13h05 Hannu Toivonen : Creativity and AI
Computational Creativity is the study of creative systems. In this talk, Prof Toivonen will discuss the relationship between AI and creativity. He will show different ways in which machine learning can be used for creativity, and he also highlights key differences between typical AI tasks and creativity. He introduces concepts that can be used to describe or analyse creative programs beyond the creativity of their products.

Music composition aided by symbolic AI

Music has a deep connection to the human condition.  Composing music is like composing poetry: it is difficult but when it succeeds it touches us deeply.  Is there a way that we can use computers to help composers?  We would like to amplify a composer’s creativity and not replace it; we want the musical ideas to come from the composer.  In the Harmoniser project, we propose to build a toolbased on constraint programming that targets Western tonal music.  The composer gives musical ideas and the tool uses them to construct a musical workin which the structure of tonal music is respected.  Constraint programming is a form of symbolic artificial intelligence that finds solutions to complexcombinatorial problems specified as logical relations.  We focus on Western tonal music because it is highly combinatorial and has been studied deeply by music theorists.  To make our tool practical, we define a four-layer iterative process for music composition that is relevant for many composers.  The process is suitablefor many shorter musical works, such as for example the songs of popularmusic.  This is ongoing work: we give examples to show how the tool works and we explain what remains to be done.  The final tool will be a plug-in for a Digital Audio Workstation.  We hope that the Harmoniser project will lead to an improved use of symbolic AI for music composition.

Haunted Comics: The inderterminate craft of prompting.

This presentation explores how generative AI reshapes the early stages of comicbook creation, focusing on the interplay between visible outcomes and unrealizedpossibilities embedded in production materials like sketches and layouts. Through a case study of Fastwalkers—a synthetic comic co-created with AI in 2020—the research investigates how algorithmic feature spaces function as conceptual terrains for iterative prototyping. By training models on domain-specific datasets such as Danbooru, the creators engaged in an evolving dialogue between human intent and machine-driven variation. This iterative, AI-assistedprocess revealed latent aesthetics and speculative narrative structures, complicating traditional notions of authorship, intentionality, and materiality. The presentation frames these dynamics through the lenses of explainablecomputational creativity, arguing that AI introduces a spectral dimension to comics production. Ultimately, it suggests that generative systems amplify the indeterminacies of multimodal craft, prompting a redefinition of what constitutesa comics object in the era of synthetic media.

Generating Games Via Evolution and Language Models

The automated creation of novel games represents a frontier in proceduralcontent generation and artificial creativity. However, it is hindered by complexchallenges such as representing game mechanics, navigating a massive design space, and meaningfully assessing originality. This work leverages the rich and diverse Ludii platform, home to over 1400 formalized board games, and combines state-of-the-art language modeling with evolutionary algorithms to generate innovative new games. Our system intelligently edits and recombines existing games, demonstrating the potential for uncovering unexplored areas of the game design landscape and pushing the boundaries of what automatedsystems can create.