Paper Reading #2: Protecting Artificial Team-Mates: More Seems Like Less

Intro:

• Protecting Artificial Team-Mates: More Seems Like Less
• Merritt, Tim, and Kevin McGee. (2012).  Protecting Artificial Team-Mates: More Seems Like Less. Proceedings of the 2012 ACM annual conference on Human Factors in Computing Systems (CHI 2012), 2793-2802.
• Author Biographies:
• Tim Merritt is a current PhD student at the NUS Graduate School for Integrative Sciences & Engineering in Singapore. He began his PhD in 2008 and is studying under Kevin McGee.
• Kevin McGee is an associate professor at the National University of Singapore. He teaches in the Department of Communications and New Media. His research revolves around partner technology design/implementation and studies for entertainment purposes.

Summary:
The goal of the research is to determine the cohesive nature that human gamers adopt for artificial intelligence beings in gaming situations. In fact, the belief that a gamer has about the identity of a teammate has tremendous impacts on behavior, despite the true identity of the teammate. Thus, a middle ground must be established between behavior and interpretations of game events.

The study focused on gamers behaviors that have the option of "drawing gunfire" away from teammates onto themselves. The experiment was repeated twice with A.I. teammates each time. However, during the second game, researchers told the participants that they were playing with a human teammate. For shorthand, the latter will be referred to as PH, for presumed human. The game interface looks similar to the figure below:

The measurements of the game involved the actual number of times the player decided to "draw gunfire" and the number of times the player reported that he/she drew gunfire after the gaming session was over. After the experiment was over, a series of eleven questions were asked to each participant to gauge self evaluation, predisposed stereotypes, personal pressures, and explanation of observed behaviors. The conclusion of the research was that humans were more cooperative with AI figures that PH figures. However, this contrasts the self-reporting at the end of the game in which gamers declared themselves more cooperative with the PH teammates.

Related work not referenced in the paper:
1) "Developing & Validating a Synthetic Teammate" by Dr. Christopher W. Myers
2) "Real-time team-mate AI in Games"  by McGee and Abraham
3) "Teammates and Trainers: The Fusion of  SAF’s and ITS’s" by Schaafstal, Lyons, and Reynolds
4) "TeamMATTE: Computer Game Environment for Collaborative and Social Interaction" by Thomas and Vlacic
5) "Behavior Modeling in Commercial Games"  by Diller
6) "Evolution of Human-Competitive Agents in Modern Computer Games" by Priesterjahn, Krammer, Weimer, and Goebels
7) "Applying Collaborative Intelligence to RoboCup" by Carrera
8) "Approaches to measuring Difficulties in Computer Games" by Costello
9) "The Evolution of Abstract Resource Sharing Dilemmas Computer Games" by Cunningham
10) "Team Based Behaviour in Artificial Intelligence for Real Time Strategy Games" by Burke

The work in most of these papers is novel. Although most are related to the gaming industry, this can be instructional for real life human-computer interaction. However, the one inconsistency I would abide in change for is the lack of the combination of sophistication and relevance. Most of the papers, including this one, contained either high technology techniques or conclusive findings that were relevant and useful, but never both. In these papers, the related work section was complete and helped direct me to other similar sources on the topic of artificial intelligent teammates in game type environments. In essence, the difference of this paper derives from the evaluation of humans and presumed humans playing on the same team and how the humans evaluate their experience.

Evaluation:
In order to evaluate the results, paired samples T-tests were used to compare the logged data during both sessions. The researchers quantitatively and recorded unbiased measured the number of times the human player distracted the gunman away from there teammate, which was more for the AI player. However, the questionnaire at the end of the game will help dissect the subjective aspects of the research conducted. The first question, was quantitative and subjective which measured the amount that the gamers thought they helped the PH more than the AI, which was 71%. A reason behind this was a sense of empathy.

The remaining questions involved the subjective nature of participants. They were instructed to evaluate what they thought the teammate was "thinking" at the time or what the objectives were. The researchers analyzed all responses on a subjective nature since the students provided open ended results or rated a given question on a scale from one to five. In essence, the researchers did an effective job of utilizing both quantitative and subjective results.

Discussion:
All in all, the paper brought about some startling discoveries related to humans and how they interact with computers and presumed humans. Although players wanted to appear more loyal to their fellow species, they actually aided the AI teammates more than the presumed human. The authors attribute this altruistic behavior to that of the humans thinking that AI was inferior to their own, and thus required more help. There are a variety of other explains offered, but this is the most logical.

My considerations of the research include that the authors touched upon a hidden gem in the collaboration between human and machine. However, their experiment was trivial and extremely primitive. I would not base any conclusions on the simplistic nature of their work. Although, I would definitiely want the researchers to conduct their experiment on a deeper level with more realistic terms. Their evaluation was mostly subjective, but they did a proper job of analyzing the feedback and noting the limitations. In essence, this topic provides some interesting insights, and should be examined further.

Paper Reading #1: BiTouch and BiPad: designing bimanual interaction for hand-held tablets

Intro:

• BiTouch and BiPad: Designing Bimanual Interaction for Hand-Held Tablets 
• Wagner, Julie, Stéphane Huot, and Wendy E. Mackay. (2012). BiTouch and BiPad: Designing Bimanual Interaction for Hand-Held Tablets. Proceedings of the 2012 ACM annual conference on Human Factors in Computing Systems (CHI 2012), 2317-2326.
• Author Biographies:
• Julie Wagner received her Bachelor and Master degrees in computer science from RWTH Aachen University in Germany. She is currently in her third year of her Ph.D at Univ Paris-Sud, specializing in Computer Human Interaction. She has student teaching internship experience at her respected undergraduate and graduate universities. She published papers in CHI in each year from 2009-2012. Her paper in 2011 received the 'Best Paper Award'. Also, she has presentation experience at CHI in 2009 and 2010.
• Stéphane Huot received his Ph.D in computer science from University of Nantes and École des Mines de Nantes in 2005. His thesis covered computer human interaction. He is currently an associate professor at Université Paris-Sud 11. He received an award from 2011-2013 that granted him leave for research at Inria as a full time researcher.
• Wendy E. Mackay received a Bachelors from University of California, San Diego in psychology. She also earned her Masters degree from Northeastern University in experimental psychology. She has a Ph.D from MIT in Management Technological Innovation in 1990. She was professor in computer science department at University of Aarhus. She then became a senior researcher at Inria for two years before she was promoted to Research Director at Inria for the past ten years.

Summary:
The use of tablet like technological devices has become ubiquitous and essential to society over the past few years. Currently, tablets support different motions, such as tap or swipe of one hand, but do not support any sort of bimanual input. We know that bimanual interaction among desktops increases not only performance, but accuracy as well. Thus, the goal of the research is to implement a bimanual design and then evaluate its effectiveness while considering the need to hold the device.

The domain of the research focuses on Apple's iPad due to the larger screen than its relatives consisting of smartphones and PDAs. The conducted research will be implemented primarily through software due to the availability of additional iPad hardware purchases that stimulate bimanual usage, but are rather impractical.

First, the researchers discovered the unconscious methods that different users hold the iPad. They discovered a total of five positions with the fingers and thumb on the iPad border, for both landscape and portrait mode. The different positions can be seen in the figure below. Also, all participants of the study switched hands once while others changed holding positions several times throughout the experiment.

A BiPad toolkit with specific widgets for a test array of applications was then designed. This allowed the researchers to measure the utilization of the non-dominant hand based on the holding position the user chose. These BiPad zones appear on sides and corner of the screen. The implemented interaction among users include the tap, which is a single press, the chord, which allows multiple finger presses, and finally the gesture, which involves the sliding in an array of directions using one finger.

First, an example of the bimanual touch implementation includes a PDF in which the user is able to sift through using the dominant hand while simultaneously accessing the menu with the non-dominant holding hand via a side bar. Furthermore, a very useful design is custom buttons for the non-dominant hand such as a space bar for ease of typing and reduce movement of the dominant hand. Lastly, another potential use is spacial multiplexing in which is user is able to zoom in and out using the non-dominant hand while the dominant hand controls primary application usage.

Related work not referenced in the paper:
1) "Bimanual and Unimanual Image Alignment: An Evaluation of Mouse-Based Techniques"
2) "HabilisDraw DT: A Bimanual Tool-Based Direct Manipulation Drawing Environment"
3) "A Comparison of Tracking- and Controller-Based Input for Complex Bimanual Interaction in Virtual Environments"
4) "Exploring Bimanual Camera Control and Object Manipulation in 3D Graphics Interfaces"
5) "Bimanual Interaction for Tablet Computing"
6) "Motor Behaviour Models for Human-Computer Interaction"
7) "Actions and Consequences in Bimanual Interaction Are Represented in Different Coordinate Systems"
8) "Determining the Benefits of Direct-Touch, Bimanual, and Multifinger Input on a Multitouch Workstation"
9) "Bimanual Interaction on the Microsoft Office Keyboard"
10) "Harnessing the Benefits of Bimanual and Multi-finger Input for Supporting Grouping Tasks on Interactive Tabletops"

Overall, the work in these papers was extremely related to the BiPad and BiTouch research. In specific, they mostly deal with the interaction of technology and two human hands. Although, each inspects a different domain of usage, their ideas are still cohesive. For the most part, these papers all provided novel work. Their objectives was to increase the speed and utilization of technology. Although no gargantuan breakthroughs are to emerge from these works, there will still be small amendments to technology that may lead to a larger discovery along the line of technological advancement in the future. In fact, these papers did discuss related work appropriately. There is a wide range of research involving the usage of technology and bimanual utilization, so the papers each had a large reference point among one another.

Evaluation:
In terms of evaluation, the researchers set out to determine if the bimanual usage is quicker than a single hand and the trade-offs among the different orientation and hold types. The researchers used a team of subjects to properly evaluate their system. They independently analyzed the differences in technique, orientation, and hold. A special program was developed to test the quickness of two hands versus one through the tapping of buttons as they appeared on the screen. In order to properly monitor the effectiveness, the trial time, BiPad reaction time, BiPad completion time were all recorded for further analysis. In addition, the comfort level of the bimanual design versus the traditional one hand design was considered for a thorough and complete evaluation of this research.

First, the researchers discovered that bimanual taps were more efficient in both landscape and portrait mode than a single hand. However, bimanual chords and gestures were only faster in portrait mode. This can be attributable to the fact that there are shorter traversal distances for the single hand in landscape mode. All methods of evaluation used time trials and were thus quantitative and unbiased in nature. The most preferred method of human-computer interaction was the bimanual tap. In conclusion, bimanual maneuvers were overall faster than the single hand on a global scale.

Second, the researchers look to analyze the trade offs of the BiPad. They discovered that having the non-dominant hand on the same side is most efficient rather than on the opposite side or underneath the iPad device. Overall, bimanual taps were similar across the board for all holds and orientations. This was later confirmed for gestures and chords.

As far as comfort is concerned, taps were significantly the winner compared to gestures and chords. There were some abnormal positions for users such as applying a chord while holding the iPad from the top position. However, the unnatural state of the maneuver would not be executed in usage due to its awkward feel.

It should also be noted that the evaluation was systemically broken down as previously described. Instead of measuring the entire new design at one time, the researchers used separation of parts to measure each gesture in each mode with each holding position. This provided the most encompassing results.

Discussion:
In conclusion, not only was the bimanual implementation able to be cohesive with the iPad, but it did outperform the traditional one hand method in the timed trials. The breakthrough of the BiPad is that it lets users decide which hold, orientation, or gesture is most appropriate for their application as well as the one that adheres to there comfort levels. This proves to be useful for users that frequently change holds or ones who desire speed.

Overall, I thought the work was extremely intriguing and directly applicable to modern society. The speed up of iPad users would inherently be widespread due to the large volume of iPad owners. One of the most interesting aspects I enjoyed was that the analysis was conducted in both landscape and portrait mode. The work provided in this research is not ground breaking by any means. It will not drastically change the lifes of many, but it is some insightful work into improving modern technology. Although I would not consider this work eye popping, it was very interesting to read and will possibly show up in a future version of a touch pad.

 This enabled insights into different ways that different people access this type of technology. The evaluation itself was sound. They explored every possible hold, with every possible position, with every possible gesture. They did this independently to fully extract all results with no overlap. I would be curious about the correlation between comfort and speed of the device. Since users who tend to feel more comfortable will most likely perform faster, then this must be analyzed to properly evaluate the BiPad.

About Me

Email:  erikkatzen@tamu.edu

Class Standing:  5th year senior

Why am I taking this class: Besides the fact it is required under one of the computer engineering tracks, I find the interaction among humans and computers extremely fascinating. The overall concept of developing systems that appeal to human utilization is intriguing.

What experience do I bring to this class:  I am double majoring in mathematics along with computer engineering. I completed the USRG at Texas A&M with Dr. Dos Reis working on his computer algebra system. However, my interests have transitioned into business recently. I just completed a summer internship with Goldman Sachs in Houston.

What are your professional life goals:  My plan involves acquiring a full time job with a prestigious strategic consulting firm. From that point, I want to specialize in management consulting for technology firms.

What are your personal life goals: Outside of my career, I want to continue my hobbies of sand volleyball, tennis, hunting, fishing, archery, and weight training. Also, I want to start an awesome family at some point down the line.

What do you want to do after graduation:  I would prefer to enter the consulting business at a firm such as BCG, Bain, or McKinsey. If I was able to work in Texas, that would be even better.

What do you except to be doing in 10 years:  I plan to attend a lucrative MBA program after pursuing a job for several years. Following my MBA, my ideal job would be at a venture capitalist firm that invests in entrepreneurial and upcoming technological ideas.

What do you think will be the net biggest technological advancement in computer science:  I imagine automated personal transportation services. This new innovation would involve hoping into my car, plugging my smart phone in, running the app, and selecting a destination. Not only could I have two free hands and one free mind while my car takes me to my destination, but also my car could drop me off, and find a parking spot on its own. The days of $20 vallet are over.

If you could travel back in time, who would you like to meet and why:  I would like to meet Carl Gauss. Not only was he the undisputed greatest mathematician of all time, but also, I would be able to delineate the type of society that I currently live in. Then, I could proceed to invite him into discussions about which direction mathematics and our society are traversing down.

Describe your favorite shoes:  I love my Margaritaville boat shoes. First, they are significantly better than any type of generic Sperry. Second, any time I put them on, it reminds me of deep sea fishing with nothing to worry about except for enjoying the simplicity of life.

If you could be fluent in any foreign language, which one would it be:  I would want to learn Italian. First, Italians are probably the coolest Europeans out there, even though they can't pay off their debt. Second, Italians just sound articulate when they speak. I could see myself retiring in Italy.

Interesting fact/story:  I picked up tennis my Freshman year of high school. By the time I was a Sophomore, the ex-pro I trained with gave me an ultimatum. He said that if I quit basketball, and devoted the same amount of time I spent on the basketball court to tennis, then he could practically guarantee me that I could earn a Division I scholarship within the next year. I declined and decided to continue playing basketball and tennis concurrently. My senior year, our basketball team went 33-3, and I still made it to the tennis state championship as well.