Paper Reading #26: Embodiment in Brain-Computer Interaction

Kenton O'Hara, Abigail Sellen, and Richard Harper are all researchers at Microsoft Research at Cambridge.

This paper was presented at CHI 2011.

Summary


Hypothesis
The hypothesis of the paper was that people react to Brain-Computer Interfaces in many different ways.

Methods
The Brain-Computer Interface (BCI) used in this research experiment is a game known as MindFlex (pictured). The "game board" has a fan underneath the construction which levitates a small ball into the air. The player controls the device via a wearable sensor on the forehead. When the player concentrates, the fan speeds up which causes the ball to rise. If the player relaxes, the fan slows down and the ball is lowered. As the game board rotates, the player must overcome obstacles by lowering or raising the ball by concentrating.

In order to test their hypothesis and to study the use of BCI interfaces in the "real world", researchers recruited 4 different groups of people. For each group, one person was in charge of getting the people for their group. The four groups were decently diverse. One group, for example, was a cohabiting couple. Another group was a family of four. These groups were all asked to play the game over the course of a week and to record themselves playing the game. The video footage was then analyzed and trends were discussed.

Results


The researchers gathered several different interesting trends.

First, they found that people attempted to concentrate by changing their body orientation. For example, one player held her breath and leaned close towards the ball. They also discovered that players developed strategies to cause the ball to lower through relaxation by looking away from the game board.

They also analyzed how people tried to visualize their concentration. For example, many of the groups who played talked about making the ball rise by thinking about making the ball rise and making the ball lower by thinking about it falling. In the game however, both of these thoughts require "concentration" so the ball would rise.

Another trend they discovered was spectator verbalizations throughout the game. Spectators played a roll in the game by either trying to make the current player mess up or by encouraging them to concentrate.

Discussion

Personally, I found this paper extremely fascinating. BCI, in my opinion, is the future of input devices. When they become more accurate and faster, the entire need for keyboards or controllers or other input devices can be completely eliminated.

This paper, while studying embodiment as it relates to BCI, gives some clues for future BCI work, in my opinion. In the thought visualization portion of the paper where participant tried to visualize lowering the ball, this can tell researchers what kind of "thought" is natural for a given action. More studies like this one can begin to map the natural "thoughts" for Brain-Computer interfaces.

Paper Reading #24: Gesture Avatar, A Technique for Operating Mobile User Interfaces Using Gestures

Hao Lu is a computer science graduate student at the University of Washington.
Yang Li is a senior researcher at Google Research

This paper was presented at CHI 2011.

Summary
Hypothesis
The researcher's goal was to create a system that allows for the easy interaction with mobile interfaces.

Methods
To test their hypothesis, the authors set out to create a system called the Gesture Avatar system. To them, one of the main problems of current mobile interfaces is the small size of many interface elements. For example, small buttons that are hard to use or tiny links on a webpage that are hard to click.

To counter this problem, the Gesture Avatar system allows the user to draw a shape or a letter near the object the user wants to activate and the system let's the user control the object from this larger "gesture avatar".
For example, a slider might be hard to activate while on the move or it might just be too small. A user could draw a box close to the slider they want to control and then slide that box like the slider.
Another example is for selecting small texts or clicking small links on a webpage. By drawing the first letter of the link they're trying to activate, the user can then click on the letter and the system attempts to figure out what link is meant by the user.

To test their system, researchers used a Motorola Droid running the Android OS. They found 12 participants to test the system. They used two different tests. The first was a letter selection test. A small amount of letters would show up on the screen and the user would have to select the highlighted letter by drawing the letter gesture and select it. The next test was simply a target selection. They also had some participants walk and use the device and some were asked to walk on a treadmill and use the device.

Results


They used three different types of criteria to judge the Avatar Gesture system: time performance, error rates, and subjective preferences.

For all letter sizes, Gesture Avatar's performance time was about constant. It was significantly faster than an alternate system (Shift) on the 10 px letters however.

Error rates also remained low and about constant over the three target size tests. The error rates are SIGNIFICANTLY lower than the alternate Shift system.

In the subjective preferences, 10 out of 12 participants preferred Gesture Avatar over Shift.

Discussion
The Gesture Avatar system is certainly a novel system of controlling a device. One of the main problems with mobile interaction is the limited screen space of a device. Screen size cannot be increased without making the device larger. Thus, for cell phones, making a larger screen is not really a viable option when it comes to fostering better user interaction.

One problem that might arise from the system is the fact that they only tested the system using two hands. Often times, when I'm in a hurry walking down the street, I only used one hand on my mobile device. I would have liked to see what would happen if someone just used their thumb to draw the gestures as opposed to a whole new hand.

Paper Reading #23: User-Defined Motion Gestures for Mobile Interaction

Jaime Ruiz is a PhD student of HCI at the University of Waterloo
Yang Li is a senior researcher scientist at Google.
Edward Lank is an assistant professor of computer science at the University of Waterloo.

This paper was presented at CHI 2011

Summary


Hypothesis
Researchers set out to prove that certain motion gesture sets are more natural than others in mobile interaction.

Methods
In this research experiment, researchers essentially allowed participants to create the gestures to be investigated. The authors created a list of tasks of several different types. The tasks would either fall into the "action" or "navigation" category. These categories were then subdivided into smaller tasks.

Participants were given the full list of tasks to be completed as well as a smartphone. They were instructed to complete the simple task (like pretend to answer a call, or go to the homescreen) with the idea that the phone would know how to execute the task.

The phone they were using was equipped with specially designed software to recognize and keep track of gestures performed by the participant.

After the test, the user was asked to comment on the gestures they used and whether they were easy to perform or good match for it's use.

Results
Through the test, researchers discovered that many users preferred and performed the same gesture.

The authors found several trends in the description of motion gestures. They found that many of the gestures mimic normal use. For example, 17 out of 20 participants answered the phone by putting it to their ear. The gestures also use real-world metaphors.

Through all these results, the authors were able to analyze the various created gestures and create a taxonomy of these natural gestures.

Discussion

Research like this, in my opinion, should occur more often. By giving participants free-range of their actions, they can come up with whatever input or gestures are natural to them. After analyzing these gestures, we can figure out which are natural.

By creating more studies like this, we can find out what types of input are preferred in a system. For example, do most people prefer to execute gesture XYZ to access the menu or do they prefer to use gesture ABC instead.

Paper Reading #22: Mid-air Pan-and-Zoom on Wall-sized Displays

Mathieu Nancel is a PhD student of HCI at the University of Paris
Julie Wagner is a postgraduate research student at the University of Paris.
Emmanuel Pietriga is a full-time research assistant at the University of Paris.
Olivier Chapuis is a research scientist at University of Paris.
Wendy Mackay is the research director at INRIA. 


This paper was presented at CHI 2011.


Summary
Hypothesis
In this paper, researchers experimented with various ways to control a wall-sized display. They had several hypotheses regarding which ways would be preferred. They believed that two handed gestures will be faster and easier to use than one-handed gestures. They also believed that users will prefer circular gestures. They predicted that small gestures would be preferred.

Methods
In the project, researchers tested various different types of gestures. First, they used uni-manual (one handed input) and bi-manual (two handed input). Each of these manual gestures could be used in conjunction with 1D control, 2D surface, and 3D free space.
The 1D control was control in one plane using a linear scrollwheel or a circular scrollwheel.
The 2D surface was control using a PDA or smartphone.
The 3D free space allowed gestures to be used.
With the bi-manual input, one hand was used to show the focus of the gesture, i.e. where to focus the zoom or pan, etc.

In the experiment, the researchers tested three factors: Handedness (one handed, two handed), Gesture (circular, linear), and guidance (1D, 2D, 3D).

Each participant was given a target in a high zoom scale. They had to zoom out to display all targets and they had to pan the targets so they were all visible on the screen.

Results
In most cases, they found that two hands (bimanual gestures) were faster and had shorter movement times
They also found that 1D path control was much faster than the other input types.
Linear control was also found to be faster than circular.

In the qualitative results, participants seemed to prefer the one handed gestures. They also preferred the linear gestures to circular gestures. They found that circular gestures on a surface were too hard to do.

Discussion
The paper itself was fairly interesting although it was slightly difficult to identify with since I've never personally used a large wall-sized display. I wasn't able to really agree or disagree with their findings since I've never had to use gestures on such a large display.

This kind of research could potentially become very helpful in the future when wall-sized displays become more widely used, though. Another user qualitative might be helpful in the future as well. By letting users design their own gestures, they could choose the most natural gestures for the large wall.

Paper Reading #21: Human Model Evaluation in Interactive Supervised Learning

Rebecca Fiebrink is an assistant professor at Princeton in the school of Computer Science but also is associated with the music department.
Perry R. Cook is a professor emeritus of Princeton in the school of Computer Science and the Department of Music.
Daniel Trueman is an associate professor of music at Princeton University.

This paper was presented at CHI 2011.

Summary


Hypothesis
The researchers main hypothesis was that while using a machine learning system, humans could evaluate and adjust the machine learning generated model to better train the system.

Methods
To test their hypothesis, researchers created a system called The Wekinator. The Wekinator is a software system that allows human users to train a given machine learning given certain gesture or real time input. Based on the human input, the system attempts to figure out the best course of action for the human input. If the system is incorrect or it isn't certain enough, the human can tell the system what the correct action is and the human can train the system by performing the gesture or input over more iterations.

To test Wekinator and discover how humans used it, they created three different studies.

In Study A, researchers had members of the Music Composition department (PhD students and a faculty member) discuss Wekinator and allowed them to use it over a given amount of time. When they were done using it, the participants sat down with the researchers and talked about how they used Wekinator and they offered various suggestions. Most of the participants used Wekinator to create new sounds and new instruments through the gesture system.

In Study B, students ranging from 1st year to 4th year were tasked with creating two different types of interfaces for the machine learning system. The first was a simple interaction where based on an input (like a gesture or a movement on a trackpad), a certain sound would be produced. The next type was a "continuously controlled" instrument that would make different noises based on how the input is being entered.

In Study C, researchers worked with a cellist to build a system that could respond to the movements of a cello bow and report the movements of the bow correctly.

Results
Researchers found that in all cases, participants often time focused on iterative model-rebuilding by continuing to re-train the system and fixing the models as they needed to be adjusted with given input.

In all of the tests, direct evaluation was used more than cross-validation. Essentially, directly evaluating the model and modifying it was preferred over the cross-validation model.

The researchers also noted that cross-validation and direct evaluation was actual feedback to the participants so that they could know whether or not the system was properly interpreting their input.


Discussion

The usage of a system like Wekinator could have some very far-reaching benefits. At the moment, machine learning seems pretty bulky and hard to do if you're not an AI programmer or scientist. But Wekinator seems to make the task of machine learning easier to understand and to put into practice. In the second study, a lot of the participants didn't know much about machine learning until they were briefed on the subject. However, they were still able to put Wekinator into practice to create new types of musical instruments.

Wekinator (and systems like it) could open the door on many different input tasks. Gesture and motion control can be better fined tuned by quick training. Since multiple people might perform a gesture differently, using Wekinator, a system's model can quickly and efficiently be "tweaked" to fit a given person. This makes analog input (like speech recognition or gestures) more reliable and more accessible to a large group of people and technologies.

Paper Reading #20: The Aligned Rank Transform for Nonparametric Factorial Analyses Using Only ANOVA Procedures

Since I completed Paper Reading #16, I'm choosing this blog as my blog to skip.

Paper Reading #19: Reflexivity in Digital Anthropology

Jennifer A. Rode is an Assistant Professor at Drexel University. She teaches in the school of Information.

This paper was presented at CHI 2011.

Summary


Hypothesis
Rode's main hypothesis and idea in this paper is to discuss the fact that ethnography are present and useful in the HCI field. She also discusses various types and aspects related to ethnography.

Methods
Through her paper, Rode talks about the needs and uses for what she calls "digital ethnography." A digital ethnography is simply an ethnography that is focused on technology in some fashion. One example that Rode gives in the paper is that of a researcher visiting super-churches and learning how each one employs technology.

To prove her points and to visit the various aspects of digital ethnography, Rode breaks down the various components. First, she covers the styles of ethnographic writing. Then she moves on to discussing ethnographic conventions before she looks at the framing ethnographic practices.

Results
First are the three styles of ethnographic writing.
The realist is interested in accurately displaying the subject. Rode writes that "the realist account strives for authenticity." The realist also goes to great lengths in order to show the reader the native's point of view.
The next kind of ethnographic writing is the confessional. In this kind of study, the writer comes up with a kind of theory on the subject, discusses it, then applies the theory to the ethnography and addresses it.
The final kind of ethnographic writing is the impressionistic. This is simply an attempt to "give an impression" of the subject. It is an attempt to capture the entire feel of the subject much like how the impressionist painters captured their subjects in a painting.

Next, she discussed the ethnographic conventions. The first discussed is the rapport that the ethnographer must build with their subject. If a researcher can't interact well with their subject, the subject is less likely to reveal information or to openly discuss a topic.
The next discussed was the participant-observation. Rode described this as simply "deep hanging out." By being around and observing the participant in their "native" environment, one can gain information.
The final ethnographic convention discussed was simply the use of theory. By coming up with theories on the subject, researchers can focus questions for focus their research.

Discussion

This paper was a good description of how ethnography applies to the digital world. This paper will be very helpful when we begin to work on the design project related to our class ethnography projects. It will help point out things to look for and ways to approach problems.