Publisher: Charles River Media
ISBN 10: 1-58450551-6
Companion Site: http://www.emergenceingames.com/
My Rating: 6/10 (Interesting)
Summary: This book is a compendium of information on methods of increasing interactivity in your games, and even though it does not deliver on the promise of developing methodologies to design for emergence in games or offer analyses of complex dynamic game systems, it does a good job of introducing the reader to the field, especially if you are willing to read all the reference material introduced in the book.
Emergence and complexity in game systems are my main research interest, so when i got my hands on this book, I was overjoyed by the prospect of reading up on innovative methods to design for emergence, and gauge the types and extent of emergent behaviors based on design templates. I picked up a hot tea, and sat down, expecting to find original ideas and a gain deep understanding of complexity in games. As always, I jotted down notes about my experience with the book, as I read on. This review is a summary of those notes.
Chapters 1 and 2 give us a basic understanding of emergence and complex systems. The approach here is to offer examples form the physical, biological and game design perspectives, and describe how those systems can show emergent properties. There is a definite lack of building up from simple concepts to the more complex, and often other books are cited, without enough explanation of the underlying concepts. This makes the book less useful for the average game developer who has no grounding in complexity studies or chaos theory, and since the citations aren’t followed by in-depth analyses and fresh ideas, this section will also not prove very interesting to anyone with a deeper understanding of complexity, trying to find new applications in games. The way the first two chapters are written, reminds me very much of university textbooks, where facts are listed and references are given for those interested in further examination of the subject. No new ideas are offered, which is strange, because the application of complexity studies to video games is such a novel subject and such an open and new field.
One more warning, before we go on: some passages in the book contain information that are stated matter of fact, but are really highly debated in the community. Read this quote from the book e.g.:
“The potential for emergence is compounded when the elements of a system have some capacity for adaptation and learning.”
We know that genetic algorithms adapt, and provide good solutions under the right conditions, and yet they are not seen to be learning. Rather, they explore the solution space for “fit” candidates, and weed out the more “unfit” solutions. This is pretty much how DNA and nature work together to create all living organisms, and life is probably the most emergent phenomenon we know! It isn’t the Lamarckian capability to learn during a lifetime and transmit that knowldege to the next generation that makes life “adaptive”, but rather the Darwinian selection process, whereby interactions with the physical world and other living beings remove the “unfit” individuals at a higher rate than the “fit” individuals. A single bacteria, will not learn and adapt by any choice of its own, and truly if it could, we might never had such huge variety of life around us! In this way, learning, may not be integral to emergence, while the ability tp “explore the solution space” certainly is.
“In brains, the large number of neurons with enormous interconnectivity gives rise to thought, emotions, and memory.”
At the risk of sounding picky, I’d have to say that a large number of interconnections will not neccessarily give you a better ability to generalize/adapt when speaking of neural networks. Fully connected neural networks, e.g. Tend to perform badly (both in terms of generalization capability and in terms of time needed to train all those weights) compared to well-designed or smartly “grown” or pruned networks. There is and understanding nowadays, that making things more complicated (i.e. And overabundance of interconnections of your subsystems, or elements) does not make systems more dynamically complex or allow for more emergence, and can very well lead to chaos, or simple oscillations.
The book references at the end of the chapters are certainly well-chosen, and I was positively surprised to see references to John Holland and Andrew Ilachinski, and if the reader picks up those books, and other books referenced by the author, she will certainly be able to gain a solid foundation of the field of complexity, and a treasure trove of new ideas in the field.
The third chapter starts with a discussion of how important ease of interaction with the game systems is to players, and how feedback on this issue can be gathered from players- hardly an issue that fits into a book about emergence. If anything, this discussion should go under the heading of “immersion”…in a different book. The discussion about player freedom and physics on the other hand is quiet necessary, though possibly superfluous for the reader who is already in game development. The discussion really picks up when sandbox games are mentioned, and emergent gameplay (e.g. Through physics in the Valve games) is introduced, but before the thing starts to get interesting the author jams in a section about Flow and GameFlow (the results of some research the author and others have done on what enjoyment is to players. Again, and not for the last time, the author discusses immersion, where she should be talking about emergence, a completely different issue!
Chapter 4 is named after the books title “Emergence in Games”, and I am hoping to finally get some in-depth discussion about designing for- and measuring emergence in games, and I did. Look at this insightful paragraph:
“The key to creating emergent gameplay is to define a simple, general set of elements and rules that can give rise to a wide variety of interesting, challenging behaviors and interactions in varying situations. The simpler and more generalizable the rules, the easier they will be to understand (for the player and the developer), test, tune, and perfect for emergent gameplay. The simplest solution that gives the desired results is always the best. As with any emergent system, the fundamental set of rules and elements stay constant, but their situation and configuration change over time. The sensitivity of the elements to changing situations and the interaction of the elements with each other and the player are what create emergent gameplay.”
Unfortunately, the ideas in that passage are never discussed. Instead we are offered examples of how previous games have managed to model interactive environments, agents and narrative. Here the author seems to try to prove that interaction will inevitably lead to emergence! The rest of the chapter, is happy to offer examples, instead of design guidelines, for emerging social interaction, economies, etc. Until we get to page 108 – Developing for Emergence. Now this is where I’ll get my money’s worth I say, and I have been fooled again! The section I was most looking forward to turns out to be a discussion of what production and QA problems you might encounter if you allow for emergence in a game! Nothing about how to actually design for emergence!
At this point I am actually disappointed, but I’d never put a book on game design down, so i keep going. The rest of chapter 5 is basically a condensed, albeit simple, introduction to machine learning methodologies such as fuzzy logic, neural networks and genetic algorithms, along with a list of games that employ those methods for some of their functionality. There is no discussion of how well those features perform, and if they could have been made to work differently, or if they increased or decreased the possibility of emergent gameplay. In one of the examples, Dirt Track Racing, the developers e.g. did not claim to have used the neural net to create emergent behavior, but rather to guide the competing cars around the track in a realistic way. [like from gamasutra]
Chapter 6 quickly moves to a description of the “Active Gameworld” project, along with some code and data structures used in the project. Active Gameworld is very similar to the environment simulation of Dwarf Fortress, where simple physics of pressure, fluid flow, heat distribution, etc. are simulated via Cellular Automata and other similar similar methods. The discussion is interesting, with real examples from the author’s own project, and because we have a simple simulation of nature (vs. The dead heightmap + obstacles that most game terrain boil down to), with many different interacting elements, the project is very relevant to complex systems and emergence in games. If you only have time to read one chapter in this book, it should be chapter 6.
Chapter 7 provides a model for agents who could live in the Active Gameworld. Their behavior is basically scripted, but based on physical parameters in their environment (e.g. They run away from a burning hot cell, presumable on fire) and simple goals of maximizing/minimizing certain values calculated from the environment. This is a great example of how a complex environment allows your agents to display complex behavior, even though the agents themselves might be exceedingly simple. Unfortunately, the idea is not fully explored. E.g. there are discussions about agents picking comfortable locations, and pathfinding, but the agents are not allowed to modify the world in a meaningful way (the way ants for example build a nest), so the world is the only real source of complexity, and the agents are mostly “reactive”. The original idea is brilliant nonetheless.
Next, we get a description of a game, based on flocking. The idea to use simple agents to generate lots of complex looking content is compelling, but the implementation details and game code take up way more space than is warranted. Those fifty or so pages read like a stand-alone final project report in a programming course, rather than being an integral part of the book.
Chapter 8 is titled “Emergent Narative”, which makes me think about automatic story and content generation, a deeply unappreciated, yet very interesting topic in the study of AI and expert systems. My guess is that much work isn’t being done on that topic, because it is hard to imagine immediate and profitable uses for an auto-generated story, while other fields, such at computer vision and classification, for example, eat up a disproportionally large share of research funds, because they have thousands of immediate applications. I am diverging… So does this chapter of the book cover automatic narrative generation? Nope! Instead we get a discussion about what storytelling is and the different forms it may take in games. Though the discussion is interesting, it is by no means to the point! Where is the emergence? And why do we need to learn the basics of storytelling for game in a book supposedly covering emergence in games?
The only relevant idea brought forth is that of creating a Hebbian Network-like structure to decide which parts of the story to expose to the player, but that idea is not pursued. It is unfortunate that the author limits herself to a recap of common knowledge on interactive writing and stories for games that can be found in any book on game design so we end up with a chapter you can happily skip if you have read other books (e.g. Creating Emotion in Games or Character Development and Storytelling for Games) on the subject.
Chapter 9 is titled “social emergence” and deals with social interactions and market economies in MMOs. The chapter gives a good overview of where to find emergent behaviours (e.g. player-ran banks in MMOs, or social strata where the game doesn’t necessarily provide a fixed mechanism for them) but does little in the way of instructing us on how to design game systems so that players find plenty of ways to create their owen “kind” of social and political gameplay.
Chapter 10 is a quick summary of what we have learned throughout the book. I’d have to mention here, that almost each chapter provides an interesting interview with a game developer, loosely related to the content of the chapter itself. The interviews are too short to be informative, but they offer a welcome change from the often repetitive tone of the writing.
Having read the book, I have a nagging question in the back of my head: what audience is the book is intended for? It is clearly not for the professional developer, as there is way too much about basic design concepts in the book. It isn’t for the adept complex systems engineer either who might want to apply his knowledge to games either, because a large number of modern findings in the field of complexity and chaos have been omitted. Is it meant for the layman, or aspiring future game developer then? Not really, because having no grounding in the science of complexity will leave you wondering what many passages mean! So…who comprises the audience for this book? Probably conference goers, who have heard about complexity and all the buzz it has created, and who are aware of video games, and all the buzz games have created, and who would happily welcome a relaxed introduction to both, while also gleaning some basic ideas on how those two subjects might be married, to generate a wealth of low-quality introductory journal papers.
To summarize, the book covers a lot of ground in game design principles and does quiet a good job of proposing methods to increase the interactivity of the game, and to immerse the player in the world. It does not deliver on developing methods to create emergence in games, despite its title and a number of brilliant ideas (such as the “Active Gameworld”). I enjoyed the book nevertheless, if only because it addresses an issue that is so dear to my heart. I hope more writers will follow in Ms. Sweetser’s wake, and that we start to see more in-depth, and more creative articles and books on this topic.