Abstract
This paper describes the game engine of the forthcoming game generation system titled The Game-O-Mat. As underlying game engines strongly constrain the space of possible games that a generator can create, the goal of this paper is to fully document the design considerations and features of this game engine that was designed with game generation as a priority. This paper is aimed at helping future developers of game generation systems.

Abstract
Non-representational art-such as works by Wassily Kandinsky, Joan Mitchell, Willem de Kooning, etc.-showcases diverse artistic expressions and challenges viewers with its interpretive open-endedness and lack of a clear mapping to our everyday reality. Human cognition and perception nonetheless aid us in making sense of, reasoning about, and discussing the perceptual features prevalent in such non-representational art. While there have been various Computational Creative systems capable of generating representational artwork, only a few existing Computational (Co)Creative systems for visual arts can produce non-representational art. How would a co-creative AI that incorporates elements of the human visual perception theory be able to collaborate with a human in co-creating a non-representational art? This paper explores this challenge in detail, describes potential machine learning and non-machine learning approaches for designing an AI agent and introduces a new web-based, multi-agent AI drawing application, called Drawcto, capable of co-creating non-representational artwork with human collaborators.

Abstract
Generative Graph grammars are an established technique for procedural content generation (PCG) of the mission space of dungeon levels, as they allow for explicit specification of design intention. However, they are difficult to control beyond the initial specification. Defining effective grammars requires both design expertise and familiarity with how grammars work, and even then the resulting grammar is not guaranteed to generate missions that fit a designer's intention. In this paper, we propose a system that attempts to allow designers to affect an existing grammar's generative space by allowing designers to control the probability distribution induced by the grammar. The system does this by learning the parameters of a probabilistic graph grammar from examples. Designers can control the generation of these examples via specification of assessment criteria, and a threshold above or below which the output is generated. We conclude with a demonstration of the efficacy of the system in shifting the distribution of the generative space.

Abstract
The Holodeck, a virtual reality simulator from the television show Star Trek, is known as the 'holy grail' of interactive narrative experiences. However, while there have been approaches to various components of a theoretical Holodeck, scene generation from dialogue is often overlooked. This paper introduces a prototype AI Holodeck application for scene generation, demonstrating the use of Natural Language Processing and a corpus of spatial data. The application creates scenes from user input text and fills those scenes with objects and relationships not explicitly defined by the user. This paper discusses potential use cases of scene generation in creating environments for interactive narrative, virtual reality, and other development opportunities.

Abstract
Behavior trees (BTs) are a popular method for modeling NPC and enemy AI behavior and have been widely used in commercial games. In this work, rather than use BTs to model game playing agents, we use them for modeling game design agents, defining behaviors as content generation tasks rather than in-game actions. Similar to how traditional BTs enable modeling behaviors in a modular and dynamic manner, BTs for PCG enable simple subtrees for generating parts of levels to be combined modularly to form complex trees for generating whole levels as well as generators that can dynamically vary the generated content. We refer to this approach as Procedural Content Generation using Behavior Trees, or PCGBT, and demonstrate it by using BTs to model generators for Super Mario Bros., Mega Man and Metroid levels as well as dungeon layouts and discuss several ways in which this paradigm could be applied and extended in the future.

Abstract
Procedural content generation (PCG) has seen relative widespread adoption in games and game development. While PCG methods are gaining traction in commercial games, there are comparatively fewer cases where PCG has been used to generate the game world, characters, and the narrative for a single game. We present our research, collaborating with a small independent game studio, on the design of a game engine designed to generate a world, characters, supporting quests and overarching narrative content for their upcoming game. We discuss our choice of game artificial intelligence algorithms, decisions made in designing our system, and our progress and implementation under the constraints and requirements given to us. We posit our system will generate a rich game world with an explainable world history for players to explore and quests to complete. We believe that our engine will meet their requirements increasing the replayability of the game, reducing the authorship burden, while providing the authors with significant control over quest and narrative structure.

Abstract
Procedural content generation (PCG) algorithms pervade the world of game development. They are useful tools that allow creators to unshoulder a degree of authorial burden of non-trivial virtual worlds. Such a goal necessitates a well-defined model of what the algorithm is producing. In this work, a PCG system called TileTerror is presented that uses a model of horror constructed from a ludological analysis of well-known horror games to embed horror story features in a procedurally generated 2D tilemap. The horror features found in the ludology are separated into a distinct hierarchy: low-level features and high-level features. These features are used in conjunction with a procedurally generated tilemap to create an annotated map denoting candidate locations for strong horror moments. This system is evaluated using a series of pathfinding agents which score the annotated tilemap. TileTerror represents a first step into exploring what story elements can be embedded in procedurally generated constructs like 2D tilemaps.

Abstract
AgentCraft is an entry for the Generative Design in Minecraft (GDMC) AI Settlement Generation Competition, including the optional Chronicle Generation challenge. It makes use of an agent-based simulation to organically produce a plausible settlement and a written history of the settlement's development. Uniquely, AgentCraft utilizes the HTTP Server version of the competition framework to show the agents constructing the settlement in real time-a visual technique that can't be achieved with the earlier MCEdit framework. In this paper, we aim to provide a point of reference for future agent-based settlement generators by describing how our competition entry works and discussing the benefits and downsides of the agent-based approach. Additionally, we propose a new optional challenge for the GDMC competition, centering on the development of settlement simulations whose inhabitants can be directly observed or interacted with by the player.

Abstract
Current agents use communication in a collaborative setting, exchanging truthful information to achieve a common plan. This paper defines games where agents may exchange information about the physical situation (both fluents and action events), arbitrarily nested beliefs, and action consequences, to manipulate other agents for their own goals, i.e. guide the other agents' own reasoning and planning. We propose a model for an 'agent mind' that can cater for all these aspects through revisable, prioritized belief bases; goal recognition including epistemic situations; or planning including speech acts with structured content. We also discuss recent algorithms to address each one of them, and propose a concrete implementation for a future stage.

Abstract
Games currently feature simple models of morality and moral reasoning. These typically take the form of a simple binary opposition between good and evil, in which game actions are sorted into these categories, and reputation systems that track the reputation of players (and sometimes NPCs) with in-game factions based on the actions taken. This paper presents a more sophisticated model of moral reasoning based on Lakoff's metaphorical family models: the strict father and nurturant parent morality. This model utilizes more rapid, surface level categorization of different situations in moral categories, and deeper reasoning that characterizes situations in terms of their relationship to moral virtues as determined by the metaphor value system. This model is used in the experimental prototype Argument Box (AB), a social argument simulator where the player argues with clients visiting the shop. Arguments in AB center around the moral virtues and vices of simulated characters in the social simulation Talk Of the Town. This paper presents the current architecture, discussing its technical details.

Abstract
Data is becoming an important central point for making design decisions for most software. Game development is not an exception. As data-driven methods and systems start to populate these environments, a good question is: can we make models developed from this data transparent to users? In this paper, we synthesize existing work from the Intelligent User Interface and Learning Science research communities, where they started to investigate the potential of making such data and models available to users. We then present a new area exploring this question, which we call Open Player Modeling, as an emerging research area. We define the design space of Open Player Models and present exciting open problems that the games research community can explore. We conclude the paper with a case study and discuss the potential value of this approach.

Abstract
Artificial Intelligence has been applied to many facets of the game design and development process. Though this has led to many advances in games and AI research, there remain few examples of games in which play centers around engagement with AI processes: the design space of AI-based games remains underexplored. By examining a breadth of playful experiences through different lenses, it is determined that games which forefront AI are beneficial for players, designers, and for the field of game scholarship itself. Moreover, there is evidence that symbolic approaches (rather than statistical) lend themselves to experiences with more agency, greater human interpretability, and more controlled authorability.

Abstract
This work aims to present the principles from the game design perspective and lessons learned of two techniques adopted for solving the NPC behavior in casual language-oriented games in a Reinforcement Learning discrete and partially observable environment. It might be useful for other game developers as it offers an example of designing companionship in NPCs from the game perspective and to speed up their developments for implementing machine learning in NPC players, showing that a designed heuristic function and Imitation Learning approach can speed up developments with respect to a Reinforcement Learning approach for a deterministic output.

Abstract
'Tokamak Elementary' is a game built around a custom conversational engine. The player can talk with 'Toki', the AI-controlled character as with a chatbot. Toki is a gentle spirit creature summoned from the world of spirits. As such, Toki starts from a mostly clean slate and it's up to the player to teach them basic concepts of our world (such as what animals are) and how to interact with humans (such as returning greetings).

Abstract
Tabletop games come in a variety of forms, including board games, card games, and dice games. In recent years, their complexity has considerably increased, with many components, rules that change dynamically through the game, diverse player roles, and a series of control parameters that influence a game's balance. As such, they also encompass novel and intricate challenges for Artificial Intelligence methods, yet research largely focuses on classical board games such as chess and Go. We introduce in this work the Tabletop Games (TAG) framework, which promotes research into general AI in modern tabletop games, facilitating the implementation of new games and AI players, while providing analytics to capture the complexities of the challenges proposed. We include preliminary results with sample AI players, showing some moderate success, with plenty of room for improvement, and discuss further developments and new research directions.

Abstract
Procedural content generation via Machine Learning (PCGML) creates game levels from examples. PCGML for platformers with physics-based player movement generally has not guaranteed the reachability of goals, relying on post-generation filtering approaches or game-specific heuristic rules. In contrast, constraint-based PCGML can provide gameplay guarantees, but has typically been applied to games with simple grid-based movement. In this work, we present a constraint-based PCGML approach for platformers with physics-based player movement that guarantees playability and provides design controllability. Our approach exhaustively precomputes all possible player movement states in example tile-based platformer levels. It extracts metatiles containing local state information and uses them in constraint-based level generation that ensures legal tile neighbors and consistent movement state transitions. The approach can ensure constraints on gameplay, such as the reachability of goals, guaranteeing level playability, and other elements like platforms and bonuses. It can also incorporate other designer-controllable constraints.

Abstract
Prior research has shown variational autoencoders (VAEs) to be useful for generating and blending game levels by learning latent representations of existing level data. We build on such models by exploring the level design affordances and applications enabled by conditional VAEs (CVAEs). CVAEs augment VAEs by allowing them to be trained using labeled data, thus enabling outputs to be generated conditioned on some input. We studied how increased control in the level generation process and the ability to produce desired outputs via training on labeled game level data could build on prior PCGML methods. Through our results of training CVAEs on levels from Super Mario Bros., Kid Icarus and Mega Man, we show that such models can assist in level design by generating levels with desired level elements and patterns as well as producing blended levels with desired combinations of games.

Abstract
Several recent PCGML methods have focused on generating game levels and content that blend the properties of multiple games. However, these works ignore the fact that blended levels must in some way have blended physics models that enable playable levels. In this work, we present an approach for extracting jump physics models for such blended game domains. We make use of variational autoencoders (VAEs) trained on level data from six platformers, encoded using a previously introduced path and affordance vocabulary. Our results show that the extraction model is able to reasonably recreate the original physics models when given ground truth paths, and is able to produce physics models that can reliably allow an agent to play the generated levels. We also find that there are promising results for blended physics models behaving intuitively between physics models of the original games being blended.

Abstract
In this paper we describe the Sharp Multi-input Editing Software (SMES), a prototype system which allows players to convert dexterity challenges to planning challenges in existing games. SMES accomplishes this by recording the player input, displaying it in a visual form, allowing the player to edit the recorded playtrace, and replay the edited sequence back again. Our system combines ideas from tool-assisted speedruns (TAS) with interaction design patterns from Casual Creators and Accessible Player Experiences. Because our implementation is based on intercepting input events at the operating system level, no engineering effort is needed to apply the tool to new games. The result is a new mode of play for existing games that further allows players to realize new behaviors in the game. SMES could support making games more inclusive by letting players adapt the challenge type they face. Furthermore we claim that AI systems that operate over playtraces to assist the player is a rich avenue to explore.

Abstract
Procedural Content Generation via Machine Learning (PCGML) refers to a group of methods for creating game content (e.g. platformer levels, game maps, etc.) using machine learning models. PCGML approaches rely on black box models, which can be difficult to understand and debug by human designers who do not have expert knowledge about machine learning. This can be even more tricky in co-creative systems where human designers must interact with AI agents to generate game content. In this paper we present an approach to explainable artificial intelligence in which certain training instances are offered to human users as an explanation for the AI agent's actions during a co-creation process. We evaluate this approach by approximating its ability to provide human users with the explanations of AI agent's actions and helping them to more efficiently cooperate with the AI agent.

Abstract
A key concern for procedural level generation is ensuring that generated levels are in fact playable. One approach for determining playability is using pathfinding agents to test if generated levels can be completed. This typically involves following an iterative loop of generating candidate levels until the agent is able to successfully complete a playthrough, discarding the failed level and starting the generation process over each time the agent fails. In this paper, we present a new approach for generating playable levels that leverages pathfinding agents, not to simply test levels for playability, but to perform level repair to fix generated levels that are unplayable. By augmenting a simple pathfinding agent to be able to take moves that would require modifying the level, our approach can fix unplayable levels without having to regenerate a new level from scratch. We compared our repair agent to a default agent on three PCGML level generators each for Super Mario Bros. and Kid Icarus, and found the repair agent was able to fix the majority of levels, taking time that ranged from similar to the default agent to around 4x its time.

Abstract
In this paper, we describe an approach to procedurally generating abridged game levels. We present a design pattern-based method for abridging a collection of Super Mario Bros. levels. These design patterns are salient regions that users describe as memorable after playing original Mario levels. We use the Mario AI Framework and Mawhorter and Mateas's Occupancy-Regulated Extension (ORE) generator to generate 20 abridged levels based upon 3 original levels from Super Mario Bros. We conduct a preliminary study for 2 abridged levels and compare the play experience with the original levels' play experience. We also compare these abridged levels to those created by previous level generators using the same patterns. Finally, we offer a critical reflection of the process taken to generate these levels, and the assumptions and values inherent in this design pattern-based approach.

Abstract
Procedural Content Generation is widely used in games, however, its use in educational puzzle games has been limited. These types of games present common challenges such as solvability and non triviality, but also the extra challenge of preserving intended learning goals. In this paper, we present a modular constructive approach to generate levels in a puzzle programming game. The approach uses a grammar to generate game elements from code and works backwards from the solution to ensure solvability, controllability over the solution, and variation, allowing for alternative solutions that preserve the learning goals.

Abstract
Procedural content generation via machine learning (PCGML) has shown success at producing new video game content with machine learning. However, the majority of the work has focused on the production of static game content, including game levels and visual elements. There has been much less work on dynamic game content, such as game mechanics. One reason for this is the lack of a consistent representation for dynamic game content, which is key for a number of statistical machine learning approaches. We present an autoencoder for deriving what we call 'entity embeddings', a consistent way to represent different dynamic entities across multiple games in the same representation. In this paper we introduce the learned representation, along with some evidence towards its quality and future utility.

Abstract
In game development, designing compelling visual assets that convey gameplay-relevant features requires time and experience. Recent image generation methods that create high-quality content could reduce development costs, but these approaches do not consider game mechanics. We propose a Convolutional Variational Autoencoder (CVAE) system to modify and generate new game visuals based on their gameplay relevance. We test this approach with Pokemon sprites and Pokemon type information, since types are one of the game's core mechanics and they directly impact the game's visuals. Our experimental results indicate that adopting a transfer learning approach can help to improve visual quality and stability over unseen data.

Abstract
Scenes in 2D videogames generally consist of a static terrain and a set of dynamic sprites which move around freely. AI systems that aim to understand game rules (for design support or automated gameplay) must be able to distinguish moving elements from the background. To this end, we re-purposed an object detection model from deep learning literature, developing along the way YOLO Artificial Retro-game Data Synthesizer, or YARDS, which efficiently produces semi-realistic, retro-game sprite detection datasets without manual labeling. Provided with sprites, background images, and a set of parameters, the package uses sprite frequency spaces to create synthetic gameplay images along with their corresponding labels.

Abstract
Many recent generative text systems combine a context-free grammar base with some set of extensions such as tagging or inline JavaScript. We argue that the restriction to CFGs is unnecessary and that the standard pattern-directed, nondeterministic control structures common to Prolog, definite-clause grammars, and many HTNs, support a superset of these capabilities while still being simple to implement. We describe Step, a generative text system for Unity games based on higher-order HTNs that so far as we can determine from published descriptions, generalizes this previous work. We then describe syntactic extensions to make Step more natural as a text-authoring language. Finally, we discuss how Step and similar systems can be implemented very compactly in modern mainstream programming languages.

Abstract
Dynamic quest generation is a topic of interest in the game AI community, with examples appearing in AAA games (Lenhardt 2012), indie games (Coxon 2016), and academia (Sullivan et al. 2012). In July 2018, we released an update to our simulative science fantasy roguelike Caves of Qud (Freehold Games 2015) that added its own take on dynamically generated villages and quests. These systems were built on top of a framework we previously developed for generating historical biographies (Grinblat and Bucklew 2017). Like that framework, our approach to dynamic quest generation-novel in the space as far as we know-was shaped primarily by two design forces at work in the project's development. The first is an architectural constraint; to support its large-scale game worlds, Caves of Qud uses a two-phase approach to world creation (Bucklew and Grinblat 2019) that delays the full fabrication of game 'zones' until the player enters them. The quest generation system was necessarily designed to fit this two-tiered architecture. The second is an aesthetic choice of procedural content design; we tend to be nonprescriptive in deciding which of our content modules can be procedurally combined, preferring instead to 'let our generators run wild,' as we say, but ensure that they smooth the rough edges off their output. For our storyful generators, this means eschewing causal logic in favor of randomness that's parameterized by the richly narrative world the generators are embedded in. The resulting outputs are ripe for apophenic readings.

Abstract
Stealth games challenge players to complete tasks while remaining unnoticed by enemies. Modern versions often provide players with the ability to silently eliminate enemies, inducing a puzzle-like nature to the task which combines finding a path to the goal with the need to discover an order in which enemies can be eliminated. This 'take-down' problem is made more complex by the need to hide the resulting enemy body from future observation. We present a version of a sample-based search that can find paths respecting the stealth constraints of both the take-down and body-hiding mechanics, including multiple variations that address different ways these problems are expressed in games. As a perhaps surprising outcome of our analysis, we show that an additional body-hiding problem can improve the success rate, despite the apparent increase in task complexity.

Abstract
In this paper we introduce DeepCrawl, a fully-playable Rogue-like prototype for iOS and Android in which all agents are controlled by policy networks trained using Deep Reinforcement Learning (DRL). Our aim is to understand whether recent advances in DRL can be used to develop convincing behavioral models for non-player characters in videogames. We begin with an analysis of requirements that such an AI system should satisfy in order to be practically applicable in video game development, and identify the elements of the DRL model used in the DeepCrawl prototype. The successes and limitations of DeepCrawl are documented through a series of playability tests performed on the final game. We believe that the techniques we propose offer insight into innovative new avenues for the development of behaviors for non-player characters in video games, as they offer the potential to overcome critical issues with classical approaches.

Abstract
Motivation establishes the framework within which goals are pursued. Different characters are motivated differently based on their personality, emotions, relationships, and social norms. In this paper, we propose a motivation model that incorporates these concepts into multi-agent narrative planning. Using our model, characters automatically adopt new goals, find plans to achieve those goals, and choose between those plans based on their personality.

Abstract
Tarot decks appeared as early as the 14th century. Once used as a game deck, today it is primarily used as a divination medium. One attempt to adapt the fortune-telling claims to modern technological pragmatism: the selection of card imagery are a representation of the omni-homo, or the every-human, or a collective consciousness, such that a card reading session becomes an interactive resurfacing of subconscious, but preexisting, knowledge. We suggest that perhaps when we create Tarot decks we are also creating a representation of the collective consciousness, in a manner of speaking. We provide a procedure of doing this creation from text-mining and take a look at the resulting artifact.

Abstract
Let's Plays of video games represent a relatively unexplored area for experimental AI in games. In this short paper, we discuss an approach to generate automated commentary for Let's Play videos, drawing on convolutional deep neural networks. We focus on Let's Plays of the popular game Minecraft. We compare our approach and a prior approach and demonstrate the generation of automated, artificial commentary.

Abstract
Automated feedback has the potential to provide significant assistance to student game creators. Here, we present a system for generating automated, critiquelike feedback for students creating games in the Study-Crafter platform. We implemented a system that builds a personalized feedback report for students based on a templated format. This critique uses automated analysis of structural and interactive aspects of the game narrative and recommends alternate games for students to examine as inspiration. To test our system, we conducted a pilot study with 10 student groups developing narrative-based games. A key understanding from the study is that determining the appropriate depth of assessment and critique without overwhelming the student is important.

Abstract
Visualizing game mechanics is an important way for game developers to refine, review, and improve their work, and to facilitate communication between designers and programmers. This challenge has seen many different manual and automatic approaches, but they have been either too abstract and distant from the game, too close to the actual code, or lacking in connections to the code. We propose a modular system that can semi-automatically build and display a graphical representation of a given game's mechanics, to stimulate reasoning about the dependencies between the game's variables. We present an initial implementation of this visualization tool, which can read a graph from a user-made file and partition it based on user parameters. We also discuss a pilot evaluation of the tool's capacities for creative support.

Abstract
Cooperation between human players and AI agents in games is a subject of great interest in the research community. In this paper we propose a new domain as a challenge for future AI research: the cooperative game Pandemic. This game represents a challenge because it requires cooperation between players in an environment with incomplete information. Additionally, we propose a first approach for a cooperative agent for the game, that uses planning in conjunction with plan recognition. We also propose an experiment to test the mentioned approach. In this paper we argue why Pandemic makes a compelling domain for AI research, the status of our project, as well as which challenges remain to be addressed.

Abstract
One of the key benefits of a machine learning system is the fact that it can adapt, change, and learn with more experience. However, pre-trained algorithms forfeit this benefit, because continued play does not continue to impact the AI's performance. With Lab Assistant, the team had two goals: to include human-driven training as part of the game loop, and to create an emotional bond between the artificial intelligence and the player. The game can be accessed for free at the website: https://lizfiacco.itch.io/lab-assistant.

Abstract
Procedurally generating high-quality content is an important problem in the design and development of video games. A key objective of procedural level generation is controllability: the ability to generate novel levels with a predetermined goal, such as maximizing difficulty. Previous work has achieved this for platformer games using deep-learning and evolutionary methods, but such methods often lead to invalid or asymmetric structures within the generated levels. On the other hand, generative approaches using LSTMs and Markov chains have succeeded in preserving local coherence but suffer from limited controllability. In this paper, we introduce a framework for generating controllable and locally coherent levels in platformer games. Our framework extends a prior GAN and evolutionary strategy-based approach with LSTM Recurrent Neural Networks and hyperparameter optimization. We use sequence transformation via a LSTM to improve aesthetic coherency, while tuning fitness function hyperparameters for specific objectives and measuring overall level quality through human evaluation. To verify our approach, we create levels for Super Mario Bros intended to be novel and challenging to expert players. We demonstrate that the framework is able to successfully generate such levels in a controllable manner, while also resolving invalid and unaesthetic local structures.

Abstract
In level co-creation an AI and human work together to create a video game level. One open challenge in level co-creation is how to empower human users to ensure particular qualities of the final level, such as challenge. There has been significant prior research into automated pathing and automated playtesting for video game levels, but not in how to incorporate these into tools. In this demonstration we present an improvement of the Morai Maker mixed-initiative level editor for Super Mario Bros. that includes automated pathing and challenge approximation features.

Abstract
Previous work explored blending levels from existing games to create levels for a new game that mixes properties of the original games. In this paper, we use Variational Autoencoders (VAEs) for improving upon such techniques. VAEs are artificial neural networks that learn and use latent representations of datasets to generate novel outputs. We train a VAE on level data from Super Mario Bros. and Kid Icarus, enabling it to capture the latent space spanning both games. We then use this space to generate level segments that combine properties of levels from both games. Moreover, by applying evolutionary search in the latent space, we evolve level segments satisfying specific constraints. We argue that these affordances make the VAE-based approach especially suitable for co-creative level design and compare its performance with similar generative models like the GAN and the VAE-GAN.

Abstract
Procedural content generation has been widely used for aiding game content designers in the content creation process. However, limited work has focused on autonomously generating novel physical gestures for non-player characters. This paper focuses on using creative strategies from theater and dance to enable the generation of novel movement sequences for characters in 3D game environments. We apply the Space component from Laban's Movement Analysis to procedurally improvise movements in virtual environments. Our framework outlines a pipeline for an improvisational AI that analyzes human movements in terms of the geometric properties of Laban's icosahedron and manipulates these projections to generate various alternative movements. In this paper, we mainly focus on describing the structural design and encoding scheme used in the framework.

Abstract
Constraint programming offers a concise and powerful approach to a variety of procedural content generation problems (A. M. Smith & Mateas, 2011). However, it requires considerable technical expertise. No constraint-based PCG languages come close to Compton's goal of casual creators (Compton & Mateas, 2015). In this paper, we describe Imaginarium, an interactive environment for developing ontologies for constraint-based PCG, targeted toward players of tabletop role-playing games. The system allows players to interactively describe the ontologies of their games using an Inform-like English subset, and explore the spaces of artifacts they generate. While much less expressive than a full logic programming language, it is quite competitive in terms of conciseness and readability.

Abstract
We introduce gym-city, a Reinforcement Learning environment that uses SimCity 1's game engine to simulate an urban environment, wherein agents might seek to optimize one or a combination of any number of city-wide metrics, on gameboards of various sizes. We focus on population, and analyze our agents' ability to generalize to larger map-sizes than those seen during training. The environment is interactive, allowing a human player to build alongside agents during training and inference, potentially influencing the course of their learning, or manually probing and evaluating their performance. To test our agents' ability to capture distance-agnostic relationships between elements of the gameboard, we design a minigame within the environment which is, by design, unsolvable at large enough scales given strictly local strategies. Given the game engine's extensive use of Cellular Automata, we also train our agents to 'play' Conway's Game of Life - again optimizing for population - and examine their behaviour at multiple scales. To make our models compatible with variable-scale gameplay, we use Neural Networks with recursive weights and structure - fractals to be truncated at different depths, dependent upon the size of the gameboard.

Abstract
'Bosses' - powerful, difficult enemies - have been a part of video games for the majority of their existence. Despite their integral role in many games, they have rarely been the focus of study. Cuphead is a run-and-gun style 2D action game with a large pool of boss battles, making it an ideal game for comparing and contrasting bosses, to determine what makes a compelling (or frustrating) boss battle. In this case study, we developed an ontology of 'shoot-em-up'-genre attacks and bosses. With this system for codifying a boss battle, we cluster the attacks using a Gaussian Mixture Model - which are then used to represent a boss as a 'bag-of-attacks'. We then use multinomial regression to predict the player experience of a boss given the parameterized boss.

Abstract
Videogames are created for human players whose common-sense knowledge of real-world objects and interactions (and their familiarity with other games) primes them for successful play. Action games feature recurring formal elements including a directly controlled avatar, moving enemies, resource pickups, and portals to new map areas; mapping these onto culturally significant symbols helps players learn to play quickly. We present a schema, annotation tool, and dataset for codifying screenshots containing game objects in terms of their affordances, which is suitable for AI agents and machine learning algorithms for a variety of interesting and significant applications.

Abstract
As a field of study, Experience Management spans a set of technology that is becoming increasingly relevant in applications that aim to improve the experiences of their users. Given the youth of the field, however, few attempts have been made to identify and discuss the common elements of systems that manage user experiences as they occur. In this paper, we consider a subset of existing AI experience managers in the context of a shared conceptual framework. We offer directly comparable summaries of the managers that we discuss, and we highlight how some of them use different technologies to perform similar tasks. While many experience managers remain undiscussed, we nonetheless demonstrate that distinct, well-defined components exist across a diverse set of managers, and that a modular method for building new managers might well be within reach.

Abstract
Deceptive games are games where rewards are designed to lead agents away from a global optimum policy. In this paper, we have developed a deceptive generator that generated three different type of traps including greedy, smoothness and generality trap. The generator was successful in generating levels for a large set of games in the General Video Game Level Generation Framework. Our experimental results show that all tested agents were vulnerable to several kinds of deceptions.

Abstract
Procedural level generation for games is an active field of research with successful applications. However, how to generate content that embodies design intent is still an open research question. Level designers lack abstractions and tools for authoring generated artifacts for affecting emotion. We propose a novel pattern language for generative level design inspired by Appraisal Theory. Its patterns enable designers to add meaning, depth, and cohesiveness to the resulting content, and modify artifacts to make the content more engaging. We illustrate how these patterns can be implemented in a generative grammar for level generation for an adventure game. Formative evaluation of generated level content demonstrates the feasibility of the approach and suggests points for further improvements. Future work could focus on other elements which seem important for affecting emotions, including pacing, perception, and expectation.

Abstract
Procedural content generation via Machine Learning (PCGML) is the umbrella term for approaches that generate content for games via machine learning. One of the benefits of PCGML is that, unlike search or grammar-based PCG, it does not require hand authoring of initial content or rules. Instead, PCGML relies on existing content and black box models, which can be difficult to tune or tweak without expert knowledge. This is especially problematic when a human designer needs to understand how to manipulate their data or models to achieve desired results. We present an approach to Explainable PCGML via Design Patterns in which the design patterns act as a vocabulary and mode of interaction between user and model. We demonstrate that our technique outperforms non-explainable versions of our system in interactions with five expert designers, four of whom lack any machine learning expertise.

Abstract
Beatmania is a rhythm action game where players play the role of a DJ that performs music by pressing specific controller buttons to mix 'Keysounds' (audio samples) at the correct time, unlike other rhythm action games such as Dance Dance Revolution. It has an active amateur Chart (Game stage) creation community, though chart authoring is considered a difficult and time-consuming task. We present a deep neural network based process for automatically generating Beatmania charts for arbitrary pieces of music. Given a raw audio track of a song, we identify notes according to instrument, and use a neural network to classify each note as playable or non-playable. The final chart is produced by mapping playable notes to controls. We achieve an F1-score on the core task of Sample Selection that significantly beats LSTM baselines.

Abstract
Humans' eye movements convey a lot of information about their intentions, often unconsciously. Intelligent agents that cooperate with humans in various domains can benefit from interpreting this information. This paper contains a preliminary look at how eye tracking could be useful for agents that play the cooperative card game Hanabi with human players. We outline several situations in which an AI agent can utilize gaze information, and present an outlook on how we plan to integrate this with reimplementations of contemporary Hanabi agents.

Abstract
Games that involve solving puzzles or mysteries often suffer from low replayability once the solution is known. Procedural content generation has been used to aid with increasing variance in game play, leading to a higher replay value. However, games that feature information that must be obtained by the player present the additional challenge of tracking the player's knowledge and ensuring that it is possible for them to finish the game. In this paper we present a detective game with procedurally generated characters that pursue a range of goals, including one that commits a murder. The player's role is to solve the crime by collecting evidence from the actual game world as well as by interrogating the characters. Even though some of the characters might be inclined to hide evidence, our system guarantees that the player is always able to solve the murder case. To demonstrate the capabilities of our system we present a short example game that we have implemented using our techniques.

Abstract
In many video games, a wide range of non-playable characters make up the worlds players inhabit. Such characters are often voiced by human actors and their accents can have an influence on their perceived moral inclination, level of trustworthiness, social class, level of education and ethnic background. We use deep learning to train a neural network to classify speech accents. Such a machine-learned tool would provide game developers with an ability to analyze accent distribution in their titles as well as possibly help screening voiceover actors applying for a role. To make the approach accessible we used a readily available off the shelf deep network and commodity GPU hardware. Preliminary results are promising with a 71% test accuracy achieved over two accents in a commercial video game.

Abstract
Neural Style Transfer was first unveiled by (Gatys, Ecker, and Bethge 2015), and since then has produced fantastic results working with 2D images. One logical extension of this field would be to move from 2D images into 3D models, and be able to transfer a notion of 3D style from one model to another. Here, we provide steps towards both understanding what style transfer in a 3D setting would look like, as well as demonstrating our own attempts towards one possible implementation.

Abstract
Game development costs are on the rise as players expect massive open-world games populated with richly interactive non-playable characters (NPC). Procedural content generation has the potential to reduce the development costs as well as make the content more player-specific. Recent work on evolving artificial intelligence for NPCs focused on combat between NPCs and the player. In this paper we propose to evolve an ecosystem of NPCs for a broader class of games. To allow players' actions and playstyles to inform the evolution, we propose to model actual players and create AI-controlled proxies to evolve NPCs against. The approach removes the time pressure from the evolution and allows for traditional quality assurance methods while keeping the evolution player-informed.

Abstract
Many games feature collectible items that are manually placed by a designer. In this work, we developed an algorithm, inspired by desire paths, for automatically placing collectible coins in a platformer game. Desire paths are paths naturally formed where people walk, rather than those laid down artificially, and are often the shortest or easiest route between an origin and destination. Our algorithm uses player trajectories to find paths along which to place the coins for each level. We ran an experiment comparing path-based placement to other placement methods. Although we did not find a difference in total time spent playing or likelihood of finishing the game, our results suggest that path-based placement leads to players collecting more coins in less time than with designer or randomly placed coins. Further, we found that players played similarly when coins were either path-based or there were no coins, and similarly when coins were either placed by a designer or randomly.

Abstract
Dread (Barmore et al. 2005) is an award-winning tabletop horror RPG that emphasizes storytelling. One of its innovative mechanics is the use of a questionnaire for character design that helps players develop a personality and history for their characters. Questionnaires vary for different scenarios and characters, but always include 'intrusive' questions such as 'why are you married if you aren't in love?' or 'what's the worst thing you've ever done to a loved one?' In this paper, we discuss work in progress on AutoDread, an implementation of Dread-style questionnaires for video games. The system presents human-authored questions and candidate answers to players to choose between, collecting the implications of those answers in a character model. As implications are accumulated, the system uses a SAT solver to filter questions and answers that are inconsistent with character facts established by previous questions.

Abstract
In recent years, the use of machine learning for procedural content generation (PCGML) has been growing. These PCGML approaches require a training corpus of levels, often annotated or represented in some abstracted way. Due to the manual effort required to annotate or translate a sufficient training corpus, most PCGML techniques have only been explored in a handful of domains. In this paper we take a step towards addressing this core issue of PCGML by presenting an unsupervised method for automatically extracting a representation for a level domain, given only images of the levels. This approach is a move towards making PCGML more broadly applicable by reducing the effort required to create a training corpus. We evaluate our approach by comparing the automatically extracted tile representation against existing PCGML training level corpus representations.

Abstract
We introduce the problem of generating Let's Play-style commentary of gameplay video via machine learning. We propose an analysis of Let's Play commentary and a framework for building such a system. To test this framework we build an initial, naive implementation, which we use to interrogate the assumptions of the framework. We demonstrate promising results towards future Let's Play commentary generation.

Abstract
Expert-crafted puzzle games are notable for their ability to wordlessly teach players deep and interesting consequences of mechanics. Cognitive science theories suggest that this learning happens through an iterative process of hypothesizing, failure, and revision of mental models. In this work, we pose the question of what it would take to generate puzzle game levels and progressions that best support player satisfaction and systematic understanding of mechanics by integrating an understanding of player mental models. We present a preliminary investigation consisting of a new puzzle game Laserverse, whose mechanics are designed to interact with each other in a wide array of combinations, a set of hand-authored levels that combine these mechanics hierarchically, and an algorithm for organizing the levels into progressions to test hypotheses based on a theory of mental model formation. This work sets the stage for empirical validation of puzzle game design principles, which in turn inform a new approach to player modeling based on mental model matching theory.

Abstract
Many computer games feature non-player character (NPC) teammates and companions; however, playing with or against NPCs can be frustrating when they perform unexpectedly. These frustrations can be avoided if the NPC has the ability to explain its actions and motivations. When NPC behavior is controlled by a black box AI system it can be hard to generate the necessary explanations. In this paper, we present a system that generates human-like, natural language explanations-called rationales-of an agent's actions in a game environment regardless of how the decisions are made by a black box AI. We outline a robust data collection and neural network training pipeline that can be used to gather think-aloud data and train a rationale generation model for any similar sequential turn based decision making task. A human-subject study shows that our technique produces believable rationales for an agent playing the game, Frogger. We conclude with insights about how people perceive automatically generated rationales.

Abstract
A number of recent models for empirically-grounded social simulation have emerged recently from games and interactive narrative research, generally exploring models of trust, emotion, and social graph changes that occur in the process of inter-character interactions. However, these models so far failed to provide realistic models of opinion change and predisposition to new knowledge. Equipped with such a notion, these emergent social simulations can express both real and fictionalized depictions of modern phenomena like adverse media influence, the spread of 'fake news,' and the polarization of ideological sects. We present a preliminary computational investigation into modeling opinion change in virtual characters with this goal in mind.

Abstract
Generative methods provide rich, emergent ways to deal with many kinds of data. In this paper, we explore projects that listen to human motion, and respond through emergent generative art in ways that are inspired by dance and puppetry.

Abstract
Recent work has shown machine learning approaches to be effective in training models on existing game data for informing procedural generation of novel content. Specifically, ML techniques have successfully used existing game levels to train models for generating new levels and blending existing ones. Such blending of not just levels, but entire games, has also been proposed as a possible means of generating new games. In this paper, we build on these works by training Long Short Term Memory recurrent neural networks on level data for two separate platformer games-Super Mario Bros. and Kid Icarus-and use these models to create new levels that are blends of levels from both games, thus creating a level generator for a novel, third game.

Abstract
Procedural Level Generation via Machine Learning (PLGML), the study of generating game levels with machine learning, has received a large amount of recent academic attention. For certain measures these approaches have shown success at replicating the quality of existing game levels. However, it is unclear the extent to which they might benefit human designers. In this paper we present a framework for co-creative level design with a PLGML agent. In support of this framework we present results from a user study and results from a comparative study of PLGML approaches.

Abstract
Chairs' Note: In this invited industry case study, Tarn Adams discusses recent extensions to Dwarf Fortress's systems for character deception. A noted opus in the history of videogames, Dwarf Fortress is a roguelike set in procedurally generated fantasy universes. It has been shown at the Museum of Modern Art and has been featured in The New York Times, The New Yorker, Wired, and many other press publications. Currently, Tarn and his brother, Zach Adams, are roughly midway through its famous 30-year development cycle. As Tarn explains in this paper, an upcoming update centered around artifacts-and what characters know about them-has had the fun consequence of necessitating that a certain class of non-player characters cultivate secret identities. This major extension has brought both technical and design challenges, as this paper illustrates. -James Ryan

Abstract
Procedurally generating rich, naturally behaving AI-controlled video game characters is an important open problem. In this paper we focus on a particular aspect of non-playable character (NPC) behavior, long favored by science-fiction writers. Specifically, we study the effects of self-knowledge on NPC behavior. To do so we adopt the well-known framework of agent-centered real-time heuristic search applied to the standard pathfinding task on video-game maps. Such search agents normally use a heuristic function to guide them around a map to the goal state. Heuristic functions are inaccurate underestimates of the remaining distance to goal. What if the agent somehow knew how long it (the agent) would actually take to reach the goal from each state? How would using such self-knowledge in place of a heuristic function affect the agent's behavior? We show that similarly to real life, knowing of one's irrational behavior in a situation can deter the agent from getting into that situation again even if it is, in fact, a part of an optimal solution. We demonstrate the "fear" with a simple example and empirically show that the issue is common in video-game pathfinding. We then analyze the issue theoretically and suggest that "fear" induced by self-knowledge is not a bug but a feature and may potentially be used to develop more naturally behaving NPCs.

Abstract
Video games often populate their in-game world with numerous ambient non-playable characters. Manually crafting interesting behaviors for such characters can be prohibitively expensive. As scripted AI gets re-used across multiple characters, they can appear overly similar, shallow and generally uninteresting for the player to interact with. In this paper we propose to evolve interesting behaviors in a simulated evolutionary environment. Since only some evolution runs may give rise to such behaviors, we propose to train deep neural networks to detect such behaviors. The paper presents work in progress in this direction.

Abstract
Several frameworks exist to describe how procedural content can be understood, or how it can be used in games. In this paper, we present a framework that considers generativity as a pipeline of successive data transformations, with each transformation either generating, transforming, or pruning away information. This framework has been iterated through repeated engagement and education interactions with the game development and generative art communities. In its most recent refinement, it has been physically instantiated into a deck of cards, which can be used to analyze existing examples of generativity or design new generative systems. This Generative Framework of Generativity aims to constructively define the design space of generative pipelines.

Abstract
We present Generominos, a set of design cards to model inter- active generative systems. While many ideation cards exist, Generominos attempts to model the constraints of making a transformative pipeline of data in the constraints of the cards themselves. For this paper, we contribute the design of the Generominos cards and a preliminary evaluation of perceived usefulness in an undergraduate alternative controller class.

Abstract
ANGELINA is an automated game design system which has previously been built as a single software block which designs games from start to finish. In this paper we outline a roadmap for the development of a new version of ANGELINA, designed to iterate on games in different ways to produce a continuous creative process that will improve the quality of its work, but more importantly improve the perception of the software as being an independently creative piece of software. We provide an initial report of the system's structure here as well as results from the first working module of the system.

Abstract
We argue that social simulation can help us understand social justice issues. In particular, modeling certain social dynamics within computational systems can be used to creatively explore and better understand the social and identity dynamics of oppression. Writing theories of oppression in code forces us to explicate everything, and question what we leave out or what we can't account for. As an early step in this direction, we present an in-progress social simulation of group discussion in activist meetings, developed in the already-existing AI system, Ensemble. Through this minimal, highly constrained social arena, we can explore wide-reaching phenomena like privilege, intersectionality, and power dynamics in nonhierarchical groups, but in a way that's grounded in concrete, person-to-person interactions. We propose that this kind of social simulation can aid in the process of unlearning hegemonic ways of being, and imagining liberatory alternatives.

Abstract
In video games, a wide range of characters make up the world players inhabit. These characters, NPCs, have traits, such as their appearance and speech accent, that determine certain things about them, including moral inclination, levels of trustworthiness, social class, levels of education, and ethnic background. But what does an accent say about a character in a video game? We use deep learning to train a neural network to detect speech accents and establish the degree to which machines can be used to recognize these accents. This research aims to help sociolinguists and discourse analysts establish critical study and content analytical findings for instance about stereotypical uses of speech accents, to better analyze who has what accent in video games, and what kind of language ideologies and social value judgments the use of accents in games construct and perpetuate. This paper presents the results of the first deep learning experiments, which were conducted on Standard North American, British Received Pronunciation, and Spanish English. We discuss our methodological considerations and some early deep learning results, which show relatively low levels of accuracy (61%). We discuss possibilities of improving our method, and of enriching our training datasets.

Abstract
We propose the problem of tutorial generation for games, i.e. to generate tutorials which can teach players to play games, as an AI problem. This problem can be approached in several ways, including generating natural language descriptions of game rules, generating instructive game levels, and generating demonstrations of how to play a game using agents that play in a human-like manner. We further argue that the General Video Game AI framework provides a useful testbed for addressing this problem.

Abstract
In this paper we describe a level design editor designed as an interface to allow different AI agents to creatively collaborate on level design problems with human designers. We intend to investigate the comparative impacts of different AI techniques on user experience in this context.

Abstract
Marahel is a language and framework for constructive generation of 2D tile-based game levels. It is developed with the dual aim of making it easier to build level generators for game developers, and to help solving the general level generation problem by creating a generator space that can be searched using evolution. We describe the different sections of the level generators, and show examples of generated maps from 5 dif- ferent generators. We analyze their expressive range on three dimensions: percentage of empty space, number of isolated elements, and cell-wise entropy of empty space. The results show that generators that have starkly different output from each other can easily be defined in Marahel.

Abstract
General videogame playing has come a long way in a short period of time, but remains at the level of solving relatively short games made up of distinct and isolated episodes. Even simple console role-playing games (RPGs) are far beyond the reach of current techniques, requiring the synthesis of cultural knowledge with compositional reasoning over several interconnected sub-games. We explore how the challenges of playing these games could spark new advances in compositional analysis of games and common-sense reasoning. General RPG playing can leverage advances in episodic general game playing and in areas like text understanding, image classification, and automated game design learning. It has direct applications in design support and AI-based game design, and the techniques used to enable it could generalize to other families of games such as adventure, open-world, and simulation games. In this paper, we describe the motivation behind general RPG playing in a sub-domain of Nintendo Entertainment System (NES) RPGs, some promising approaches to some of its fundamental issues, and immediate next steps; we conclude by describing a few concrete benchmark problems on the path towards automated play of these complex games.

Abstract
A procedure that uses phonetic transcriptions of words to produce a continuous vector-space model of phonetic sound similarity is presented. The vector dimensions of words in the model are calculated using interleaved phonetic feature bigrams, a novel method that captures similarities in sound that are difficult to model with orthographic or phonemic information alone. Measurements of similarity between items in the resulting vector space are shown to perform well on established tests for predicting phonetic similarity. Additionally, a number of applications of vector arithmetic and nearest-neighbor search are presented, demonstrating potential uses of the vector space in experimental poetry and procedural content generation.

Abstract
We present a system for quickly prototyping AI code for modeling social interaction among the simulated residents of a skyscraper. The system is built on top of a commercially released indie game but replaces its character AI with a general-purpose logic programming language. These two features together simplify the more tedious parts of creating prototype AI code from scratch, enabling more effort to be focused on the meatier parts of research.

Abstract
In this invited industry case study, Tanya X.Short introduces eight mechanisms that designers can utilize to better harness procedural character personalities in games. Tanya is co-founder and captain of Kitfox Games, a Montreal-based independent game studio, and a veteran developer and designer known for her expertise in procedural generation and systems-driven game design. With Tarn Adams, she edited the 2017 volume Procedural Generation in Game Design. As Tanya explains in this paper, there is an emerging pattern in game design that utilizes character personality as a central gameplay system. Using a number of examples spanning her experiences as both developer and player, in this paper Tanya distills her collected design knowledgeinto an actionable recipe for building better character personality systems.

Abstract
Proofdoku is an AI-based game design that extends Sudoku. In addition to playing by the rules of the traditional logic puzzle, players must explain their reasoning. An AI system checks this reasoning and provides hints that guide the player to discover new reasoning patterns for themselves. Co-developing the game design and the AI system, implemented using the technology of Answer Set Programming (ASP), guided us to include features that depend on high-complexity combinatorial search and optimization. We developed Proofdoku to better understand the implications of designing and deploying game systems that depend on ASP for live interaction. This paper offers design tradeoffs and makes suggestions for future deployments of ASP-backed game designs.

Abstract
Procedural content generation via machine learning (PCGML) has been growing in recent years. However, many PCGML approaches are only explored in the context of linear platforming games, and focused on modeling structural level information. Previously, we developed a multi-layer level representation, where each layer is designed to capture specific level information. In this paper, we apply our multi-layer approach to Lode Runner, a game with non-linear paths and complex actions. We test our approach by generating levels for Lode Runner with a constrained multi-dimensional Markov chain (MdMC) approach that ensures playability and a standard MdMC sampling approach. We compare the levels sampled when using multi-layer representation against those sampled using the single-layer representation; we compare using both the constrained sampling algorithm and the standard sampling algorithm.

Abstract
A challenge of game design is in providing affordances to players so that they can learn and improve their skills. Advances in Procedural Content Generation (PCG) suggest this type of game content is a candidate for automatic creation. Some work in PCG has been successful in customizing game difficulty to achieve desired player experience; however, this often involves bringing the difficulty of the game to a level appropriate for the player's current skills. Players desiring to improve their performance in a particular game may be willing to tolerate relatively higher levels of frustration and anx- iety than are targeted in experience-based approaches. As an initial step in this line of inquiry, we introduce ProcDefense, an action game with a modular difficulty control interface, as a platform for future inquiry into the effectiveness of differing PCG techniques for performance-training, dynamic difficulty adjustment.

Abstract
Dynamic Epistemic Logics are a set of multimodal logics that deal with knowledge and change. We argue that the theory for this formalism is mature enough to start a practical implementation, while at the same time having a sizable amount of theoretical expansions. We also claim that these two factors make it an attractive formalism for new features in games featuring an agent's internal model, like planning for gathering information, acting without revealing one's goals or feeding possibly false information to a set of agents to influence their beliefs, augmenting already existing mechanisms.

Abstract
Games have used different mechanisms along its history to provide agent behavior: FSMs, utility systems, behavior trees and planning methods. In this paper, we present an architecture that aims at incorporating all these approaches into trees of event handling nodes with behaviours as leaves, using rules for combining actions akin to utility systems. This formalism aims to develop hybrid systems in an easier way.

Abstract
The use of artificial intelligence and procedural content generation algorithms in mixed reality games is an unexplored space. We posit that these algorithms can enhance the gameplay experience in mixed reality games. We present two prototype games that use procedural content generation to design levels that make use of the affordances in the player's physical environment. The levels produced can be tailored to a user, customizing gameplay difficulty and affecting how the player moves around the real-world environment.

Abstract
This paper describes an avenue for artificial and computational intelligence techniques applied within games research to be deployed for purposes of physical therapy. We provide an overview of prototypical research focussed on the application of motion sensor input devices and virtual reality equipment for rehabilitation of motor impairment: an issue typical of patients of traumatic brain injuries. We highlight how advances in procedural content generation and player modelling can stimulate development in this area by improving quality of rehabilitation programmes and measuring patient performance.

Abstract
Automatic analysis of game levels can provide as- sistance to game designers and procedural content generation. We introduce a static-dynamic scale to categorize level analysis strategies, which captures the extent that the analysis depends on player simulation. Due to its ability to automatically learn intermediate representations for the task, a convolutional neural network (CNN) provides a general tool for both types of analysis. In this paper, we explore the use of CNN to analyze 1,437 Infinite Mario levels. We further propose a deep reinforcement learning technique for dynamic analysis, which allows the simulated player to pay a penalty to reduce error in its control. We empirically demonstrate the effectiveness of our techniques and complementarity of dynamic and static analysis.

Abstract
Players in Dear Leader's Happy Story Time are placed in the role of contestants in a reality TV show where they are forced to audition for roles in the upcoming film of the host, a deranged billionaire who has inexplicably been elected president. The stories are produced by a story generator that combines stock plots and characters to produce kitsch story outlines. The players then collaborate to improvise a camp performance of the outline. The game design provides a context for experimenting with automatic story generation within a narrative game, as well as an opportunity for experimenting with knowledge representation schemes for expressing the tropes of popular narrative. The story generator uses a higher-order logic for describing tropes, and an HTN planning algorithm based on Nau et al.'s SHOP.

Abstract
We describe an application of neural networks to predict the placements of resources in StarCraft II maps. Networks are trained on existing maps taken from databases of maps actively used in online competitions and tested on unseen maps with resources (minerals and vespene gas) removed. This method is potentially useful for AI-assisted game design tools, allowing the suggestion of resource and base placements consonant with implicit StarCraft II design principles for fully or partially sketched heightmaps. By varying the thresholds for the placement of resources, more or fewer resources can be created consistently with the pattern of a single map. We further propose that these networks can be used to help understand the design principles of StarCraft II maps, and by extension other, similar types of game content.

Abstract
While generative approaches to game design offer great promise, systems can only reliably generate what they can 'understand,' often limited to what can be handencoded by system authors. Proceduralist readings, a way of deriving meaning for games based on their underlying processes and interactions in conjunction with aesthetic and cultural cues, offer a novel, systematic approach to game understanding. We formalize proceduralist argumentation as a logic program that performs static reasoning over game specifications to derive higher-level meanings (e.g., deriving dynamics from mechanics), opening the door to broader and more culturally-situated game generation.

Abstract
Platformer level generation has often used a beat metaphor to relate to how players interact with level geometry. However, this conceptualization of beats is different from the musical concept of `beat', limiting the utility of theories and tools developed in music analysis for platformer levels. A gameplay gestalt, a pattern of interaction that the player enacts or performs in order to make progress in a game, may fit the beat metaphor. By taking a very similar lens and viewing players playing platformer levels as enacting a series of gameplay gestalts through time, gestalt music analysis (GMA) does fit into the platformer domain. This paper details work on transforming a GMA model to work with the Platformer Experience Dataset (PED), and some promising first results of the transformed model.

Abstract
Interactive narratives (IN) are stories that branch and change based on the actions of a participant. A class of automated systems generate INs where all story branches conform to a set of constraints predefined by an author. Participants in these systems may create invalid branches by navigating the story world outside the bounds of an author's constraints. We approach this problem from an adversarial game perspective, where the IN system's goal is to prevent the player from creating invalid branches. From this perspective, one way an IN system can take action is to transition the game world between alternate states that are consistent with the player's observations during gameplay. In this paper we present a method of modelling and updating sets of world states consistent with player knowledge as a single superposed data structure. We discuss how this data structure can be used in an IN framework to maximize the probability that author constraints are maintained during gameplay.

Abstract
We present Juke Joint, a small work of interactive storytelling that demonstrates an extension to the Talk of the Town framework by which characters form thoughts, expressed in natural language, that are elicited by environmental stimuli. Juke Joint takes place in a procedurally generated American small town, in a bar with a haunted jukebox and two patrons facing personal dilemmas; the player is a ghost whose only action is to select which song from the jukebox will play. As the lyrics of the song emanate from the machine, thoughts are elicited in the minds of the patrons, constituting streams of consciousness that may eventually lead them to resolutions of their respective dilemmas. In this paper, we outline the game and also the AI architecture that makes it possible; the latter combines a light simulation of stimulus processing with a novel approach to natural language generation.

Abstract
Computationally assisted performance is a burgeoning area for AI applications, and an important stepping stone toward the dream of generative and personalized narrative experiences. As more pieces of computationally assisted performance are developed, it will become ever more important to develop a vocabulary with which to describe them. Inspired by previous work in creating taxonomies for other related domains, this paper outlines a taxonomy for performance-based experiences, drawn from digital games, traditional theatre, and the hybrid of the two. Having such a taxonomy not only creates a common language with which to discuss such experiences, but reveals unexplored design space in the field, and the particular applications of artificial intelligence necessary to realize them.

Abstract
Boss fights are a memorable and climatic point of many games. In this work we present a programming model for defining boss experiences in 2D action games. The domain we focus on is characterized by real-time movement through a continuous space in which the player and opposing boss damage each other through physical collisions, and player and boss behavior is governed primarily by finite state machines. Our programming model consists of primitive systems such as kinematic physics to store object state and detect collisions paired with finite state machines to define behavior. We describe our model and demonstrate its expressiveness with examples of three classic boss fights from games in The Legend of Zelda, Castlevania, and Sonic the Hedgehog. Our future goals include procedural generation of boss encounters, and we report on the research challenges involved in achieving this goal.

Abstract
In this paper I present ClueGen, a murder mystery game that generates its own narrative. I discuss the genre's suitability for combining gameplay and story using procedural generation techniques, before explaining the implementation and rationale behind ClueGen, utilizing the genre's common structure as an opportunity to define the scope of the generation requirements. In particular I detail a novel system for telegraphing deceptive characters in procedurally generated games with text-based dialogue, by using audible cues.

Abstract
Analysing strategic decision-making in eSports is an increasingly important problem -- for players, for teams, for commentators, for viewers and for broadcasters. Such analysis is extremely difficult, however, because of the comparatively small quantities of data, the ever-shifting state of competitive play, and the huge complexity of the game. In this paper we describe a system for predicting drafting decisions in DOTA 2, and evaluate both how the system performs compared to human experts, as well as the new kinds of analysis made possible by automation.

Abstract
A touted use of Procedural Content Generation is generating content tailored to specific players. Previous work has relied on human identification of player profile features which are then mapped to level generator features. We present a machine-learned technique to train generators on Super Mario Bros. videos, generating levels based on latent play styles learned from the video. We evaluate the generators in comparison to the original levels and a machine-learned generator trained using simulated players.

Abstract
Creating reasonable AI for sidekicks in games has proven to be a difficult challenge synthetizing player modelling and cooperative planning, both being problems hard by themselves. In this paper, we experiment with designing around these problems: we propose a cooperative puzzle-platformer game that was designed to look similarly to the mainstream of the genre, but to allow for an easy implementation of a quality sidekick AI, letting us test player reactions to the AI. The game was designed so that it is easy for the AI to find optimal solutions while the problem is relatively hard for a human player. We gathered survey responses from players who played the game online (N=28). While the AI sidekick was reported as likeable and helpful, players still reported greater enjoyment of the game when they were allowed to control the sidekick themselves. These findings indicate that the AI itself is not the only obstacle to truly enjoyable gameplay with an AI sidekick.

Abstract
In this paper we present a technique for procedurally generating sections of 3D level geometry using computational evolution and guided by the visibility of certain game objects or areas during play. We show that certain level design goals can be achieved in the resulting levels, such as encouraging or dissuading player sightings of certain objects or locations. We also give details of a simple study of players on the generated levels, and discuss how this might be expanded to incorporate more complex problems related to level design.

Abstract
We sketch the process of creating a novel video game by blending two video games specified in the Video Game Description Language (VGDL), following the COINVENT computational model of conceptual blending. We highlight the choices that need to be made in this process, and discuss the prospects for a computational game designer based on blending.

Abstract
We present Fiascomatic, a mixed initiative system for generating consistent scenarios for the indie storytelling RPG Fiasco. Players can repeatedly generate scenarios, locking down aspects of a scenario they like and regenerating aspects they don't, until they arrive at a scenario they find entertaining. It is not a story generation system; it generates scenarios from which players then generate stories. Nor is it intended to generate optimal scenarios; it generates random scenarios which the players can then curate according to their taste. Fiascomatic presents an interesting intermediate point between non-automated table-top RPGs and fully automated systems such as story generators or autonomous characters. It is a tool that can be used by Fiasco players to speed the generation of game setups while preserving creative input on the part of the players, and by Fiasco playset authors to make automated playsets.

Abstract
Research has indicated that role models have the potential to boost academic performance. In this paper, we describe an experiment exploring role models as game avatars in an educational game. Of particular interest are the effects of these avatars on players' performance and engagement. Participants were randomly assigned to a condition: a) user selected role model avatar, or b) user selected shape avatar. Results suggest that role models are heavily preferred. African American participants had higher game affect in the role model condition. South Asian participants had higher self-reported engagement in the role model condition. Participants that completed <= 1 levels had higher performance in the role model condition. General trends suggest that the role model's gender and racial closeness with the player, could play a role in player performance and self-reported engagement as consistent with the social science literature.

Abstract
We present a system for autonomously generating puzzles in the form of a 2D, tile-based world. Puzzle design is entirely dependent on tile characteristics, which are implemented as abstract classes that can be modified by the system. Thus, the system controls not only the base-level puzzle design but also (to some extent) the meta-level component design. The result is a rich space of possible puzzles that the system explores with a combination of evolutionary computation and Q-learning. The system autonomously produces a variety of puzzles of varying difficulty to create a game called Loki's Castle. The system is almost completely autonomous, requiring only a minimal description of what a puzzle should include, and the abstraction allows extensibility so that future versions can invent entirely new classes of tiles. Several puzzle examples are presented to demonstrate the system's capability.

Abstract
This paper describes the structure of a web service able to generate simple game levels via constrained evolutionary optimization. The provided web service allows users to generate playable game levels without needing to understand the underlying process and without having to allocate computational resources for doing so; combined with the highly expressive and customizable generator, a broad range of levels for different genres and purposes can meet many user needs.

Abstract
We encounter the results of category formation every day, from demographic categories like race and gender, to role-playing-game classes like "fighter" or "mage". Category membership is often not simply based on the possession of discrete properties but instead constructed from and reflect the highly nuanced relationships (gradience) between members and best-example individuals called "prototypes". In this paper, we present The Marginal, an artificial intelligence (AI)-driven game that (1) computationally models the cognitive categories that players develop when customizing videogame avatars and (2) generates challenges for players to use their perception of visual, textual, and numerical data to progress in a game created using these models. We use archetypal analysis, an AI clustering approach for identifying boundary points in data, to generate tasks in The Marginal for its gameplay. It shows how AI can be combined with games to model and evaluate cognitive categorization phenomena.

Abstract
Knowledge and its attendant phenomena are central to human storytelling and to the human experience more generally, but we find very few games that revolve around these concerns. This works to preclude a whole class of narrative experiences in games, and it also damages character believability. In this paper, we present an AI framework that supports gameplay with non-player characters who observe and form knowledge about the world, propagate knowledge to other characters, misremember and forget knowledge, and lie. We outline this framework through the lens of a gameplay experience that is intended to showcase it, called Talk of the Town, which we are currently developing. From a review of earlier projects, we find that our system has a novel combination of features found only independently across other systems, and that it is among the first to support character memory fallibility.

Abstract
Procedural Content Generation (PCG) using machine learning is a fast growing area of research. Action Role Playing Game (ARPG) levels represent an interesting challenge for PCG due to their multi-tiered structure and nonlinearity. Previous work has used Bayes Nets (BN) to learn properties of the topological structure of levels from The Legend of Zelda. In this paper we describe a method for sampling these learned distributions to generate valid, playable level topologies. We carry this deeper and learn a sampleable representation of the individual rooms using Principal Component Analysis. We combine the two techniques and present a multi-scale machine learned technique for procedurally generating ARPG levels from a corpus of levels from The Legend of Zelda.

Abstract
Placing decorative content in any virtual environment is not a trivial task, as the location of even non-critical content influences player movement choices through the level, as well as the general game aesthetic. In this work we use path visibility properties to define and investigate three methods for content placement in a game level, aimed at generating shared, unique, and never seen content. We compare and evaluate the three approaches using a human study where each player was given a level layout based on our three methods, showing our directed process has a non-trivial and controllable impact on player movement.

Abstract
Markov chains are an enticing option for machine learned generation of platformer levels, but offer poor control for designers and are likely to produce unplayable levels. In this paper we present a method for guiding Markov chain generation using Monte Carlo Tree Search that we call Markov Chain Monte Carlo Tree Search (MCMCTS). We demonstrate an example use for this technique by creating levels trained on a corpus of levels from Super Mario Bros. We then present a player modeling study that was run with the hopes of using the data to better inform the generation of levels in future work.

Abstract
Procedural content generation is often seen simply as a means to generate stuff, elaborating on or rearranging abstract data types that describe levels or modular pieces of gameplay. Generating content which is situated in an understanding of the real-world is a much harder task; it requires access to large amounts of knowledge, and a good technique for parsing and using that knowledge. In this paper we describe A Rogue Dream, a game prototype which can generate new visual content and change its design based on an input word from the player at the start of the game. We describe the game and the tools it makes use of to do this, and use the game to discuss ways in which such techniques might enable unique kinds of gameplay or new directions for intelligent design tools.

Abstract
Due to the virtual nature of Virtual Environments (VE), there is an affordance to change game and avatar features asymmetrically, where the features that are rendered for one participant can be different from those rendered for other participants while still sharing the same virtual world. We are calling these kinds of VEs Asymmetric Virtual Environments (AVE). In this paper we present work on a platform for studying online interactions using an AVE. We do so through a preliminary study that explores the effects of avatar colors on individual and group performance when solving a scavenger hunt in a 3D virtual environment followed by a location identification task in a 2D virtual environment. While the results of this preliminary study are not conclusive, our data seem to indicate that players' avatar colors can indeed influence online performance.

Abstract
Large online repositories of player-generated game content are a popular component of modern interactive digital entertainment. These add-ons are typically referred to as "mods", and the communities of co-creators that coalesce around them "modding communities". We propose a method to augment these communities with computational agents that encourage creative submissions, which we define as those that are both unexpected and valuable. We propose to combine player experience modelling, computational creativity evaluation and mixed-initiative co-creation to drive modding communities towards more fun and diverse gaming experiences.

Abstract
Procedural game generation is the automatic creation of all aspects of a playable computer game. Procedural game generation systems require specialized knowledge, virtual worlds, and art assets. In this paper, we show how 3D graphical scenes for interactive fictions can be automatically generated with only knowledge that is readily available in existing knowledge bases or can be acquired via crowdsourcing. The key to 3D scene generation is commonly accepted spatial relationships between different types of objects in different types of scenes. We use a crowdsourcing game to automatically and rapidly acquire spatial relations. The spatial relations are used by an intelligent scene generation system that selects and configures 3D assets within a virtual geometric space.

Abstract
Reasoning using expressive symbolic representations is a central theme of AI research, yet there are surprisingly few deployed games, even within the AIIDE research community, that use this sort of 'classical' AI. This is partly due to practical and methodological issues, but also due to fundamental mismatches between current game genres and classical AI systems. I will argue that if we want to build games that leverage high-end classical AI techniques like commonsense reasoning and natural language processing, we will also have to develop new game genres and mechanics that better exploit those capabilities. I will also present a design sketch of a game that explores potential game mechanics for classical AI.

Abstract
The invention of fictional ideas is often a central process in the creative production of artefacts such as poems, music, paintings and games. Currently, fictional ideation is being studied by the Computational Creativity community within the WHIM European project. The aim of WHIM is to develop the What-If Machine, a software system capable of inventing, evaluating and presenting fictional ideas with cultural value. In this paper we explore the potential applications of the What-If Machine in the context of games. Specifically, we propose ways in which the What-If Machine can be used as an assistant for the design of games, by providing ideas about characters, the environment, etc., as well as a creative system during gameplay, through interesting interactions with the player.

Abstract
Mediation is a plan-based interactive narrative generation algorithm that creates a cascading policy of plans. This policy represents a branching story and can be used by an execution manager in a game to control the series of events that unfold. With a few modifications we show that mediation's search space can embed all possible traversals through a game world. This shift allows gameplay to be modeled as an on-line mediation search where the game's interface is a representation of the underlying graph traversal. In this paper we outline these modifications and present a text-based implementation.

Abstract
In order to advance the field of neuroscience, we must continue motivating youth to pursue science education. In this report we tested the idea of combining neurogaming with education. We developed a pair of electroencephalography (EEG)-controlled neurogames using inexpensive and/or free tools to teach students about the fundamentals of neuroscience and brain machine interfaces (BMI) through a fun, interactive activity. We report on the particular concepts they allowed us to introduce, the techniques and methods we used, and the effect of the activities on stimulating students' interest in neuroscience, and discuss how to optimize the learning experience. We conclude that educational neurogames could be a key tool for furthering and motivating neuroscience education.

Abstract
The future of procedural content generation (PCG) lies beyond the dominant motivations of 'replayability' and creating large environments for players to explore. This paper explores both the past and potential future for PCG, identifying five major lenses through which we can view PCG and its role in a game: data vs. process intensiveness, the interactive extent of the content, who has control over the generator, how many players interact with it, and the aesthetic purpose for PCG being used in the game. Using these lenses, the paper proposes several new research directions for PCG that require both deep technical research and innovative game design.

Abstract
Platformer games let players solve real-time, physics-based puzzles by jumping and moving around to reach different goals. Designing levels for this context is a non-trivial task; the placement of well-timed jumps, moving platforms, interesting traps, etc., has a complex relationship to in-game challenge and the existence of possible solutions. In this work, we describe three different search algorithms (A*, MCTS and RRT) that could be used to simulate player be- haviour in the platformer domain. We evaluate and compare the three approaches applied to three non-trivial levels, showing a possible iterative workflow of use to designers, and research progress in designing search algorithms for platformer games.

Abstract
The General Mediation Engine is an architecture for producing game experiences with world mechanics drawn from a planning domain and problem. The system models possible player actions, state updates, and system responses in a graph representation derived from planner input. Paired with an interface that displays state and action information to the player, GME structures gameplay as an on-line expansion of its graph representation.

Abstract
This paper introduces a software framework, currently in prototype stage, designed to permit automatic integration of pathfinding navigation in isomertric camera-driven games. This framework is built with an emphasis towards touchscreen-driven gameplay on tablets and smartphones. This enables pathfinding to be adopted not by characters in-game, but also permits a range of traditional and experimental control schemes to utilise this feature. In addition, the framework provides for a range of metrics to be established based on user activity; enabling quality assurance to be conducted based on raw-user input and pre-processed data.