Back to search

Hellenic Open University

Collective artificial intelligence and personalized educational interactions in learning new skills

Abstract

dc:description

Social Learning has been an exciting area in Artificial Intelligence for researching learning mechanisms for in multi-agent game playing systems, besides being a useful concept in diverse science areas, like sociology, economics and marketing. The agent-based simulation of social organizations, based on the investigation of agents’ training and learning tactics and strategies, has been inspired by the ability of humans to learn from social environments which are rich in interactions between peers and partial or hidden information. Such richness is a source of complexity that an effective learner has to be able to navigate. Introducing socially inspired dynamics in multi-agent environments is perceived as a promising approach to simulate realistic human-like playing behaviors and highlight the importance of collectively gained knowledge.In this thesis, we first investigate the performance of synthetic agents in playing and learning scenarios in a turn-based zero-sum game and we highlight the ability of opponent-based learning player models to demonstrate a competitive playing edge in social environments. Synthetic agents are generated based on a variety of combinations of some key learning parameters, such as exploitation-vs-exploration trade-off, the ability to look back in time and the ability to project in the future, as well as the speed of learning. These synthetic agents, then, interact over a very large number of games and the experimental data is analyzed to identify clusters of agents which demonstrate interesting associations between the learning parameters and eventual performance ranking. Additionally, we examine how synthetic agents interact in social environments employing a variety of training strategies against diverse opponents. This training and playing experimentation indicates that quality playing relies more on the correct set-up of the learning mechanism than on experience. We, also, investigate the learning progress of inexperienced agents in competitive game playing social environments, and we aim to determine the effect of a knowledgeable opponent on a novice learner. An investigation of the impact of the environmental complexity on the game playing-and-learning behavior of synthetic agents also sheds light on how an effective player has to adapt its learning and playing profile to maintain a given performance profile, when the complexity of a social environment changes. To support our claims, we also conduct a large-scale evaluation of existing rating methods and, as we highlight an inconsistency in how these methods treat synthetic agents, we come up with a proposal for a new rating approach. This approach can be regarded as a base for developing rating systems for multi-agent systems and we demonstrate experimentally its effectiveness.We conclude the thesis by presenting a method, where a synthetic agent experiments with varying the ordering of opponent selection from a group of diverse agents to determine a “best fit for learning” opponent sequence; initial results suggest that good progress and stable evolution arise when an agent faces opponents of increasing capacity, and that an agent with a good learning mechanism set-up progresses better when it faces less favorably set-up agents.The simulation of social societies, as in the previously mentioned experiments, requires vast amounts of computing resources, which must be managed over distributed or high performance computing infrastructures to provide for cost-effective experimentation. To that end, this thesis presents a novel platform for the segmentation and management of social simulation experiments in game-playing multi-agent systems. The platform is managed through a web-based graphical user interface, to combine the advantages of powerful grid infrastructure middleware and sophisticated workflow systems in a way that some generic functionality is sacrificed for the benefit of obtaining a smooth and brief learning curve, without compromising security. Last but not least, this platform can serve as a generic test bed for fellow researchers to experiment with their own turn-based zero-sum games.

Degree

thesis:*
Grantor dc:publisher
Hellenic Open University
Year dc:date
2017

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Kiourt, Chairi
  • Κιουρτ, Χαϊρή

Subjects

dc:subject × 16

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/41990

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Kiourt, Chairi; Κιουρτ, Χαϊρή. Collective artificial intelligence and personalized educational interactions in learning new skills. Hellenic Open University, 2017. http://hdl.handle.net/10442/hedi/41990