In the academic writing and research design I will use qualitative method. In games design and public administration fields, the qualitative research design is commonly used and proper for the topic. And in following case studies, comparative case studies will be used to illustrate pros and cons. Frostpunk by 11 bit studios and Civilization VI by Firaxis Games will the two research cases. Also, I will focus on how 11 bit studios present the history and outcomes by child labor policy in Frostpunk’s Law system.

In public policy analysis, the first step of research is to define the problem correctly, according to Bardach and Patashnik(2019). So what is child labor? The common explanation is the exploitation of children’s human rights and intervention of their schooling. Child labor situation is often related to low productivity, lack of legislation and adult labor market. As Grootaert and Kanbur(1995) point out, even though the first World Summit for Children in 1990 ”were no explicit goals relating to child labor, but the target included basic education for all children. These goals, if met, will reduce child labor.” The problem was and will never be independent from child rights protection. In policy making fields, there are many solution a government can make to improve children situation. Some are even considered unrelated in the first place.

According to Grootaert and Kanbur(1995), poverty, population policy, employment opportunities for parents can all be policy intervention factors affecting child labor. In Frostpunk, the first thing we will face is poverty – lack of heating resources. In order to search for resources and food supplies, player will choose whether they will use child labor. One thing to note is that you can choose child labor from the very start of the game and it will affect till the end. Actually according some interview with players from Reddit, child labor is the first policy they make to increase productivity.

And will 11 bit studios stop at here discussing child labor? No, they took a step further. As we have talked before, the improvement of children education have positive effect on reducing child labor. In Frostpunk, if player were not choosing child labor, they can send children to training schools as apprentices. Yeah they did it into the first policy making for players. Though it will cost resources and one opportunity for policy making, it will later bring more productivity. In actual research it have the exact same outcome. For a sole family, education can be a burden at first, but later it can bring more chances to get rid of poverty for the whole family. In game, 11 bit studios erase the uncertainty and make it profitable in a long period.

One more thing to notice is that, according to Dolkemeyer(2020), Frostpunk provides a “Living-with and Dying with community” to the players. Citizens in game will react to player’s policy making as a community and a family. In making the child labor approval law, citizens will lose hope, which is a system that affect gaming process and ending. Citizens are complaining about that kids should be in school and not be working as adults.

In conclusion, we have a quick review of policy analysis process and child labor situation and affecting factors. Combining with decision making system in the game Froskpunk (11 bit studios, 2018)., we have a quick review on both child labor and the game.

Bibliography

Grootaert, C. and Kanbur, R., 1995. Child labor: A review. Available at SSRN 620526.

Dunn, W.N., 2015. Public policy analysis. Routledge.

Bardach, E. and Patashnik, E.M., 2019. A practical guide for policy analysis: The eightfold path to more effective problem solving. CQ press.

Dolkemeyer, L., 2020. Autocracy for the People. Modes of response-able Action and the Management of Demise in Frostpunk. gamevironments.

Devlog 1 choosing the topic

In the first week of deciding the research topic, I chose two topics including both what I am interested in and what I did in my bachelor degree. The first topic is “moral dilemma” in games. It is the very first reason I chose games design as my lifetime career. In Witcher series by CD project RED, I faced so many moral choices which have pros and cons in both sides. Under so many choices I felt like I am more into Geralt of Rivia and the cruel worlds in Witcher. In Bioshock by 2k games, the choice between kill the little sister and save them is still a hard decision to make. In a corrupted world like Rapture, the best decision would be to harvest the little sister to become stronger in every aspect, but humanity is telling you to save them. Those decision making system are charming in Roleplaying games, making it both immersive and impressive.

Another topic is related to my degree in management. I often concentrate more on public policies of a game while playing. Is it acceptable in that situation? Do it have a prototype in real life? What would I do to change the system if I were a designer? But in the following research, I found that this question might be too broad. Almost all games have a decision-making system, while some of them are related to ruling the citizens. If I chose this topic, it must be narrowed down to a more specific question.

Bringing up these topics, I had a discussion with my course leaders and course mates. The moral dilemma is more attractive than the other, but considering the size and content of this module, I could choose this topic as my final major thesis. And the policy making system can be the practice of my academic writing and research.

One major problem of the “policy making system” topic is how to narrow it down as a specific research question. The first version of the question is “How to refer policies and outcomes in video game design”. In that topic, it includes multiple policies and its outcomes, which might lead to a rough conclusion and unrelated case studies. For example, one policy might be about work overtime policy progression while the other might be immigration problems. These study areas are way too broad to discuss under 2500 words limit and are quite easy to lose focus.

After discussing again with Maddy and David, I changed the topic into a more distinct one: policy on child labor. The reason why choosing it is that in one of the case studies, Frostpunk by 11 bit studios have a perfect example of pros and cons in child labor usage. And until today, there are still areas in the world which have serious child labor problems. Children are not getting enough education and leisure time while they are working as an adult. And also with the contrast of different games, I can conclude more on how designers refer policies made by different governments and period of history.

In the end, the research question for my academic writing is how to design a decision-making system in video games learning from child labor: a case study.

Remnant: From the ashes (Gunfire Games, 2019)

Attributes:

• Boss fight design with certain patterns
• Souls – like shooting game
• Cooperation up to 3 players
• Various Gear builds and multiple playthrough required

Remnent: From the ashes is a brave tryout in Souls-like games, based on From Software’s well known Soul franchise. It has a punishing difficulty curve which will bring sudden death to unprepared players. But the innovation and combination of TPS and Teamwork against monsters is still worthy playing for 20 hours. I personally play through the game solely which makes it even harder. In different gear builds, players will have the chance to try many build plans, which is the progression loop that drive players to challenge themselves. Besides, the strictly control of materials output ensures that the levels are always challenging. According to developers, players can only reach 45% of the game content during their first playthrough, so each player will experience different dungeons and bosses’ combination.

The reason that I choose Remnent: From the ashes is its boss design. Even though some critics and players do not enjoy it that much judging from game mechanics, there are still a clear pattern in boss design. The skills of each boss are normally restricted within 4 kinds. And they can be dealt with 1 move (Dogde, in Remnent: From the ashes) towards different directions. So in balancing with learning cost and playability, they found the combination is suitable.

Elements can be used:

• Boss fight design

Alien: Isolation (Creative Assembly, 2014)

Attributes:

• AI-driven stealth game
• Perfect art design and details based on original film in 1979
• Great horror atmosphere

Alien: Isolation is a great salute to the famous Ridley Scott’s sci-fi film Alien. It has a breathtaking horror atmosphere which reminds me of Amnesia and Outlast. (Which two, by the way, I am just too scare to finish). Player cannot kill the Xenomorph during the gameplay. Only choice is to sneak away from him. But as the alien is extremely sensitive to any sound, players are still hard to escape from it.

So in this game, the AI design of the Xenomorph is critical. Player must feel that the alien is dreadfully smart. Then they will be satisfied when they are “smarter” by finishing the game. This experience is ensured by 2 AI system in the game: Alien AI and the Director AI. While the Director AI knows player’s position the whole time, it can only tell the Alien AI roughly where the player are. And then Alien AI is there to hunt player down. This design is much like the director AI design in Valve’s Left 4 Dead, which can track how much pressure the player is taking and adjust zombie’s aggressiveness from it. This works similarly in Alien: Isolation. And in Xenomorph itself, AI is based on behavior tree containing over 100 nodes in total and almost 30 nodes in the highest level. Certain parts of the behavior tree are locked as the game starts, but in late game phase, these will be triggered to make player believe that – the Alien are learning from my pattern. This is what I what to achieve in my game.

Elements can be used:

• AI-driven mechanism
• Behavior tree design (unlock certain nodes under conditions)

Metal Gear Solid V: The Phantom Pain (Kojima Productions, 2015)

Attributes:

• open world stealth game
• a Hideo Kojima production

Metal Gear Solid V: The Phantom Pain is an open world stealth game developed by Kojima Productions and published by Konami. The game has large content that I could possibly use, so I will just focus on one design. In MGS V, there is an enemy response system. Once you clear the camp with more of headshots than other approach, the soldiers in the next level will wear helmets avoiding the player finishing the game with only one way. This AI-based system is to ensure the complete player experience in combat system, lowering the marginal benefit of using only one method. In my game, it could be used as a core mechanism that player cannot use one method to play through the whole game. They are encouraged (and forced) to discover other way to achieve the goal.

In the industry, the highest comment of an AI system is that each AI has its own characteristics. Some might be cowards but others could be brave as hell. The major usage of Behavior tree is assuring this surprising attribute. But on the other hand, even these AI enemy have their own personality in games, they are still working for player’s gaming experience.

There are trends that machine learning is replacing original AI project in many ways. Let the AI refine itself seems way more efficient than artificially changes. Like training the Alpha-Go for the Go, or other games like StarCraft II. But in actual game develop industry. I do not think machine learning will be the main stream. For whatever reason, game designers bring AI into a game not to destroy player’s experience.

For example, a Soul-like game have an extremely hard boss which use the real-time machine learning, it will have probing attack and learn from player’s positioning and habits. Like for certain kind of attack, the player will tend to dodge leftwards. Then next time the boss will have double attack, one in the original place and one in its left. If the player plays long enough without changing his or her habit, the boss will be unbeatable. It sounds cool but it will totally ruin player’s experience which is totally unacceptable in the industry.

Another key problem in machine learning is the differences between copies. In standard development process, the game should have the similar experience for all players. But once there are machine learning implemented in the game, causing different difficulty for all players, the experience are no longer similar and standard. Players can choose to be different characters in a RPG game but they should have the same experience once they made same choices.

For a long time the Behavior tree will stay with game development. But for indie project or experimental project in games design, I am eager to see an unkillable demon AI that destroy all players.

Due to personal reasons, I have to give up some of the settings and plots and stick to the original mechanics – an evolving AI.

In implementing the plan, the game will not choose behavior tree as development tools. The reasons are listed in the last development log talking about the difference between Behavior tree and Finite state machine.

In achieving the mechanic, I will divide the game into different levels. Each level will record how the Death Bringer died so far. And if player try to kill him with repetitive solution or used solutions, boss will know and mock the player. In original plan, there are 5 levels and 4 solutions. In final stage of the game, player have to try all the repetitive solutions and then there will be a boss fight. It could be a melee fight or a shooting game like Raiden. And for now the plan will be canceled and the level will be less due to certain reasons.

In my original plan of level design, there should be an easy solution first, and then there will be difficult choices. But during the playtesting, players will choose the easy way frequently and have the impression that this game is a simple game. And then in the scenes of the Death Bringer knowing about what they did before, the players were surprised.

But the progression is not that good in the beginning, players found that the first solution is too easy but the second is harder than imagine. The first impression make them weaker in thinking solutions. Because there are not many levels, this cannot teach players gradually but is misleading them. If for a social game which have 100 levels, designer could use first 10 or 20 to teach the players gradually with every detail and mechanic. But for an indie game with much lesser levels, this seems not working that well.

Progression is key to the sense of achievement, making player feel that they are stronger and more powerful. But the tutorial level is not suitable for every game. Like in Sekiro: Shadow Die Twice, there are no easy mode or traditional tutorial level for players to learn. Player will learn from their own deaths and failures. In my project, there are no multiplayer content and the length of the game is rather short. So I cancel the so-called tutorial level and try to make each puzzle the same difficulty.

I will develop in this path and later have it playtested.

The idea of making an AI – driven game comes from a simple idea: What if the enemy knows how I beat him before. It gives the enemy AI ability to read the past, so you can call it a meta game as well.

In many AI related games, there are ways for players to cheat. Like in a stealth game, player can walk to some soldiers’ back without being noticed. And a ‘good’ game AI must explain all the time what will they do next to give players enough information. That is not fair with the AI.

The goal of this game is quite simple – beat the enemy certain times. There will be only one scene and one enemy to kill. Seem easy enough? Except that the enemy will know what method you have used so far which means player cannot beat him with the same method twice. That gives a little advantages to the AI.

Also, if possible, the dialogue between AI and player can also be an enjoyable part of the game. Maybe in the end they will become friends rather than stay hostile.

The brief story of the game will be like this: a strange doctor is building a time machine/space ship which can travel through time. And player will act as a brave soldier to stop him, because the world in the future have been destroyed by the machine.

So in developing the idea of this game, a clear scene occurs to my mind: nerdy doctor like Rick in Rick and Morty, messy lab which will give player many items and weapons to use, a time machine. These elements combined will grant creative player ways to beat the boss.

Looking back, I always thought of a scene in the game first rather than mechanics. So sometimes I get extremely frustrated when realizing that I may not be able to rebuild the scene due to lack of certain art assets or programming skills. As an indie developer, I think it is the best to develop your game from game mechanics instead of fancy scenes. In that case, the cost will be acceptable and tasks will be easier.

And in this term’s development, I will try to follow my course mates’ habit – stick to the plan of timeline. Use Trello or other software to record problems, thoughts and maybe possible innovations and further plans.

But let us hope that this simple idea will grow into a complete game.

In game industry, Behaviour tree is commonly used concerning about AI-driven development. but what we learn about AI first is normally Finite State Machine(FSM). In my project, I will use FSM as my method to code the main enemy and there are couple of reasons for this choice. And literature review grants me a clear view of these two main solution.

In my internship with a FPS project, the enemy AI is a critical part of all. What should enemy AI do at this moment? Do they always have the right option? What should they do under complex situations? Or more advanced, do they have their own personality?

The designer I talked to give me a model for these AI-related problems. They called it the OODA loop – Observe, Orient, Decide and Act. That is from strategist and U.S. Air Force Colonel John Boyd[3]. But this procedure is for fighter pilots, which cannot be directly used as an AI judgment model. In short, there will be a blackboard or Director AI to do the Observe and part of Orient job. And in actual contact, each AI-driven enemy will do the rest. Limitation of information input of each enemy will ensure the flexibility of changing circumstances and make them feel “real”. And yes, there are different personality for each enemy which result in various actions facing players.

All actions above is implemented by the Behavior Tree. Here I will not fully explain how exactly behavior tree works. Tree in computer science is a data structure which provide a way to store information. And behavior tree is more like judgment procedure using the Tree structure. There are father nodes and children nodes. Decisions will only goes from father nodes to children nodes all the way down to minimal nodes, which is basic actions like running or shooting. And in different conditions, father nodes will choose different child node to make the right choice.

In comparison, there are finite state machine to operate this procedure easier to understand.

FSM is also a concept in computer science. Each state represent a situation where the AI stays in. And with factors changes, the current state will change to another like from running to standing. There are also many conditions that helps FSM to decide.

Differences

And the main difference between these two methods is extendibility. In FSM, each state added will bring huge pressure to calculation and storage part. From description above, we can conclude that for N states in total, there will be N^2 conditional judgments max in one tick. And also it is heavy coding for developers.

There are also problems in behavior tree. The biggest one would be that different trees might conflict with each other. Like in what condition, should the AI run from players instead of shooting players? Director AI might have a decision but enemy AI have another. So in actual development, there are more structure added to avoid conflictions. Also the behavior trees will consume huge amount of memory and calculation ability when they are numerous. But there are also ways to avoid that.

Conclusion

In conclusion, in AAA project in game industry, the behavior tree will be used frequently due to its good extendibility. The designer need no programming skills to arrange fixed tree nodes. Coding team can focus more on basic nodes development. But for smaller team or project, FSM is also acceptable for simple AI enemy. And it is also convenient for indie developers who act as programmer and designer simultaneously. For my project, I will choose FSM and see how it goes. If there are problems in maintaining, I will switch to behavior tree structure instead.

Bibliography
Angerman, W.S., 2004. Coming full circle with Boyd’s OODA loop ideas: An analysis of innovation diffusion and evolution.

Saini, S., Chung, P.W.H. and Dawson, C.W., 2011, August. Mimicking human strategies in fighting games using a data driven finite state machine. In 2011 6th IEEE Joint International Information Technology and Artificial Intelligence Conference (Vol. 2, pp. 389-393). IEEE.

Nicolau, M., Perez-Liebana, D., O’Neill, M. and Brabazon, A., 2016. Evolutionary behavior tree approaches for navigating platform games. IEEE Transactions on Computational Intelligence and AI in Games, 9(3), pp.227-238