Machtinstrument zu nutzen und den Leser sowohl emotional als auch unbewusst zu beeinflussen. Daher habe ich mich zunächst mit den Grundlagen und der Definition auseinandergesetzt, damit ich im Anschluss eine eigene Typographie für einen spezifischen Anwendungsfall entwickeln konnte.
Sprache: Deutsch
Kunst besondere LernleistungMai 01, 2017, Anaïs Siebers
In meiner Facharbeit habe ich einen Ansatz für die Entwicklung eines KI-basierten Assistenten beschrieben. Der Assistent soll im Gegensatz zu anderen Assistenten die Individualität und Persönlichkeit des Nutzers würdigen. Das Prinzip des Assistenten ist an den Lern- und Wachstumsprozess des Menschen angelehnt.
Im Gegensatz zu anderen Assistenten soll er mehr auf den Benutzer und dessen Interessen eingehen. Andere Assistenten werden überwiegend von wirtschaftlichen Interessen des Anbieters bestimmt. So wird bei Google, zum Beispiel, meistens zuerst Amazon kommen, wenn man nach einer Ware sucht. Doch bei diesem Assistenten soll es möglich sein, seine eigenen Quellen auszusuchen und zu priorisieren.
Aufgrund der Komplexität des Themas befindet sich der Assistent allerdings immer noch in einem Beta Stadium und ist Gegenstand weiterer Entwicklung und Forschung.
Sprache: Deutsch
Facharbeit