Teaching AI-Based Game Design

[Music] so my name is Myriam Paulo sorry Allan hari and I'm a game designer I just started in the studio just a few days ago I'm also a researcher affiliated to the University of Malta and to Stockholm University in Sweden so I've been teaching games for for a while since 2004 and this talk is a case study of of a course concerned with AI based game design and computational expression and the reason that I want to give this talk is to show you how it's possible to enable students to expand the technological palette how they can make AI technologies as part of their artistic practice as designers it's

something that's super empowering to know that that you have lots of technologies at your fingertips and for those who are already proficient they will in coding and using these technologies it's also a way to show that ok so this is the shorter territory and that makes it possible to see what is uncharted and where you can do innovations and new things in designing this course I wanted to make it accessible for both programmers and non-programmers so they kind of saw the mix of technologies that we have been using are things that that can be used as off-the-shelf even if you don't know how to program however the students that I've had are mostly programmers just shortly with AI based

game design means it's simply when the AI is in the foreground and some calls for designing a game simply said so back in 2011 I was spending a lot of time in Santa Cruz and having lots of good discussion with some of my favorite collaborators Josh McCoy Julian Smith and Anne Sullivan and we started to compare how do we work with AI and design and we could see commonalities first of all we could see that we usually kind of looked at three different knowledge domains all across that the case studies that we did usually having one rich metaphor of a system like personality psychology or social behavior and then some sort of either existing AI technology or

whatever we were building and also the type of genre that we were kind of adhering to when we were doing our development if any we also kind of saw that back then the main challenges that we were facing in our own practice was how to make the the AI really be foregrounded how how do you make it transparent to the player that something interesting is happening so that you don't have a super tailspin life system where you don't see what's on the surface and how it's played another one was looking at particular sweet spots for emergence again it is fun and understandable but still rich and something that you can explore so we reasoned a lot about that and then when I designed the course computational expression I try to find a

way to give this to the students like at the same time give them a method for how to work with it and design and also give them a palette of AI method but not only AI methods also some interesting into technologies and saying like get oculus or or connect to things like that and practically about the course it's a five ec TS course and in Europe this means that 180 TS is 25 hours of student work and one semester is 30 ECT SS in total so this was a six of a semester time and the semester is four in the house month-long so the students that I have they were from this was given at the University of Malta mostly they were bachelors in computing

science but also a few of them didn't have programming backgrounds they also already knew how to aggressively design and develop games because they were students in our masters game design master of course we used the great book expressive processing by no awarded proven as a kind of redline in our tech seminars and and we had workshops and some guest lectures and this is kind of what the press cannot see that the course was divided into two main parts where the first six weeks but it was an intense period of them of knowledge gathering so having seminars and workshops and this is kind of like the things which are kind of on that top top top row there those were the kind of the

technological poletik i wanted to give my students i mean whoever kind of like when you use this you use the things which you think is okay what are the technologies that exist and could be useful for game design and these were the ones that we cannot look at basically having AI based game design that was one of the starting points and also looking at the methods for research prototyping like how can you answer a research question by posing it as if as a prototype looking at agents and characters interactive narrative systems for that game worlds how you build the full rich ontology for game systems for health healthy places looking as procedurally generated content methods data mining big data and then also the kind of the tool box on the toy box using things

like bio data and other types of inputs like like the connector or things like that mainly looking at a huge area of the present research and emergence and this was kind of also interspersed with hands-on workshops when students use these technologies so and at the end of this six-week period the students delivered a game concept like what are we going to build when we have the tech seminars we divided up all the readings among the group so each students picked the technology he or she was most interested in and then champion that and kind of held that seminar and presented the text to or or the tool or the approach to the group and we had a discussion about that but in addition to just reading papers I invited the guest

speakers who could give demonstrations like okay Richard Evans sketched in and said this is how I developed versus this is how I think about narrative when I build a system it was also a good opportunity for the students to ask about basic TBI agents architectures Gillian Smith showed her prototype the endless web and students also read her publications so that when she tried to sketch then they could both kind of get the demonstration of her work but also ask questions about their academic publications on that Bryan McGuirk Oh showed his viewpoints AI where we're advancing closest player is kind of represented in a very artistic manner also an interesting way to use computation for artistic expression we

had Noah also skyping in if we had his book as the red thread and the talk if in the in the whole course so mainly talking about the different effects like how you kind of like how you can use the Eliza effect how how you try to not be in the tailspin effect when when this you're super impressed and fantastic AI system is not really visible to the player even though it's a fantastically beautiful system and instead use the sim citty effect where you have this rich metaphor really feeding into the design and making it accessible for it for the player in the workshops we we were both looking at digital systems but we were also doing a paper prototyping AI and this was especially to make it accessible for non-programmers like if you do a paper prototyping of the strict

system for example building building a story together in groups you literally can't walk out of the room not understanding plan based problem solvers it's a it's something that I really kind of recommend as doing when you teach AI to also combine it with these kind of more tangible hands and methods for teaching so the the tasks that that the that the students got was two when they brainstormed look at choose one AI technology from from the Colet and choose one rich metaphor from real life so those were two starting points the third one looking at the kind of the game conventions in the design it was something that they should be aware of so that they so if they were using game

design conventions they would be kind of know when they made something innovative the development process it was iterative they already knew how what cycles to go through with play testing and iteration so they cannot did that naturally I didn't have to teach that part in the course so I'm going to show a few quick examples of games so one this is highway code it's a small racing game where the students use machine learning as the central technology and use kind of traffic systems on Malta as their main knowledge domain so in the game that the cars learned how to have the kind of the right speed console blocks is another game that the students made were we're

using procedural generated platforms for music composition so so just playing by playing the game you would compose music as you went along and the little and the little platforms they're kind of they emerged in real time during play so that's also very in line the physical prototypes were made you know in just a few days of time they didn't have much time for this kind of development another example is organic Tron where students use PCG in in combination with the genetic algorithms to make a robot work a strategy game that was played by Titian share the keyboard and kind of watch their little robot evolve it was very fun another one was dungeon crawlers

procedurally generating so it was kind of this mix the mixed-media thing that you would get support for how to build your dungeons for your kind of board game bros like that you would say so this was looking at quest flags and PCG quest so they were also to a couple of people groups used in such technologies instead of AI as a central part which is totally viable as well at least within the scope of this course so once were using the gyro as the central thing and the hero class messed with with being attacked by Birds mounting mounting the a cell phone using the gyro to play a level of that another was using multi device to kind of have to play to play a kind of arts game so

to win your squigglies need to kind of traverse to the end of the screen and use have have a kind of art piece being created as you play and they were inspired by a minimalist art and use the yellowtail codebase for doing that and do their thing so the resulting things this is kind of like the competitive mode or the fight mode and this would be the kind of result of that play those the collaborative play part what we're resulting in in in in images that can looked at it like this [Music] so after this month of prototyping there was a development stop in order to

reflect this was a very important part of the course because you can't really make sense of all these technologies and unless you get time to really reflect about how can you use this in your own artistic practice as a designer in the future and some of the questions they were looking at was kind of going back to how the technology and the design affects each other so looking at the course structure that left the three ending months of the course after the knowledge gathering phase was spent one month prototyping and testing then a full reflection phase where people wrote reports and played each other's games but then also making sure that they had a full month for polish so that the feedback that they got from each other and for me was something that they could use in their absolute last iteration the

what what I saw as a teacher which was what was strangely obvious was that those groups and students whose who were really embracing the idea to have a rich metaphor from real life that they were kind of using in their design work we're the ones who made the most innovative and in games during this short at that time that they have another thing that that I saw was that because I was using this kind of traditional style defense having kind of people defend their argument in Class A combining that academic rigor together with practical work was something that it general fed into each other nicely the outcomes for the students was that they both games to put into their portfolios some good seeds for research papers that they then were writing it was also kind of feasibility

test for their forthcoming thesis work something that would spend a year doing so that is kind of a time-saver there and an important thing too was to have a hands-on experience of what it means to build an AI module specifically for a game because the process of iterations together with the game design is really fun but it's also resource heavy and this is something it which is very good to know when you go out in the industry that is not you know sometimes it's a super good idea but maybe not always but most important was that the students have really gotten an idea about Apple s for how to use technology and also see where where can we innovate and where can we do new things that doesn't exist already the course materials and stuff

it's possible to download I just threw out all the PDFs and things on that address so if anyone is interested it is all there I'll put the slides there as well and that's it question [Applause] [Music] [Applause] [Music] [Applause] hi thank you for a talk if I understand well in this course they have to program their own AI we have other course for the same student when the user found out the truth AI you find in engines and using and natural to respond yes in addition in the program there is a

several game AI courses and also PCG courses for those who do the technologies from so so they already most of the students already have lots of things general technologies on their fingertips that they can use so this is long this course doesn't exist in isolation okay thank you so much for listening [Applause] [Music] [Applause]