Stratmapper
Motivation: Game data is exceedingly complicated, made up of large quantities of data points each representing actions taken by a human. Those actions, in turn, are informed by human decisions, which are, in turn, informed by ever-changing contextual elements. How, then, do we empower game data analysts or even players themselves to extract accurate, insightful, and actionable insights from gameplay data?
Description: Stratmapper is an interactive spatio-temporal data visualization system designed to make complex gameplay behavioral data interpretable in a context and process-sensitive manner. The system features a map that populates with data points to represent players’ movements and actions in terms of where they occurred and a timeline that represents them in terms of when they occurred. The user can highlight segments of time on the timeline to control what is shown on the map and can move the highlight across the timeline to animate the data points on the map. The system also allows users to apply labels to segments of data that can be exported for analysis via other tools or methods.
Results: The tool has been well-received by esports players and has produced publications at ACM CHIPlay and ACM CHI
Team: Erica Kleinman (UX Research and Design Lead/UI Design Lead) Andy Bryant (Lead Front End Developer) Sabbir Ahmad (Lead Data Scientist and Back End Developer) Zhaoqing Teng (UX Research Assistant and Back End Development) Nikhita Preetham (UX Research Assistant) Truong-Huy D. Nguyen (Co-PI) Magy Seif El-Nasr (PI)
Process Visualizations to Support Self-Reflection in Complex Gameplay
Motivation: The culmination of my dissertation work, through interviews and systematic review I found that existing computational tools for supporting players’ learning and reflection processes lack causal information, making it difficult for players to understand how their actions resulted in the outcomes they experienced. The problem to address, then, is finding a way to present players with enough information about their process so that they may better identify mistakes and adapt their strategies.
Description: The images here depict different prototypes for process-visualizations meant to be viewed by players after gameplay. By leveraging insights from the domain of process mining, these visualizations present the player with all of the actions they took, in the order they took them. The topmost image depicts this concept in the context of the esport League of Legends while the bottom two depict it in the context of an educational game.
Results: The results of two different experimental studies (one between subjects and one within) have demonstrated that these visualizations have significant impacts on performance, quality of reflection, and willingness to adapt one’s strategy. Some of this work has been accepted for publication at ACM CHI and ICEC and the rest is currently under review at ACM CHIPlay
Team: Erica Kleinman (Abstraction and Visualization Design Lead/User Research Lead) Jennifer Villareale (Research Assistant) Murtuza Shergadwala (Research Assistant) Zhaoqing Teng (Technical Support) Andy Bryant (Developer) Jichen Zhu (Co-PI) Magy Seif El-Nasr (PI)
Interactive Visualization and Exploration System (IVES)
Motivation: Identifying human behavior is the first step to preventing unwanted behaviors, an objective inherent to domains such as cyber social attacks, where agents will attempt to influence public opinion. Behavioral models are often trained to identify behaviors in this way using logged activity data, but how do we ensure that the model understands the behavior correctly? Especially when the behavior may be vast, complicated, and heavily influenced by contextual factors.
Description: Building on Stratmapper and the HAP behavior language, IVES is a system meant to visualize a behavioral model’s understanding of a parsed behavior to a domain expert. Using HAP’s tiered goal and behavior approach, the model’s understanding of the behavior is displayed as a tree of interconnected nodes that represent either goals, behaviors, or actions alongside how confident the model is that it has identified each goal or behavior accurately. Through a series of UI elements, the domain expert can rename nodes, add or remove nodes, and adjust the connections between nodes in order to correct the model’s understanding. The user can also filter nodes based on type, the model’s confidence, or the level of abstraction. Further features are in development based on the results of an early user study including a view that locks the camera in a top-down view and one that locks it in a side view.
Results: Proof-of-concept visualizations have been created for a fictional DotA2 scenario and a fictional cyber-social attack scenario. The DotA2 scenario is featured in a publication recently accepted at FDG 2023
Team: Erica Kleinman (Abstraction and UI Design Lead/User Research Lead) Andy Bryant (Lead Software Developer) Zhaoqing Teng (Software Developer) Nikhita Preetham (Abstraction Designer and Software Developer) Stefany Arevalo Escobar (Software Developer and UX Research and UI Design Assistant) Derusha Baskaran (UX Research and UI Design Assistant) Spencer Lynn (Project Manager and Co-PI) Bryan Loyall (Co-PI) Magy Seif El-Nasr (PI)
A Human-in-the-Loop Method For Analyzing Behavioral Sequences
Motivation: There are many domains, ranging from games to education to health, that would benefit from a robust and interpretable way to analyze and understand behavioral sequences. Being able to analyze such data would allow, for example, educators, to make more informed decisions about how to adjust learning content to meet a student’s needs or to identify those students who need help.
Description: This mixed-method approach to behavioral data analytics combines algorithms, visualization, and qualitative analysis to allow a human to work iteratively with an algorithm to identify interpretable behavioral clusters, outliers, and common decision-making processes. While the method requires the user to either have or collaborate with someone possessing programming knowledge, we demonstrate through a case study with an educational game that this method detects behavioral clusters that more closely match those identified by domain experts than existing methods.
Results: The full details of the method are published in the Journal of Learning Analytics (See Publications)
Team: Erica Kleinman (Data Abstraction and Method Design Lead) Murtuza Shergadwala (Data Abstraction Designer) Jennifer Villareale (Data Abstraction Designer and Parallel Designer) Zhaoqing Teng (Glyph Back End Developer) Jichen Zhu (Co-PI) Magy Seif El-Nasr (PI)
Rough Draft
Motivation: Meta-gaming mechanics, specifically rewinding and redoing choices, have become more common in commercially available games with several critically acclaimed games, including Life is Strange and Undertale, not only encouraging but requiring the player to return to the past and revisit events (and sometimes remake choices) to progress the story. The question, then, is how engagement and immersion in the narrative are impacted by this constant rewinding and redoing.
Description: Rough Draft is a relatively simple interactive story game about a writer, Denise, trying to work her way through writer’s block and finish a children’s fantasy story. In the story she is writing, Reina, a princess, must return Lyre, a baby dragon, to his mother to free her brother from a curse. However, due to her writer’s block, Denise keeps writing the characters into corners and pushing the plot to dead ends. Thus, the player must regularly rewind the game back to a previous choice and remake it in order to progress. Moving forward in the story will typically uncover new information that, when the player rewinds, will unlock new choices. For the purposes of answering the questions described above, three versions of the game were created. One served as a control and featured no rewinding. A second featured a localized rewind, in which the player could only return to their most recent choice. The third featured a global rewind, which allowed the player to return to any previously made choice. This game was used as a part of a user study that examined the impact of each design on player engagement.
Results: Results of the study found that players enjoyed being able to remake choices, as it was something they would often do anyway, and that allowing them to do so did not have any significant negative impact on their experience. The project has directly resulted in 2 peer-reviewed publications in ICIDS 2016 and the Journal of Entertainment Computing and inspired the work of two more in ICIDS 2019 and FDG 2019. Rough Draft (all versions) are available online here. Read more about the game in my Master’s Thesis here.
Team: Erica Kleinman (Lead Designer, Programmer, and Researcher) Valerie Fox (Faculty Advisor) Stefan Rank (Faculty Advisor) Jichen Zhu (Factuly Advisor)
Matsya
Motivation: Matsya was created as a group project for a two-term game design course at Drexel. The goal for the project was to create a relaxing, zen-game experience, while also creating a game that incorporated elements of Hindu myth and art, which are under-represented in mainstream western games. Through the analysis and evaluation of a number of culturally inspired games including Okami and Never Alone, and inspired by an Indian flood myth, the game aims to discretely convey elements of Hindu culture and philosophy.
Description: Matsya is intended to be played on an iPad and to invoke a sense of balance and flow. The gameplay features rings of water that the player can spin by dragging their fingers on the iPad screen. The rings slowly move towards the center and disappear as new ones spawn on the outer edges. In the center of the rings is a boat that will randomly shoot out fishing spears in different directions. There are fish swimming in the rings and the player can spin the ring to direct the fish into, or away from, the path of the spear. The rings and fish were programmed to move with a slight amount of lag, to emulate the phenomenon of dragging one’s finger through actual water. The game features ten levels with explicit rules, created to help appeal to those who prefer a more structured gameplay experience. Advancing through each level unlocks a new piece of the myth’s story in the main menu, allowing the player to step through the entire narrative as they complete the levels. The game also features a “zen” mode in which the player can play infinitely. All of the art is based on traditional Hindu artwork.
Results: Matsya was well received by players and won runner-up at the 2015 CHIPlay student game design competition. More can be read about the game here.
Team: Joe Baranoski (Producer, Lead Game Designer, Programmer) Erica Kleinman (Art Director, Game Designer, Writer/Narrative Designer, User Research Lead) Zheng Wang (3D Artist and Animator, Programmer) Martina Tucker (2D Artist, UI Designer) Jonathan Ahnert (Early Production Lead, Concept Designer) Jonas Schell (Programmer, Game Designer) Rohan Doshi (Lead Programmer, Game Designer, Cultural Consultant) Stefan Rank (Faculty Advisor) Jichen Zhu (Faculty Advisor)