An analysis of the structure and evolution of the scientific collaboration network of computer intelligence in games
Introduction
Games have long been seen as an ideal test-bed for the study of artificial intelligence (AI) [1]. In the past, much of the academic work on AI and games focused on traditional board games, such as checkers and chess, and these games were in turn used to check the goodness/efficacy of AI techniques. Although this traditional focus is still being used, there is now also a growing body of work on applying computational intelligence (CI) to game development (including video games) with the aim of improving the game itself. CI comprises a collection of nature-inspired methods–such as evolutionary algorithms, artificial neural networks and fuzzy logic–for the resolution of complex problems, and can be applied to optimize game development from different perspectives: for instance, from the user’s point of view, current players demand, not only outstanding graphics, but also other non-visual features that CI can help to significantly improve such as intelligent/adaptive behavior of non-player characters (NPC), an interesting narrative, or more attractive contents (in the form of, for instance, levels, maps, weapons, armor, terrains, music or even the game rules themselves) and many more. In addition, from the industry’s point of view, CI is starting to be viewed as a mechanism to improve the process of game development as well as a tool to extend the commercial life of games. Regarding the first issue, CI is being used to automatically generate game content, such as terrains, maps or even music [2], whereby the automation of this process would lead to a reduction in the production costs as the content would not be entirely hand-created and thus human intervention might be reduced. Moreover, CI techniques applied to achieve this objective in one particular game might presumably be adapted to a similar game belonging to the same genre and this would also reduce the expense of implementing new games. With regard the latter issue, the capacity of CI to produce automated NPC behaviors as well as multimedia content is interesting (and promising) as, in theory, it would be natural to extend a successful (i.e., a best seller) game by procedurally creating new content. This new content in turn would produce a new (commercial) version of the game (the recursive application of this idea sculptures the concept of infinite game) which represents a way of extending the earnings at a minimum cost. Moreover, the industry perceives CI as a tool from which it is possible to gain additional benefits as it can be used to extract interesting players’ data that can be employed, for instance, to evaluate player satisfaction (e.g., via emotional analysis) or to construct specific games according to the profile of the users.
However, the application of CI techniques to game development goes beyond the idea of just generating pure entertainment and also covers the purpose of solving hard (and, in many cases, socially accepted) problems; this is for instance, the case of the so-called “serious games”. These games prioritize additional objectives than of being simply fun such as training, learning, evaluation, management, marketing, or advertising among others. Serious games have already been implemented in a wide number of heterogeneous areas such as education, health care, politics, or defense [3] and CI has already been applied successfully to the construction of serious games—see Ref. [4].
In general, one can be certain that CI will have a great impact on game development and for this reason the application of CI to games is currently attracting increasing interest from both industry and academia. The consequence of this interest is that this field is a burning issue that seems to have undergone a huge transformation in recent years. To study this transformation, we have embarked on an analysis of the social network of researchers working on CI in games (CIG). Social network analysis–see Ref. [5] for a survey–although an old discipline, has recently been given new impetus and tools from the field of complex networks. A review of complex networks is presented in Ref. [6]. The study of all kinds of networks has been an extremely active research topic in the recent past, following the introduction of models for power-law [7] and scale-free networks [8], which, in turn, has induced the study of many different phenomena in this new light. The particular class of network the study of which is addressed in this work is a co-authorship network [9]: nodes in these networks are paper authors, joined by edges if they have written at least one paper together. Note that the co-authorship data is a social network since collaborating in a research study usually requires that the coauthors become personally acquainted. Thus, studying the ties, their structure, and their evolution enables a better understanding of the factors that shape scientific collaboration. In this sense, this network analysis has been done in many other fields, like mathematics [10], [11], evolutionary computation (EC) [12], [13], [14], computer-supported cooperative work (CSCW) [15], social sciences [16] and physics, biomedicine and computer science [17], [18], to cite some examples.
This work analyzes the collaboration network of the CIG community on three different scales–macroscopic, mesoscopic and microscopic–aiming to discern the status and distinctive features of the community, the rules governing scientific cooperation within it, and the direction in which it is heading as well as trying to identify which seems to be some of the most active research topics in the area.
Section snippets
Materials and methods
The bibliographical data used for the construction of the scientific-collaboration network has been obtained from the Digital Bibliography and Library Project (DBLP1) bibliography server. This database, maintained by the DBLP Team at the University of Trier, provides bibliographic information on major computer science journals and proceedings. It lists more than 1.9 million publications and several thousand computer scientists (as of March 2012). Besides this wide
Analysis of the collaboration network of CI in games
Using the data gathered from the DBLP database we can represent the collaboration network as a weighted graph where each vertex represents an author and each edge indicates that authors and have co-authored (possibly with other additional co-authors) at least one paper. The number of collaborations between two authors is captured by . More precisely, is a function with signature such that is the number of papers authors and have co-authored.
Conclusion and future work
This work has provided the first steps towards understanding the dynamics of the Computational Intelligence in games (CIG) co-authorship network and its modeling, using an individual-based model. CIG is an active and vibrant field, attracting new researchers and generating new papers, having a stable growth in the number of authors, which had an accelerated growth up to the early/mid-2000s. The number of published papers per year has been increasing since 2005, sustaining the steady growth of
Acknowledgments
This work is partially supported by Spanish MICINN under project ANYSELF4 (TIN2011-28627-C04-01), and by Junta de Andalucía under project P10-TIC-6083 (DNEMESIS5).
References (39)
- et al.
Evolution of the social network of scientific collaborations
Physica A. Statistical Mechanics and its Applications
(2002) - et al.
Empirical analysis of the evolution of a scientific collaboration network
Physica A
(2007) Growth and structure of Slovenia’s scientific collaboration network
Journal of Informetrics
(2010)- et al.
A graph-theoretic perspective on centrality
Social Networks
(2006) Centrality in networks: I. Conceptual clarification
Social Networks
(1978)Computational intelligence and games: challenges and opportunities
International Journal of Automation and Computing
(2008)- et al.
Search-based procedural content generation: a taxonomy and survey
IEEE Transactions on Computational Intelligence and AI in Games
(2011) - et al.
Self-adaptive games for rehabilitation at home
Social network analysis
Sociology
(1988)
The structure and function of complex networks
SIAM Review
Emergence of scaling in random networks
Science
Collective dynamics of ‘small-world’ networks
Nature
Understanding patterns of international scientific collaboration
Science, Technology & Human Values
The evolution of the mathematical research collaboration graph
Congressus Numerantium
The complex network of EC authors
ACM SIGEVOlution
Where is evolutionary computation going? A temporal analysis of the EC community
Genetic Programming and Evolvable Machines
Six degrees of Jonathan Grudin: a social network analysis of the evolution and impact of CSCW research
The structure of scientific collaboration networks in scientometrics
Scientometrics
Cited by (20)
Statistical analysis of risk assessment factors and metrics to evaluate radicalisation in Twitter
2019, Future Generation Computer SystemsCitation Excerpt :For instance, the information network made of citations between academic papers in which every node is a paper and there is a directed edge if a paper cites another [15]. Other approach is the one followed by [16–18], in these works the authors of the academic papers are represented in the nodes of the co-authorship network and the different edges imply there is at least an academic paper written by them. Regarding Social Networks [19], there are studies on business relationships between companies [20], supply chain context [21] and obesity [22], to name a few.
From ephemeral computing to deep bioinspired algorithms: New trends and applications
2018, Future Generation Computer SystemsPolarization in the social sciences: Assortative mixing in social science collaboration networks is resilient to interventions
2018, Physica A: Statistical Mechanics and its ApplicationsCitation Excerpt :The structure of scientific collaboration is often analyzed by drawing on proxy measures such as co-authorship networks. A co-authorship network is comprised of researchers as vertices and joint publications as edges [1–8]. Modeling the structure of co-authorship networks can be instrumental for designing better higher education and research institutions, mapping knowledge domains [9–12], understanding how innovation comes about [13,14], and understanding in how far the network structure has repercussions on individual researchers’ performance [2,15,16].
Seagull: A bird's-eye view of the evolution of technical games research
2018, Entertainment ComputingCitation Excerpt :This timeframe for emergence seems realistic considering the community did not fully solidify until the formation of IEEE Computational Intelligence and Games conference in 2005, and the transformation of the AAAI Spring Symposium on Artificial Intelligence and Interactive Entertainment (AAAISAIIE) to the AAAI conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE) in 2004. Furthermore, the appearance and growth of the AI community in our analysis matches the findings of Lara-Cabrera et al. [20] which identified a marked increase in the number of papers and authors from 2005 onwards, sustaining the steady growth of the field. Thereafter, the community continues to develop and interact with others, especially interactive narrative.
A new approach to understand metro operation safety by exploring metro operation hazard network (MOHN)
2017, Safety ScienceCitation Excerpt :Zhou et al. (2014) studied the topological characteristics of subway construction accident network (SCAN) to understand the complexity of subway accidents by using complex network theory. Lara-Cabrera et al. (2014) presented a network-based analysis of the scientific collaboration network of computational intelligence in games at the macroscopic, mesoscopic and microscopic levels. Deng et al. (2015) proposed a network modeling approach to identify the critical and vulnerable modules in metro system.
An overview of self-adaptive technologies within virtual reality training
2016, Computer Science Review