Stellar Collaboration: The Forging of Evolutionary Spaceships 

ABSTRACT

Co-creative systems, leveraging the collaboration between human designers and computational systems, offer innovative avenues for creative exploration. In this paper, we embark on a journey through the cosmos of co-creation, guided by the principles of speculative design practices. Our focal point is Evolutionary Spaceships, a dynamic co-creative system engineered to engage users in the captivating art of spacecraft design. Employing evolutionary algorithms and interactive interfaces, users embark on an iterative quest to sculpt and refine spacecraft designs, forging a symbiotic bond between human ingenuity and artificial intelligence. Through meticulous evaluation, we unearth the system's prowess in fostering user enjoyment, expressiveness, outcome satisfaction, and ownership. However, amidst the celestial dance of collaboration, opportunities for enhanced synergy beckon, underscoring the ongoing evolution of co-creative technologies. This exploration not only illuminates the frontier of co-creative systems but also propels us towards a future of boundless creative potential in human-AI collaboration. (Abstract written by ChatGPT)

I. CO-CREATIVITY RESEARCH ✨

Co-creative systems are interactive interfaces in which the output is produced through the collaboration of a human designer and a computational system. Research into co-creative systems conducted in collaboration with generative systems, with our companion ChatGPT, led us to recognise domains within which collaborative computational creativity is achieved, namely generative art and design, interactive storytelling, game development, code generation, and more. A few examples of the application of co-creative systems include Versu, No Man's Sky, TabNine, GitHub Copilot, and DeepAI. 

Collaborative technologies are being developed in a variety of creative and technical domains, fostering collaboration between humans and technologies in the creative process. Co-creative systems research mainly focuses on quantifying the applicability of the system, emphasising the evaluation of the creativity of the system's results instead of assessing the creative functionality of the system (Karimi, P., et al. 2019). Collaboration with ChatGPT highlighted certain human-AI collaborative systems that have been explored in generative art and design, such as tools like DeepDream or RunwayML that allow users to produce artistic output. Interactive storytelling is another interesting domain where users can develop dynamic environments based on interaction with AI, for example, AI Dungeon. Generative systems also incorporate code generation, assisting in collaborative coding experiences with systems such as GitHub Copilot. Generative systems like Tableau or Power BI are used in the domain of data visualisation, where users produce interactive and insightful visualisations. 

Exploring co-creative systems through the lens of speculative design practices (Ullstein, C., & Hohendanner, M., 2020), this research employs science fiction as an instrument to foster an immersive experience of co-creation. An enhanced interactive experience is gained when the user is provided with the agency of world-building through the engaging speculative narrative that our co-creative system is based on. ChatGPT advised in the composition of the introductory narrative composed in our collaborative system. Furthermore, in the framework of our design approach, we discovered that other accessible generative systems, such as GitHub Copilot, TabNine, and DeepAI, can be implemented to materialise our system in its entirety. 

As co-creative systems are intended to support their human users, facilitate their creativity, foster inspiration, and stimulate continuous interaction with the interface (Karimi, P., 2019), our co-creative system, Evolutionary Spaceships, invites its collaborators to design spaceships with the components that they desire in their spaceship design by interacting with its interface. It is an intuitive system that allows real-time collaboration and continuous engagement with its users. ChatGPT and GitHub Copilot assisted in developing a user-centred interface as well as determining essential parameters in the design of the spaceship that otherwise we would not have considered. We believe that the development of the project has been a co-creative process in its own essence. 

According to ChatGPT, operating on parameters such as autonomy, variety, adaptability, and exploration and exploitation makes evolutionary algorithms an effective generative system that aligns with this research. Evolutionary algorithms as generative systems, in the framework of artificial intelligence and computational creativity, refer to systems that can produce novel and varied outputs autonomously, as they are capable of generating diverse solutions to a problem through an iterative and evolutionary process. Therefore, the fundamental generative system that this research is based on is the application of an evolutionary algorithm, which essentially enables the generation of uniqueness in each spaceship design. 

In the context of this research on playful spaceship design, evolutionary algorithms are implemented to generate the wide range of potential designs adapted over generations based on fitness evaluations and generate novel and optimised solutions. Its characteristic of autonomy enables the generation of novel design iterations without direct human intervention as it relies on the principles inspired by natural selection, including mutation and selection. The feature of variety enables evolutionary algorithms to inherently introduce diversity into the population of solutions through genetic operators such as mutation and crossover, which ensures exploration of diversity in design outputs. Adaptability allows the spaceship's design to evolve based on its fitness function, favouring the solutions that are closest to its representation. Lastly, the characteristic of exploration and exploitation facilitates the equilibrium between refining the already-generated designs and providing novel spaceships. In summary, evolutionary algorithms share characteristics with generative systems, making them valuable team members for generating diverse and adaptive solutions to complex problems and assisting in the creation of evolutionary spaceships. 

II. DESCRIPTION OF THE CO-CREATIVE SYSTEM 🚀

The interface of Evolutionary Spaceships’ is designed as a game website to enhance the experience of interaction. The co-creative system consists of 13 components that the algorithm can create spaceships with. Each component has it's own function within a ship, and the existence of these components within a ship determine how good a ship performs in certain factors. The user can input the parameters in the way they want their spaceship to perform, as the correlation between the parameters will define the overall design of the spacecraft. 

Evolutionary spaceships are developed using evolutionary algorithms, and the fitness function of the evolutionary algorithm defines how well a spaceship performs. When the user alters the wanted parameters of the ship, the user is altering the fitness function of the algorithm. The parameters can be modified from 'more' to 'less', such as increasing or decreasing the number of humans on board. Additionally, the users can reset the parameters and even download their designs in a png format after they have designed a spaceship to their needs. The panel on the right of the screen displays information about the best performing ship in the current generation, based on the established wants of the user. This panel can be used by the user to perceive how well their ship is performing and improving.

The properties of the ship that define the fitness of the ship in our co-creative system are as follows:  

1. Spaceship Size: refers to the diagonal size of the spaceship generated.  

2. Weight: refers to the cumulative weight of the components within the craft. 

3. Connectedness: refers to the number of loose ends available for more components to attach to the spaceship. 

4. Power generation: refers to the amount of power that is being produced by the solar panels in the spaceship. 

5. Power usage: refers to the power that is being used by the components of the spaceship. 

6. Power efficiency: refers to the distribution of power throughout the spaceship.  

7. Oxygen Generation: refers to the amount of oxygen produced. 

8. Humans: refers to the amount of humans aboard, based on how many the ship can support using its components. 

9. Science: refers to the amount of scientific knowledge generated by the spaceship consisting of components such as the optical instruments, observatory and laboratory.  

10. Novelty: refers to the novelty of the components in the ship, compared to the other ships in the generation.

Evolutionary spaceships is a form of system that assesses the significance of a component required to build a desirable spaceship. Of which, the parameters are seen as a fundamental basis for the evolutionary algorithm to produce efficient designs based on the given parameters. Furthermore, with each iteration, the system develops unique configurations. It takes the previous generation's best performing ships, based on the users wants and needs, and tries to make a new generation by crossing over spaceships and applying mutations. It is a highly involved and interactive system; if the design does not suit the user's preferences, the user may modify the parameters while it is running to nudge it in the desired direction. For instance, if the user adjusts the power generation parameter to increase the scale of the spaceship, the design evolves to require more power, leading to an increase in solar panels. If the power must be distributed evenly across all components, the rest of the ship will evolve to accommodate the rise in power. 

III. EVALUATION 📃

In the framework of assessing the process of co-creation, three domains of creative processes are defined to determine the user-interface interaction: generate, evaluate, and define. Depending on the creative process, the co-creative AI can take on the roles of generator, evaluator, or definer (Rezwana, J., and Lou Maher, M., 2023). Evolutionary Spaceships co-creative system plays the role of generator, whereas the user is alternating between the evaluator and definer roles. As the user defines the spaceship design based on the modifications in the parameters of the components and evaluates whether the desired result is achieved, it offers creative freedom to the user and a sense of greater control over the outcome. In the generative process, the co-creative system produces innovative design compositions depending on the guidelines defined by the user. 

Evolutionary spaceships cater to anyone who is an astronomy enthusiast and is interested in creating their own spaceship design(s); individuals who are inspired by science fiction narratives and are curious to investigate the anatomy of a spacecraft by assessing the parameters of spaceship components and their dependency, relationship, correlations, and assessment in system configuration. The special emphasis on evaluation is specifically on the process of co-creation of our system, which is assessed by the amount of time the user spends engaging with our system, the designs that the user generates, etc. After interacting with the co-creative system, the participants are invited to share and rate their experiences through a questionnaire analysis. The user analysis is conducted by participants who have backgrounds in animation, computer science, linguistics, social science, and more.  

The questionnaire evaluation is based on creativity matrices as discussed by A. Kantosalo and S. Riihiaho (2019) in ‘Experience evaluations for human-computer co-creative processes: planning and conducting an evaluation in practice’. The first metric refers to fun, which is generally used for evaluating user interfaces for children. The second and more relevant question for our interface is the Enjoyment metric, which is assessed in the questionnaire by asking how enjoyable the experience of the user with this interface was and whether the user would engage with it again, serving as a measure of enjoyment. The other metrics are Expressiveness, Outcome satisfaction (Cherry and Latulipe, 2014), and Collaboration (Jordanous, 2012). In which Expressiveness refers to the extent to which users can express their creativity during the creative process. Outcome Satisfaction suggests how content the user is with the end result of their collaborative experience with the algorithm. And the Collaboration metric refers to the mutual influence, sharing, and feedback with the algorithm (Jordanous, 2012). And the last metric is Ownership, which is based on the research of Compton and Mateas (2015), emphasising the notion of ownership that the user experiences during the creative process. 

Through questionnaire analysis based on the aforementioned metrices, users have indicated that Evolutionary Spaceships successfully approach the Enjoyment, Expressiveness, Outcome Satisfaction and Ownership metrics. The evaluation has shown that Evolutionary spaceships is limited in Collaboration metric in its current stage. Users have communicated that the co-creative system is engaging, with the approximate length of engagement being 10-15 minutes, and the user interface is intuitive. The perceived creativity of the generated designs is 50:50.  

Participants highly appreciated the saving of their spaceship design, referring to Ownership and the Outcome Satisfaction metric. Additionally, participants have communicated the lack of clarity in what the 'Reset’ and ‘Run Algorithm’ buttons did. There is a remark on the user interface referring to the lack of clarity in the conclusion of the algorithm, as the user assumed that altering all parameters for the spaceship design would bring the algorithm to conclude with the optimal design. The users stated that the evolutionary process of the spaceship designs is interesting to observe when the user modifies the parameters, and the algorithm responds in real-time to the user’s alterations. 

REFERENCES 

1. Karimi, P., Maher, M. L., Davis, N., & Grace, K. (2019). Deep learning in a computational model for conceptual shifts in a co-creative design system. arXiv preprint arXiv:1906.10188. 

2. Ullstein, C., & Hohendanner, M. (2020). Exploration of the future of co-creative systems through collaborative speculative design practices. In A Workshop on Human-Computer Co-Creativity (ICCC’20). Association for Computational Creativity. 

3. Jeba Rezwana and Mary Lou Maher. 2023. Designing Creative AI Partners with COFI: A Framework for Modeling Interaction in Human-AI Co-Creative Systems. ACM Trans. Comput.-Hum. Interact. 30, 5, Article 67 (October 2023), 28 pages. https://doi.org/10.1145/3519026.  

4. Kantosalo, A., & Riihiaho, S. (2019). Experience evaluations for human–computer co-creative processes–planning and conducting an evaluation in practice. Connection Science, 31(1), 60-81. 

5. Cherry, E., & Latulipe, C. (2014). Quantifying the creativity support of digital tools through the creativity support index. ACM Transactions on Computer-Human Interaction (TOCHI), 21(4), 1-25. 

6. Jordanous, A. (2012). A standardised procedure for evaluating creative systems: Computational creativity evaluation based on what it is to be creative. Cognitive Computation, 4(3), 246–279. 

7. Compton, K., & Mateas, M. (2015, June 29–July 2). Casual creators. Proceedings of the sixth international conference on computational creativity, Park City, Utah, pp. 228–235. 

8. Davis, N. Co-creative AI, https://sites.google.com/view/cocreativeai/co-creative-ai-research/evaluating-co-creative-ai.  

9. Station Assembly Elements, https://www.nasa.gov/international-space-station/international-space-station-assembly-elements/. 

10. Anatomy of a Spacecraft, https://www.esa.int/Science_Exploration/Space_Science/Anatomy_of_a_spacecraft.