Quick Facts

(more pics)
Name: baylor wetzel
Job: Lecturer
Department of Computer Science
E-Mail: wetzel@cs.umn.edu
Phone: No phone
Office: Shepherd Labs 583
Mailing Address: University of Minnesota
4-192 EE/CS Building
200 Union Street SE
Minneapolis, MN 55455

Frequently Asked Questions

Teaching

Spring 2014

Fall 2013

Spring 2013

Fall 2012

Current Projects

Project Insight. In many fields of human endeavor, from sales, marketing and design to debate, discussion and dating, an important part of success is adapting to your audience. The goal of Project Insight is to collect data on game players and use that data to automatically adapt the game to the player.

PaperDoll RPG. A role-playing game where fashion matters. Less a research project than an actual game. Being developed by Shikigami Games, which is the tiny company setup for me, friends and maybe a student or two to work on games focused more on entertainment than research.

Past Projects

GopherTD. We are studying how people learn, represent, transfer and apply problem solving knowledge. We are specifically investigating how people solve complex spatio-temporal reasoning problems and how they improve with experience. We are collecting data on human performance in the tower defense domain and using it to create a computer agent capable of solving arbitrary TD puzzles. Our primary interests are in strategy representation, strategy-based transfer learning and reasoning with spatial affordances.

PaperDoll Fashion Reaction. PDFR began as a research project investigating explicit cultural representation and scalable agent authoring tools and has gone on to become a commercial game being developed by Shikigami Games. In addition to the standard commercialization issues (art, polish, etc.), we are investigating knowledge merger techniques (specifically conflict resolution), influence attenuation and inheritance-based design by exception for cultural representations.

Research

My work is a mixture of artificial intelligence and experimental psychology. i build generative behavioral models, which means i study how humans solve problems then build computer programs that can solve problems the same way. i also work in natural computing, which means i use human problem solving as an inspiration for building better computer programs. The first is like studying birds and then making a mechanical bird. The second is like studying birds and then using what you've learned to build an airplane.

My specific interests are in ecological rationality, environmental cues, fast and frugal heuristics, concept representation and generalizable learning. i'm less interested in rationality and optimal behavior and more interested in what real people do, why they do it, what information they use and how they represent that information. While numerous laboratory experiments have proven that people are irrational, i work off the belief that, in the real world, people make the best possible decisions.

My domains of study generally fall into the average guy category. Specifically, i'm interested in problems that almost every human can solve easily but which computers have a hard time. The easiest tasks in this category involve video games. Games have three great advantages. First, games are the right difficulty level for humans. They can't be too easy or the player gets bored. They can't be too hard or else the player gets frustrated. Second, since the task is done on a computer, it's easy to collect raw data. Finally, most games represent problems that are incredibly difficult for current AI techniques.

My dissertation (Representation and Reasoning for Complex Spatiotemporal Problems: From Humans to Software Agents) used tower defense games to study how people represented and reasoned about space and time, how that enabled them to get better with experience, how they used knowledge derived from old problems to solve new ones and how all of that could be used to make a fairly nifty artificial agent that could solve novel tower defense problems at an expert level on the first try.

i also do a little work on games and simulations that is not based on simulating real human behavior, although a lot of it does use human psychology (especially all the cheating and player management algorithms).

i'm also in data mining. Right now that means strategy inference. That means that i try to infer a person's strategies from a set of generated artifacts (in this case, solutions to game problems). To make life easier, i assume the the set of base strategies is known. Unfortunately, in real problem solving, agents can use an arbitrary number of strategies (including 0) and strategies can be merged into novel, unknown strategies. Also, many strategies generate the same outcome. It's pretty tricky.

Games

This is how i got started with computers. It was a video game programming class at The University of Texas that convinced me to leave journalism and learn how to program a computer. i liked role playing games, Austin was the home to Origin (maker of the most popular RPG games in the world) and many of my friends worked as designers and programmers at Origin, so it should be no surprise that i started by writing RPGs (i also made, with now industry veteran Allen Jackson, an Axis & Allies game i was particularly proud of).

Since then, i've made many games, some for research, some for fun, some to learn and some to teach. In addition to my current projects (GopherTD and the commercial PaperDoll Fashion Reaction), we've built GopherTD (a tower defense game used to study human spatial reasoning), AI Game (a scriptable turn-based strategy based on Heroes of Might & Magic IV used to teach AI programming), The Dating Game (a dating sim used to teach production rules), Counter-Strike Tower Defense (a hostage rescue game based on cs_militia used to study opponent modeling), the Harry Potter simulator (a virtual reality spell casting game used to study 3D gesture recognition), The Game of Love (an adventure game about feelings), and various other small games. Professionally i worked on Alelo's Tactical Language and Culture series (Tactical Iraqi, Tactical Pashto, etc.) as well as a few indie titles and maybe one or two much older commercial titles. i also have a friend or two at Rockstar, Nintendo, Sony, Origin, NCSoft, Iguana, Digital Anvil, Bungie, Monolith and whoever it is that made Silent Hill, Worms, Killzone and a bunch of other games like that.

For a partial list of games i've written in the last couple of years, click here.