M.Sc. Simo Linkola defends his doctoral thesis Creative Systems, Agents and Societies: Theoretical Analysis Tools and Empirical Collaboration Studies on Tuesday the 19th of April 2022 at 16 o'clock in the University of Helsinki Exactum building, Auditorium B123 (Pietari Kalmin katu 5, 1st floor). His opponent is Professor Dan Ventura (Brigham Young University, USA) and custos Professor Hannu Toivonen (University of Helsinki). The defence will be held in English. It is possible to follow the defence as a live stream at https://helsinki.zoom.us/j/61526788115.
The thesis of Simo Linkola is a part of research done in the Department of Computer Science and in the Discovery Research group at the University of Helsinki. His supervisor has been Professor Hannu Toivonen (University of Helsinki).
Creative Systems, Agents and Societies: Theoretical Analysis Tools and Empirical Collaboration Studies
Creativity is a multi-faceted phenomenon that can be observed in diverse individuals and contexts, both natural and artificial. This thesis studies computational creativity, i.e. creativity in machines, which can be broadly categorised as a subfield of artificial intelligence. In particular, the thesis deals with three important perspectives on computational creativity: (1) identifying properties of creative individuals, (2) studying processes that lead to creative outcomes, and (3) observing and analysing social aspects of creativity, e.g. collaboration which may allow the individuals to create something together which they could not do alone.
One of the key interests in computational creativity is how computational entities may exhibit creativity in their own right, implying that the creative entities and their compositions, roles, processes and interactions are potentially different from those encountered in nature. This calls for theoretical analysis methods specifically tailored for artificial creative entities, and carefully controlled empirical experiments and simulations with them. We study both of these aspects. The analysis methods allow us to scrutinise exactly how creativity occurs in artificial entities by providing appropriate conceptual elements and vocabulary, while experiments enable us to test and confirm the effectiveness of different design decisions considering individual artificial creative entities and their interaction with each other.
We propose three novel, domain-general analysis tools for artificial creative entities, i.e. creative systems and creative agents, and collections of them, called creative societies. First, we distinguish several conceptual components relevant for metacreative systems, i.e. systems that can reflect and control their creative behaviour, and discuss how these components are interlinked and affect the system's creativity. Second, we merge elements from sequential decision making in intelligent agents, i.e. Markov Decision Processes, into formal creativity as search model called the Creative Systems Framework, providing a detailed account of various elements which compose the decision-making process of a creative agent. Third, we map elements from an eminent social creativity theory, the Systems View of Creativity, a.k.a. Domain-Individual-Field-Interaction model, into the elements of the Creative Systems Framework and show how creative societies may be analysed formally with it.
Each of the proposed analysis tools provides new ways to analyse creativity in artificial entities. The analysis of metacreative systems assumes an architectural point of view to creativity, which has not been previously addressed in detail. Deconstructing the decision-making process of a creative agent gives us additional means to discuss and understand why or how a creative agent selects certain actions. Lastly, the contributions to the creative societies are the first formal framework for their analysis.
We also investigate in two consecutive case studies collaborator selection in creative societies. In the first study, we focus on what kind of cues, e.g. selfish or altruistic, assist in choosing beneficial collaboration partners when all the agents can observe from their peers are the individually created end products. The second study allows the agents to adjust their aesthetic preferences during the simulations and inspects what emerges from society as a whole. We conclude that selfish cues seem to be more effective in choosing the collaboration partners in our settings and that the society exhibits distinct emergence depending on how much the agents are willing to change their aesthetic preferences.
Availability of the dissertation
An electronic version of the doctoral dissertation is available on the e-thesis site of the University of Helsinki at http://urn.fi/URN:ISBN:978-951-51-8026-1.
Printed copies will be available on request from Simo Linkola: firstname.lastname@example.org.