Software Design Patterns of Computational Creativity (PDF)

Software Design Patterns of Computational Creativity: A Systematic Mapping Study Porter Glines, Isaac Griffith, and Paul M. Bodily Department of Computer Science Idaho State University 921 S. 8th Ave, Pocatello, ID 83209 USA [email protected], [email protected], [email protected] Abstract Software design patterns can be helpful in describing the ar- chitecture of a system. Our objective is to obtain a broad overview of the current state-of-the-art of software design patterns used in Computational Creative (CC) systems. We conducted a systematic mapping study using manual and snowballing search techniques. Only 7 primary studies are identified in the CC community that explicitly mention the use of design patterns. Within these primary studies, 14 de- sign patterns are mentioned, 12 of which are user-interaction design patterns rather than software design patterns describ- ing the architecture of the system. The small number of pri- mary studies indicates a gap in CC literature regarding the use of software design patterns in CC systems and motivates the need for research to identify software design patterns specific to CC systems. Introduction As computationally creative (CC) systems strive to become more creative, they tend to require an increasing number of behaviors. CC systems strive to have behaviors such as self- evaluation, a knowledge-base, and ultimately an understand- ing of the world — all of which are added to push creative systems further along in the spectrum of creative systems (Ventura 2016; 2017; Glines, Biggs, and Bodily 2020). Each desired system behavior adds another design chal- lenge when creating a CC system. Each design challenge comes with an opportunity to introduce elements into the system that are not easily maintainable, reusable, or un- derstandable. To avoid common design challenges, system builders can use software design patterns to implement com- mon system behaviors. Software design patterns, which we will refer to as “de- sign patterns”, are general, reusable solutions to commonly occurring problems in software development. In 1994, a team of researchers referred to as the “Gang of Four” iden- tified the foundational 23 design patterns in their work “De- sign Patterns: Elements of Reusable Object-Oriented Soft- ware” that has since become a standard part of collegiate computer science curriculum (Gamma et al. 1994). Since then, many more design patterns have been identified like model-view-controller, delegation, blackboard, etc. Practi- tioners use design patterns with the goal of improving main- tainability, scalability, reusability, understandability, among other quality attributes (Zhu 2009). However, there is debate as to the effectiveness of design patterns. Some studies have shown that design patterns can negatively affect quality at- tributes, but conclude that more research is needed (Khomh and Gueheneuc 2008). Though design patterns may nega- tively affect quality, they have shown to improve maintain- ability (Zhang and Budgen 2012). Overall, it appears that design patterns should be applied to the problem they solve while considering the consequences they can bring. The potential benefits of design patterns motivates the de- sire to identify design patterns for CC systems. Ventura (2017) identifies a general architectural pattern for building a CC system for any arbitrary domain — see Figure 1. In the architectural pattern, a system builder first chooses a do- main in which the system operates. Then internal and ex- ternal representations of artifacts are designed, a knowledge base of the domain is collected, a conceptualization or model is chosen to generate artifacts, an aesthetic is chosen to in- fluence how the system learns, and finally an evaluator for artifacts is designed. Note that the architectural pattern for CC systems describes easily separated components that feed into an overarching architecture: a knowledge base, genera- tor (model), aesthetic, and evaluator. Ventura argues that it is worth spending “significant” time searching for existing implementations for these components rather than building them from scratch. This argument for more reuse in CC sys- tems further motivates the desire for maintainable and easily communicable code underlying these systems. The goal of this paper is to provide a broad overview to re- searchers and practitioners of the state-of-the-art in CC sys- tems with regards to the software design patterns used to build them. To provide this broad overview, a systematic mapping study will be performed as described by Peterson et al. (2008; 2015). Our systematic mapping study consoli- dates the use of software design patterns from 30 papers that present CC systems. This paper aims to provide insights to researchers and practitioners regarding how and when de- sign patterns are used in CC systems to facilitate building systems that are easier to maintain, reuse, and understand. There are currently no studies providing an overview of design patterns used in CC systems. This paper aims to fill this gap in the literature. Proceedings of the 12th International Conference on Computational Creativity (ICCC ’21) ISBN: 978-989-54160-3-5 218  Table 1: Research questions along with their rationale. Research Question Rationale RQ1 What design patterns are mentioned in CC literature? The aim is to explore works published in the field of CC, identify what design patterns are commonly used, and in- form researchers what those patterns are. RQ2 Are there software design patterns devised specifically for CC systems? The aim is to inform researchers, especially those new to the field, of field-specific design patterns. RQ3 Of the CC literature that mentions design patterns, what conferences and journals are represented? The aim is to inform researchers when and where these papers are being published to better direct research re- sources. Table 2: Inclusion and exclusion criteria for selection of pri- mary studies. Inclusion criteria • English language articles. • Peer-reviewed conferences or journals articles. • Articles published between January 2010 and Septem- ber 2020. • Studies that relate to the field of CC. • Studies that mention a keyword identified as relating to design patterns, whether high level or otherwise. Exclusion criteria • Studies that do not mention a design pattern of any kind. • Articles that present frameworks for building CC sys- tems but do not identify or name a design pattern. Methods Systematic mapping studies are used to provide a broad but rigorous review of the literature with the goal of revealing gaps in the literature. The systematic mapping study in this paper is conducted as described by Peterson et al. (2008; 2015). Planning of the mapping study is described, includ- ing identified research questions and search procedures. The process by which papers are screened is described in the in- clusion/exclusion criteria. Then, data extraction procedures are described. Extracted data is kept in a database to be queried for later analysis. Planning Stage Identified research questions and their motivations are shown in Table 1. For clarity, we define CC literature as any literature containing concepts related to CC as defined by Colton and Wiggins (2012): [CC is] the philosophy, science and engineering of computational systems which, by taking on particular responsibilities, exhibit behaviours that unbiased ob- servers would deem to be creative. Figure 1: A diagram from Ventura’s (2017) paper: “How to build a CC system”. The paper describes a general approach to constructing a CC system; components described in the approach could have design patterns associated with them. Search Strategy The CC community has yet to estab- lish a formal database of published literature facilitating ad- vanced keyword search. Therefore, a manual search using the search term “design pattern” was performed on the pro- ceedings of the International Conference on Computational Creativity (ICCC). The search term was changed to just “pattern” after an initial search yielded few papers. Once a starting set of primary studies was selected, a forward snowballing search was performed as described in Wohlin’s (2014) guidelines. Selection of Primary Studies Identified papers are then systematically marked to be included or excluded from the study. The criteria for including or excluding a paper is shown in Table 2. This criteria is first applied to study ti- tles and abstracts; then to the introduction, conclusion, and methods sections; and finally, full papers are read to evaluate whether the study should be included as a primary study. Identified Keywords To address RQ1 and RQ2, we use additional keywords to help detect the mention of design patterns. The 23 design patterns presented by the Gang of Four are included as keywords, e.g., “factory”, “flyweight”, and “mediator”, as well as the three categories “creational Proceedings of the 12th International Conference on Computational Creativity (ICCC ’21) ISBN: 978-989-54160-3-5 219  design pattern”, “structural design pattern”, and “behavioral design pattern”. Additionally, a keyword is included for the “blackboard” design pattern. Generic keywords “pattern”, “design pattern”, and “software design pattern” are included. Data Extraction Papers identified as primary studies have the following in- formation extracted from them: title, year published, confer- ence or journal, and pattern(s) mentioned. This information is extracted and stored in a spread sheet1 for later analysis. Results There are a total of 458 papers within the eleven ICCC conference proceedings from 2010 to 2020. Within the conference proceedings, a pilot search was performed with the search term: “design pattern”. This preliminary search yielded only four results. Thus the search was expanded by using the more generic term: “pattern”. This new search yielded 195 papers. Of these resulting studies, six satisfied the inclusion/exclusion criteria and were identified as pri- mary studies and as a starting set to perform a snowballing search. The forward snowballing search yielded an additional 86 new papers from the starting set of six studies. Of the newly found papers, one satisfied the inclusion/exclusion criteria. The combined search results yield the following seven primary studies: (Compton and Mateas 2015; Concepci´on, Gerv´as, and M´endez 2019; Goel 2015; Kreminski et al. 2020; Petrovskaya, Deterding, and Colton 2020; Abdellahi, Maher, and Siddique 2020; Chang and Ackerman 2020). Findings Regarding Research Questions RQ1: What design patterns are mentioned in CC liter- ature? There are 14 design patterns mentioned in the pri- mary studies: 1. Strategy design pattern 2. Instant feedback design pattern 3. Mutant shopping design pattern 4. Chorus line design pattern 5. Simulation and approximating feedback design pattern 6. Entertaining evaluations design pattern 7. No blank canvas design pattern 8. Limiting actions to encourage exploration design pattern 9. Modifying the meaningful design pattern 10. Saving and sharing design pattern 11. Hosted communities design pattern 12. Modding, hacking, teaching design patterns 13. Turn-taking pattern 14. Biologically inspired design patterns 1Link to data: https://tinyurl.com/y4w7najp Figure 2: Number of primary studies published per year that mention design patterns and are related to CC. We can see that 2020 sees the most studies published; however, there are only seven CC studies mentioning design patterns. With so little data, it is impossible to state whether there is a trend. Only one design pattern mentioned is a traditional “Gang of Four” design pattern, namely the strategy pattern. The strategy pattern is used in a story generation system to select and apply heuristics used to weave generated story plot-lines together (Concepci´on, Gerv´as, and M´endez 2019). Design patterns 2 to 13 are user-interaction design patterns, e.g., the instant feedback pattern where users observe an arti- fact, make a change, and observe the result of the change at a glance. We note that user-interaction design patterns in- form system architects of how users will interact with the system. However, they do not inform system architects of how objects and classes interact within a codebase, i.e., how a codebase itself is designed, like a software design pat- tern would. Biologically inspired design patterns describe generic patterns by which biology is used as inspiration to solve a problem. Like user-interaction design patterns, bio- logically inspired design patterns do not inform system ar- chitects of how objects and classes interact. RQ2: Are there software design patterns devised specif- ically for CC systems? The 11 user-interaction design patterns identified by Compton and Mateas (2015) are pre- sented in the context of CC but are not specific to CC. Out of the seven primary studies, there are no identified design patterns that are specific to CC. RQ3: Of the CC literature that mentions design pat- terns, what conferences and journals are represented? The one primary study found during the snowballing search came from the 2020 AAAI Conference on Artificial Intel- ligence and Interactive Digital Entertainment (AIIDE). The remaining six primary studies come from various years of the ICCC. As shown in Figure 2, five out of the seven pri- mary studies were published within the last two years (2019 and 2020). Proceedings of the 12th International Conference on Computational Creativity (ICCC ’21) ISBN: 978-989-54160-3-5 220  Discussion With only seven primary studies, the results indicate that de- sign patterns are not often mentioned in CC studies. In other words, the results indicate a gap in CC literature in regards to the use of design patterns in CC systems. Perhaps unsurprisingly, the design patterns most men- tioned in CC literature are user-interaction design patterns. These patterns are particularly relevant to the CC commu- nity in helping design co-creative systems. While user- interaction design patterns do not describe how to design the system codebase, like a more traditional “Gang of Four” design pattern would, they do provide guidance for build- ing the overarching system architecture. Only three papers (Chang and Ackerman 2020; Abdellahi, Maher, and Sid- dique 2020; Kreminski et al. 2020) present systems where a user-interaction design pattern is used, indicating that many papers presenting co-creative systems do not disclose the use of a user-interaction design pattern. The lack of results could indicate that the CC community is unaware that they are using design patterns. In this case, CC researchers would be less likely to mention them. Ulti- mately, our results indicate that we cannot gain a true under- standing of the pervasiveness of design patterns in CC with- out reviewing the actual CC systems. This suggests the need for an empirical study on CC systems as future work. An ad- ditional avenue of research would be the evaluation of Ven- tura’s architectural pattern for building a CC system — iden- tifying common approaches (or patterns) for implementing each of the components. The suggested future work would achieve an understanding of the patterns used in building CC systems and furthers the idea that Ventura identified an architectural pattern. We see design patterns in the CC community as an op- portunity for pedagogical benefit — a way to bring in new CC community members and programmers by facilitating systems that are easier to reuse and understand. Conclusion This systematic mapping study aimed to provide a broad overview of the use of design patterns in papers presenting CC systems. However, the results show that there is a lack of discussion in CC literature regarding design patterns. This motivates future work to identify design patterns in CC sys- tems to gain a full understanding of how design patterns are used in building CC systems. The results also motivates the need to open a conversation on design patterns in the CC community to create CC systems that are easier to main- tain, reuse, and understand, facilitating more collaborative research. References Abdellahi, S.; Maher, M. L.; and Siddique, S. 2020. Arny: A co-creative system design based on emotional feedback. In 11th International Conference on Computational Creativity. Chang, J., and Ackerman, M. 2020. A climate change edu- cational creator. In 11th International Conference on Com- putational Creativity. Colton, S.; Wiggins, G. A.; et al. 2012. Computational creativity: The final frontier? In 20th European Conference on Artificial Intelligence, volume 12, 21–26. Compton, K., and Mateas, M. 2015. Casual creators. In Proceedings of the Sixth International Conference on Com- putational Creativity, 228–235. Concepci´on, E.; Gerv´as, P.; and M´endez, G. 2019. Evolv- ing the ines story generation system: From single to multiple plot lines. In Proceedings of the 10th International Confer- ence on Computational Creativity, 220–227. Gamma, E.; Helm, R.; Johnson, R.; and Vlissides, J. M. 1994. Design Patterns: Elements of Reusable Object- Oriented Software. Glines, P.; Biggs, B.; and Bodily, P. M. 2020. A leap of creativity: From systems that generalize to systems that fil- ter. In Proceedings of the 11th International Conference on Computational Creativity, 297–302. Goel, A. K. 2015. Is biologically inspired invention differ- ent? In Proceedings of the Sixth International Conference on Computational Creativity, 47–54. Khomh, F., and Gueheneuc, Y. 2008. Do design patterns impact software quality positively? In 2008 12th European Conference on Software Maintenance and Reengineering, 274–278. Kreminski, M.; Dickinson, M.; Osborn, J.; Summerville, A.; Mateas, M.; and Wardrip-Fruin, N. 2020. Germinate: A mixed-initiative casual creator for rhetorical games. In Pro- ceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, volume 16, 102–108. Petersen, K.; Feldt, R.; Mujtaba, S.; and Mattsson, M. 2008. Systematic mapping studies in software engineering. Petersen, K.; Vakkalanka, S.; and Kuzniarz, L. 2015. Guide- lines for conducting systematic mapping studies in software engineering: An update. Information and Software Technol- ogy 64:1–18. Petrovskaya, E.; Deterding, C. S.; and Colton, S. 2020. Ca- sual creators in the wild: A typology of commercial gener- ative creativity support tools. In 11th International Confer- ence on Computational Creativity. Ventura, D. 2016. Mere generation: Essential barometer or dated concept? In Proceedings of the Seventh International Conference on Computational Creativity, 17–24. Ventura, D. 2017. How to Build a CC System. In Eighth In- ternational Conference on Computational Creativity, 253– 260. Wohlin, C. 2014. Guidelines for snowballing in systematic literature studies and a replication in software engineering. In Proceedings of the 18th international conference on eval- uation and assessment in software engineering, 1–10. Zhang, C., and Budgen, D. 2012. What do we know about the effectiveness of software design patterns? IEEE Trans- actions on Software Engineering 38(5):1213–1231. Zhu, Z. 2009. Study and application of patterns in software reuse. In International Conference on Control, Automation and Systems Engineering, 550–553. Proceedings of the 12th International Conference on Computational Creativity (ICCC ’21) ISBN: 978-989-54160-3-5 221