Using Artificial Intelligence in board game beta testing

Undoubtedly, artificial intelligence is rapidly pervading our daily lives, from entertainment to professional applications. We are immersed in a technological context where the competency to effectively utilize diverse tools is essential for optimizing our actions and outcomes. In this post, I will discuss the methodology I employed in using AI for the beta testing phase of my last two board game projects.

First and foremost, I must clarify that my application of AI is not a substitute for creative effort. I avoid using lazy or generic prompts such as, "create a card game with Roman Empire theme using trick track mechanics." I consider this approach to be ineffective, as it would likely yield a generic, derivative product lacking a distinct core concept or "soul."

Instead, I employ AI in a pragmatic and iterative manner. The process begins with establishing the game's core concept, writing the comprehensive rulebook, and compiling a complete list of assets (board, components, dice, etc.). This initial stage results in a prototype ready for preliminary testing. However, prior to testing with human players, I utilize Manus AI to conduct a complete virtual beta test session.



The process is straightforward: I submit the rulebook (PDF) and the file detailing the game pieces (PDF). I then provide the initial prompt: "Manus, I request you to analyze the content regarding my new game. Upon completion of this analysis, please identify any aspects that are not completely unambiguous or 'crystal clear'." After addressing the system’s initial queries, the final prompt is executed: "I now request 100 simulations encompassing two, three, and four players. The required output is a detailed feedback report assessing game balance, pointing out potential rule errors, and suggesting improvements for the core mechanics." A significant advantage of Manus is its capability to generate Python programming code to facilitate testing on external platforms beyond the application itself.

It is crucial to note that testing with human players will always remain a fundamental part of the development cycle. Nevertheless, the initial results demonstrate that the prototypes are significantly more refined before the first human playtest. For example, in a recent beta testing session for my new game, the prototype was substantially more polished prior to the initial live test. The overall iteration process becomes considerably more efficient when AI is applied judiciously.

I am currently refining this methodology and anticipate being able to share further insights and impressions soon.

#GoGamers

The significance of artificial intelligence in contemporary game design

In this post I want to briefly discuss the importance of artificial intelligence in the gaming industry. As Togelius (2024) highlights, AI is increasingly prevalent in our daily lives, and games, as prominent elements of the technology industry, are no exception. While the author acknowledges that some recent AI systems exhibit a degree of generalizability, most remain highly specialized: they can only perform a single task within a specific context. For instance, a spell checker cannot perform mathematical calculations, and the world's best chess program cannot play Tetris.

As Togelius (2019) also says, recent advancements in artificial intelligence have significantly enhanced our capabilities, particularly in the realm of game development; as we continue to explore and refine these techniques, their potential applications within games extend far beyond the traditional role of providing skilled opponents. It is imperative that we adapt our game design methodologies to fully leverage the power of advanced AI algorithms and to understand how to use generative AI to help us to create better game designing practices.



Therefore, artificial intelligence has been a constant companion in the evolution of video games, shaping the way we interact with virtual worlds. From the very beginning, AI in games has sought to simulate intelligent behaviors in non-player characters (NPCs), making gaming experiences more immersive and challenging. According to Millington (2009), the early stages of AI in gaming were modest, with simple algorithms determining basic movement patterns and reactions of NPCs. However, as technology advanced, so too did the possibilities. Today, AI in games is capable of generating complex and adaptive behaviors, making NPCs more realistic and challenging. Based on the ideas of Togelius and Millington, it is important to observe the evolution of techniques enabling the use of artificial intelligence in recent decades. Comparing 1962, when MIT students interacted with the Space War game prototype on a static screen with minimal input commands, to 2017 and Echo game, where enemies learn from the player's combat patterns and become increasingly dangerous and unpredictable, highlights this dramatic progression. Since 2017, advances have not stopped and, every year, we are surprised by a new use of AI in games.

In other words, when contemplating the application of AI in gaming, we are not merely considering the development of more efficient adversaries that learn from player behavior or environments that adapt to the specific challenges a player seeks. AI tools can be instrumental in brainstorming processes and in establishing the foundational pillars of game creation. Particularly for small-scale studios and producers with limited resources, access to generative AI tools can serve as a powerful asset to facilitate game development.

 #GoGamers

---

References:

Millington, I., 2009. Artificial Intelligence for Games. Burlington: Morgan Kaufmann Publishers.
Togelius, J., 2019. Playing smart - on games, intelligence, and artificial intelligence. London: The MIT Press.
Togelius, J., 2024. Artificial general intelligence. London: The MIT Press.

Using AI for game balancing

Since last year, AI has become a more integrated part of my daily life. It's inevitable, and we need to learn how to utilize these technologies effectively, making them our allies in the development process. While I'm not a proponent of using AI to solely generate game ideas, I believe it holds immense potential for game balancing.



Let me share an example from a card game I'm currently developing. It uses trick-taking mechanics with a food theme. The AI I'm working with is Google's Gemini (formerly Bard). Here are some prompts and commands I've found valuable in this project:

1. Gauging AI Knowledge: I began by assessing Gemini's knowledge of trick-taking mechanics. I asked, "How familiar are you with trick-taking mechanics in card games?" The response was positive, and the AI provided several examples within this category.

2. Visualizing Game Mechanics: To help the AI understand the gameplay flow, I shared a sequence of images depicting the activity loop of each round, presented in a comic book format. I phrased it like this: "I'd like to show you a sequence of my game in comic panels. Can you tell me if you understand how a round works?"

3. Asking Specific Questions: Instead of open-ended prompts, I focused on specific goals. For example, regarding card distribution, I asked: "Based on the core mechanics I've shown you, can you suggest a better way to distribute cards in my game to create a balanced experience for players?"

It's important to remember that AI isn't a magic solution. The first suggestion Gemini provided needed some refinement. However, through further prompts and iterations, I arrived at a more viable structure for my game.

Next Steps: The next step is to create a simple prototype for initial beta testing. Player feedback will be crucial for further refinement.

Overall, AI presents a powerful tool that can assist with various aspects of game development, including balancing, code, narrative, and concept ideas.

#GoGamers

Why do you play games?

In the first post of this year, I want to share some content from the book I’m reading at the moment: Playing smart - on games, intelligence, and artificial intelligence by Julius Togelius. The author discusses many aspects on how games challenge us and what we can expect from games that use artificial intelligence in the near future.



One of the first points discussed by Togelius is about the question I already brought many times in this site: why do we play games? It’s not easy to answer (and we have many different views for this subject) but I think it’s essential to gather multiple points of view to create a more solid opinion.

Togelius launches the question: why do you play games? And starts his answer with a very interesting argument that most of the time we are playing games for many reasons but all the time – despite the game we are playing – we are doing an exercise of intense planning.

Below, I want to share this excellent content from his book and recommend the reading for all the followers of this site:

Why do you play games? To relax, have a good time, lose yourself a bit? Perhaps as a way of socializing with friends? Almost certainly not as some sort of brain exercise. But let’s look at what you are really doing: You plan. In Chess, you are planning for your victory by imagining a sequence of several moves that you will take to reach checkmate, or at least capture one of your opponent’s pieces. If you are any good, you are also taking your opponent’s countermoves into account and making contingency plans if they do not fall into your elaborately laid traps. In Super Mario Bros., you are planning wheter to take the higher path, which brings more reward but is riskier, or the safer lower path. You are also planning to venture down that pipe that might bring you to a hidden treasure chamber, or to continue past it, depending on how much time you have left and how eager you are to finish the level. You may be planning to eat the power-up that lets you get through that wall so you can lick a switch that releases a bean from which you can grow a beanstalk that lets you climb up to that cloud you want to get to. In Angry Birds you are planning where to throw each bird so as to achieve maximum destruction with the fewest birds. If you crush the ice wall with the blue bird, you can then hit that cavity with the black bomb bird, collapsing the main structure, and finish off that cowardly hiding pig with your red bird. (TOGELIUS, 2019, kindle edition – position 412)

#GoGamers

---

Reference:

TOGELIUS, Julius. Playing smart - on games, intelligence, and artificial intelligence. London: The MIT Press, 2019.