Pseudo Random Design - The Next Frontier?

by Gabriel Putnam

I've heard it claimed that nearly every RPGamer flirts with the idea, even if only idly, of being a game designer at some point in their life. Its a natural extension of our favorite hobby, and has the potential to be a perfect fit of doing what you love and getting paid for it. For many, it would also be the opportunity to take part in a world community and tell a new generation of gamers the stories which filled our youth; kind of a modern day myth progression. However, consider this. What if the progression of our culture, the translation of our stories, and the creation of epics for our next generation was taken from the hands of designers? They would no longer be needed as stories could take on a life of their own and expanded from a tiny foci into an infinite variety of paths. What then? Would this be good or bad, and what on Earth could evoke such a change? For the first two questions, the answer is open to debate, as opinions on the topic would vary wildly. However, for the third the answer should be obvious. Stories and programs which can write themselves.

Now, I can understand the apprehension that this might all just be pie in the sky talk, with a product which could not materialize until years from now, if ever. However, this is part of the issue which I'd like to discuss, so for now, lets start with a basic outline of what I envision for such a program, and an explanation of its mechanics. Also, it should be noted that much of this idea is spun from conversations on the Game Development board, with twentytwo and Locrian being the two main initiators. That said, the general concept is based upon what is commonly referred to as a neural network among programmers, psychologists, and the like. Such a network operates in a vast state of parallel distributed processing, where numerous possibilities are tried and those which are the most fit, in regard to a set of proposed heuristics, are then propagated onward or stored for further refinement and calculation. In a roundabout way, this is simply saying that such a network contains the possibility for learning. Now, with the potential for refinement of concepts, there begins to be the real possibility for creation of information which is effectively pseudo-random in nature. I say pseudo-random, because one of the main factors in the operation of a neural network is the creation of the above mentioned heuristics. These heuristics provide a set of guidelines which help the neural network to make choices among varied sets of data and determine which of its is options are most fit for further propagation.

The way that this concept of learning networks would then tie into game design would be through the creation of a network based game client. This piece of software would be crafted with the purpose of creating sets of randomized game structures in which all aspects of the game were determined with only limited input from designers in the form of heuristic guidelines. Now of course, there are some obvious problems which should be immediately apparent in this proposal. The first of these being simply the amount of time and effort which would be necessary to create such a complex framework of neural selection. I've often heard it said that the creation of an operating system has become such a task these days that it would take the equivalent of 100's if not 1000's of human lifetimes for one person to program it alone. In a sense, a neural network project of the proposed magnitude would be little different. It would be a vast undertaking which might very well require the finances of a major corporation to reach completion. However, there are a number of factors which could help it avoid some of these difficulties. The first is that there is a vast body of available programmers on hand in the populace these days, many of whom would enjoy nothing more than to work on such a ground breaking project - possibly even in their spare time as a hobby. The second is that the very research which would be undertaken to create such a design, could also be easily leveraged to increase efficiency. Heuristics for a number of coding sub tasks to the project could be created which would alleviate much of the mundane or repetitive workload on programmers so they would be free to pursue more of their core goals.

A second problem which many have already envisioned with this type of design is the worry of a final product which would emphasis quantity over quality. Now, there are a couple of tacts to take in discussing this. The first is simply that the ratio of quality to quantity would be the same formula as with any other aspect of game design. If a particularly motivated team of individuals sit down to make a classic example of an RPG, there is probably a good chance that their final product is going to be enjoyable to play, interesting from a creative standpoint, and financially successful among the RPG community. The same theory would also apply to a team of individuals who are gathered to create a structure than can then design limitless games itself in the future. If this team is composed of motivated and hard working people, then they will take the time to ensure that the final products produced by this system are of an acceptable quality to appease their fans. They will check to make sure that the machine designs stories and dialog which are at least to the level of a human programmer, and they will thoroughly playtest and evaluate the final products of this machine to make sure they encompass a level of fun which reflects well on the team as a whole. The other way to deal with this problem would simply be to say that randomness, in and of itself, is an interesting and engaging medium, which offers numerous possibilities for enjoyment if you are so inclined. As a break from the standard fare of RPGs, randomly created adventures/stories, with only a limited heuristic of cohesion to common guidelines, could offer up a whole new venue of games, in which the off the wall aspects were almost as important as the game's core features.

One of the next major points which should probably be covered in regards to pseudo-random game design is the inherent limitations of those who you would have working on such a project. As the old saying goes, we are indeed only human, and if a project of this magnitude were undertaken, there would almost assuredly be aspects or ideas that were forgotten along the way. For example, it might initially be created with only a fantasy theme in mind, and as such, all of the stories it spit out would only have access to fantasy themes, graphics, and ideas during their creation. If you wanted to play a sci-fi or modern day epic using such a system, you would probably be out of luck due to a simple lack of foresight among the human staff. However, there are a couple of ways in which this limitation might be overcome. The first would be to offer limitless content in the way of graphical updates, literature insertions, and thematic upgrades. These upgrades could be official ones, generated by the community or corporation in response to consumer requests, or they could be individually motivated; performed by the owner or player so that their favorite type of settings would be reflected in the games they played. This would allow for periodic insertions of lifeblood into a sound, but possibly outdated architecture, that could then let it keep up with the times, and current entertainment trends. The other option for overcoming this problem would simply be to code the network with its own ability to generate underlying content. For example, whenever a player wasn't using the system, they could instead set it about generating graphics and phrases which fit a general set of heuristics based around the next game concept they wanted to play. This would, once again, require a significant amount of initial skill to make sure it was implemented correctly, however, these concerns have been addressed above, so they would simply be rolled into one more hurdle of design.

Finally, however, there is a single problem which could easily be the death of such a concept - it would be a horrible development in the eyes of the major gaming studios. Such a machine would effectively strip away their ability to provide the masses with a desirable or innovative product, as the tools to create such a product might be sitting under every gamer's TV at home. One of the potential avenues which this creation process could take instead then, would be for the tools of randomized design to be held by the companies who created it, and the final products of this process then sold to the waiting masses. This would be ideal for the companies, as they would have traded a large initial investment in programming labor, for a limitless supply of randomized content. The public at large could even be kept unaware of this process for the most part, and they could effectively ride on their laurels until gaming itself became outdated. The other option to solve this problem, would instead be for gaming companies to focus on only the sale of targeted and human created games. There will always be those who wouldn't believe in the ability of machine powered creative thought, and as such, there would still be a fertile market of consumers for human creations from the heart.

Overall, I believe that the avenue of pseudo-random game design represents one of the great untapped paths of gaming's progression. The major stumbling block on the road of this process is of course the huge initial investment which would be required to really make it work. Parts of the design, such as the creation of efficient, targeted neural networks, could of course be done piecemeal beforehand to pave the way. However, the final act of assembling these disparate pieces and forming a cohesive whole would represent a vast undertaking even for most corporations. However, the potential return on such an investment is simply staggering. A wealth of ready made titles, which might require only limited tweaking from a team of designers, available for sale from now until the end of electronics based gaming; a nearly limitless source of income, produced at a fraction of the cost once initial investment had been overcome; or a supply of concepts for the next new craze, created by a system only partially tied to the strictures and guidelines of common human thought.

Anyhow, lots of brave new world talk which can be easily ignored by those who feel that randomized game design is a topic best left for theoretical discussion. However, someone will eventually accomplish this task. And when they do, everyone will have wished they'd invested in them back when they were small or used the idea themselves.