CS 391 - Selected Topics: Game Artificial Intelligence Course Information

Course Overview

This research-based course will teach and apply a variety of game artificial intelligence (AI), machine learning, and data science concepts in the context of creating competitive AI players for a modern board game.

While we will learn to apply techniques to smaller simpler games, play of a modern board game will be the focus challenge problem for teams to periodically test their AI approaches in competition. After each periodic competition, students will present their innovations, encouraging all to bring improved players to the next competition. Hence, we will mix competition and open collaboration to encourage collective research progress.

Learning Objectives

• Understand the foundational modern algorithms for game AI from minimax with alpha-beta pruning and expectiminimax, to Monte Carlo tree search.
• Implement reinforcement learning for AI agents to approximately evaluate the utility of a given game state.
• Apply a variety of machine learning and statistical models and techniques from data science to yield insight to the prediction of state utility.
• Use game play statistics to observe weaknesses in game design that may be iteratively improved, e.g. over/under-powered actions in game play, in order to achieve AI-assisted game design.

Text

There is no text for this selected topics course.  Rather, students will be reading from a selection of recommended and found resources online. As students will be taking this course from a variety of background preparations, students will be encouraged to read so as to fill their greatest gaps in Game AI, Machine Learning, and Data Science understanding, as is most relevant to our research efforts.

Instructor

Grading

70% Work Logs
20% Contest Performance
10% Class Attendance / Participation

Accommodations

Students needing special accommodations should request a meeting to get a Verification of Accommodation Letter (VAL) at the beginning of the semester, and should schedule a meeting time with me during the first week of classes to discuss the VAL.

Attendance

Class attendance and participation is required.  If you attend all classes and are willing to participate, you'll get 100% for this part of your grade.  Even if you know enough to give a particular lecture, please consider the value of helping your peers during in-class exercises.

Woody Allen is quoted as saying "80% of success is just showing up."   While our class attendance/participation grade is not 80% of the final grade, it is critical that late arrivals and unexcused absences are not excessive.  Missing more than half of class unexcused is considered being absent.  An unexcused late arrival is counted as a half absence.  If the total number of absences counted this way exceeds 20% of class meetings, i.e. 6 absences or more, the student will have failed the course.

Work Expectations

Most of your course grade will be based on honestly logged work beyond class. You are expected to work an average of 9 hours per week beyond class time.  Gettysburg College policy, in accordance with federal and state standards, equates 1 credit unit with an average of 12 hours of work per week with 75-minute classes counting as 1.5 hours of work.  During the remaining 9 hours beyond class, a student is expected to learn from assigned readings, complete exercises related to such readings, attend required colloquia, and complete assignments.

Think of your college studies as a more-than-full-time job, and engage in it with passion.  After all, you get out of it what you put into it, and it is my hope that you'll gain much from your investment in this course.  If you'd like to learn more about how to better track tasks and manage time as a student, consider watching my short tutorial on getting things done.

Honor Code

Honesty, Integrity, Honor.  These are more important than anything we will teach in this class.  Students can and are encouraged to help each other understand course concepts, but all graded work must be done independently unless otherwise specified (e.g. group work).  Submitted work should be created by those submitting it.  Submission of plagiarized code or design work is a violation of the Honor Code, which I strictly enforce.  For detailed information about the Honor Code, see http://www.gettysburg.edu/about/offices/provost/advising/honor_code/index.dot.

What is permitted:

• AI-generated code where (1) you include your prompts in comments, and (2) you have thoroughly understood, can explain, and have tested such code. Required reading to abide by: Prof. Presser's Generative AI Guidelines
• Discussing algorithms, syntax, and problem specifications with other students
• Seeking assistance (as needed) from CS professors, students assistants, and approved tutors
• Use of "starter code" supplied by your professor, or code you have previously developed

What is not permitted:

• AI-generated code where (1) you have not included your prompts in comments, or (2) you do not understand, cannot explain, or have not tested such code.
• Sharing code: Verbal dictation of code, transference of code (e.g. e-mail, file transfer, entry of printed output, etc.).  An important component of this experiential learning process is developing the ability to independently devise algorithms and express them in program code.
• Viewing and/or using unauthorized code (e.g. solutions found online, downloads from code repositories, etc.)
• Submitting work that implicitly takes credit for work that should be credited to others
• Efforts to obscure the source of unauthorized work
• Outsourcing work

Put simply, students may discuss assignments at an abstract level (e.g. specifications, algorithm pseudocode), but must actually implement solutions independently or in permitted groups.  Credit should be given where credit is due.  Let your conscience be your guide.  Do not merely focus on the result; learn from the process.

For CS 391, use of generative AI will be permitted in all assignments. However, keep in mind the policy for permitted use. If you do not comment, thoroughly review and test, and/or fully understand the code that you present as your own, it is effectively considered a plagiaristic Honor Code violation.