A Workshop on Computational Creativity, Embodiment, and Problem Solving

Capturing creativity formally, computationally, and cognitively remains a frontier in the cognitive sciences and in artificial intelligence. One popular approach to the formal capture of cognitive processes is the “problem solving” framework first developed by Allen Newell and Herbert Simon, and one popular approach to studying creativity within that framework is the use of “insight problems.” An intuitive conception of insight is the kind of task in which an agent aims to find a solution, gets stuck or seems not to make progress, and then discovers the solution all-at-once in a “Eureka!” or “a-ha!” moment. In this formal conception of creativity, researchers presuppose that the agent works within a “problem space,” “conceptual space,” or “design space” containing the agent’s symbolic representation of the problem. The creative “insight” is then construed either as a change in the representation of the problem space, or a change in the heuristics for traversing it.

The theme of this interdisciplinary workshop is to gather approaches which both work within and challenge these more traditional frameworks of problem solving and creativity. Talks by Sune Vork Steffensen (University of Southern Denmark) and Wendy Ross (London Metropolitan University)  will present contemporary empirical results in the problem solving literature from a cognitive psychology perspective. A talk by Sotirios Kotsopoulos (MIT/National Technical University of Athens)  will present a computational framework for understanding visual creativity in design which avoids explicit commitment to representations, while work by Ana-Maria Olteteanu (Constructor University Bremen) will investigate visual and verbal creativity in a series of structured games with trained A.I. models drawing from the classic Remotes Associates Task in the creativity literature. Finally, a talk by Benjamin Angerer (University of Bielfeld) will present results from a qualitative experimental task involving paper folding, exploring how the representations involved in the traversal of a “problem space” might get constructed in the first place.

The event will be held as part of the Radical Embodiment series at the University of Southern Denmark. This series will run from 18 June to 21 June 2025, with this workshop occurring on the 21st. Take a look at some of those other events at CFP: Dimensions of Radical Embodiment 3 - PhilEvents

Find the abstracts below

Abstracts of Talks 

Benjamin Angerer (University of Bielfeld) --"Where Do Problem Spaces Come From? On Metaphors and Representational Change"

The challenges of problem solving do not lie exclusively in how to perform heuristic search, but begin with how we understand a given task: how we cognitively represent the task domain and its components can determine how quickly we are able to progress towards a solution, whether advanced strategies can be discovered, or even whether a solution can be found at all. While this challenge of constructing and modifying representations has been recognised early in problem solving research, it has largely been sidestepped by focusing on simple, well-defined problems whose representation is almost entirely determined by the task instructions. Thus, the established theory of problem solving as heuristic search in problem spaces has little to say about it.

In this talk, I will present a study designed to explore this issue, with the main challenge being to find and refine an adequate problem representation. In this exploratory case study, I investigated how pairs of participants familiarise themselves with a complex spatial transformation task in the domain of iterative mental paper folding over the course of several days. Participants are required to understand the geometry of edges that arise when they repeatedly mentally fold a sheet of paper in alternating directions without the use of external cues. Faced with the difficulty of handling increasingly complex folds in the face of limited cognitive capacity, participants are forced to search for ways to represent folds more efficiently. In a qualitative analysis of video recordings of participants' behaviour, the development of their conceptualisation of the task domain was traced over the course of the study, focusing in particular on their use of gestures and the spontaneous occurrence and use of metaphors in the construction of new representations. Based on these observations, I will conclude the talk with some theoretical speculations on the roles of metaphor and cognitive capacity in representational change.

Sotirios Kotsopoulos (MIT/National Technical University of Athens)-- "Design without Representation"

Shapes are perceived unanalyzed, without rigid representation of their parts. They do not comply with standard symbolic knowledge representation criteria; they are treated and judged by appearance. Resolving the relationship of parts to parts and parts to wholes has a constructive role in perception and design. This paper presents a computational account of part–whole figuration in design. To this end, shape rules are used to show how a shape is seen, and shape decompositions having structures of topologies and Boolean algebras reveal alternative structures for parts. Four examples of shape computation are presented. Topologies demonstrate the relationships of wholes, parts, and subparts, in the computations enabling the comparison and relativization of structures, and lattice diagrams are used to present their order. Retrospectively, the topologies help to recall the generative history and establish computational continuity. When the parts are modified to recognize emergent squares locally, other emergent shapes are highlighted globally as the topology is re-adjusted. Two types of emergence are identified: local and global. Seeing the local parts modifies how we analyze the global whole, and thus, a local observation yields a global order.

Ana-Maria Olteteanu (Constructor University Bremen)  -- "How Not to Be A Lonely A.I.: Exploring Linguistic and Visual Problem Solving Through Games"

Human creativity is inherently multimodal, with insights often emerging in visual as well as linguistic forms. To explore structured creativity, we introduce linguistic and visual adaptations of the Remote Associates Test (RAT). Future AI-human collaboration should be both creative and cooperative, yet understanding and supporting these interactions requires systematic study. Games, which blend intelligence, creativity, and social reasoning, provide a rich testbed for this research. We propose a series of games incorporating both linguistic and visual elements to investigate multimodal creativity in a structured way, offering new insights into how AI and humans co-create and anticipate each other’s moves

Wendy Ross (London Metropolitan University) -- Title and Abstract TBD 

Sune Vork Steffensen (University of Southern Denmark) -- Title and Abstract TBD