Paper Reading #23 - Automatic Warning Cues

Comments:
Comment 1
Comment 2

References:
Title: Evaluating Automatic Warning Cues for Visual Search in Vascular Images
Authors: Boris van Schooten, Betsy van Dijk, Anton Nijholt, and Johan Reiber
Venue: IUI 2010, Feb. 7-10 2010

Summary:
In this paper, the authors perform a study on automatic warning systems for MRA machines. They find that warning systems that warn more often but produce false positives work better than both the ones that warn less or ones with no warnings at all.

The authors created a warning system for viewing images of the vascular system. In order to test it, they created a series of test vessels for users and the system to find. The vessels were classified by difficulty and type of problem. They then had users attempt to find these errors using a system that warned more, a system that warned less, a system with no warnings, and a "perfect" system that generated no errors.

They found first of all that the perfect system did the best. After that, the warning system with more warnings did next best, even though previous studies have shown the opposite to be true. Following that, the less warning and the no warning systems placed second-t0-last and last, respectively.

Discussion:
I think that this research is of great importance, since proper detection of blood vessels could prevent heart attacks and other vascular issues. Furthermore, I am sure these warning systems have other applications that are just as useful.

One issue I have with the paper, however, is that I do not think that their study was big enough. The results from the small group that they tested show nearly even results between false positives and negatives; they should have done more research to see which one did better.

(Image courtesy of: Imaging Group)

Special Reading #4 - Media Equation

Comments:
Comment 1
Comment 2

References:
Paper 1:
Title: Machines and Mindlessness: Social Responses to Computers
Authors: Clifford Nass and Youngme Moon
Venue: Journal of Societal Issues, Vol. 56-1, 2000

Paper 2:
Title: Computers are Social Actors
Authors: Clifford Nass, Jonathan Steuer, and Ellen R. Tauber
Venue: CHI '94, April 1994

Paper 3:
Title: Can Computer Personalities Be Human Personalities?
Authors: Clifford Nass, Youngme Moon, BJ Fogg, Byron Reeves, and Chris Dryer
Venue: CHI '95, May 7-11 1995


Summary:
In these papers, the authors show that people unconsciously think of computers as people -- in some respects, anyway. When it came to factors like gender, race, and aggressiveness, their tests indicated that indeed humans applied these qualities to machines, even though they said doing so was ridiculous.

They used many methods for the tests, but the most common was the three-computer setup shown at right. One computer feeds information to the user, then they take a test on computer number two, and then the third evaluates their scores. Different selections of male and female voices as well as aggressive and less aggressive language to use.



Discussion:
I wasn't very surprised by the results of these papers. I have referred to computers in a human way before, especially when I get frustrated. Also, when you place human voices on the machines, you make them seem more humanlike, so it's not surprising that gender and racial biases get applied.

However, the fact that these results are so evident means that we must be careful about what sound clips we use when we design programs with a voice component, as well as how we give information through the user through text. Incorrect usage could make our program seem rude and cause people to dislike it.

(Image courtesy of: Paper #2)

Paper Reading #22 - POMDP Approach

Comments:
Comment 1
Comment 2

References:
Title: A POMDP Approach to P300-Based Brain-Computer Interfaces
Authors: Jaeyoung Park, Kee-Eung Kim, and Sungho Jo
Venue: IUI 2010, Feb. 7-10 2010

Summary:
In this paper, the authors describe a method of optimizing the number of attempts a brain-computer interface (BCI) needs to use to find information from the user. They did this by creating a program that displayed a 2x2 or 2x3 matrix to the user and then flashed prospective letters at them (at left).

They used a learning system called POMDP to attempt to guide in the flashes to what the user wanted. This system required a lot of training before the system could be used on human subjects. However, when they tested it on humans, they found that the accuracy was much higher by 30 flashes than the normal algorithm, and maintained that accuracy all the way to the maximum of 60.

Discussion:
This paper was interesting because we get to see some more bleeding-edge research in BCI. The paper this time was significantly more readable than the previous one, but I still got lost in the probabilities.

One thing that bothers me again is I don't really see the big benefits of this at this stage. Of course, EEG-based programs are still in their infancy, but I don't see how flashing letters can later become navigating the pointer or typing with my mind. Maybe in a few years I will be able to see the connection.

(Image courtesy of: this paper)

Microblogs #9 - Media Equation

Part 1:
Summary:
In this paper, the authors show that people have a tendency to mindlessly apply human characteristics to computers, even if consciously they believe the idea to be ridiculous. They show in a series of tests that people profiled computers based on gender, ethnicity, and loyalty, just by slightly changing the stimulus that the computer gave. Additionally, they displayed social behaviors as well.

Discussion:
To be honest, I didn't find this very surprising. I always refer to my computers like they are people, especially if they act up. Now that I write that down here, that is kind of weird to say, but it's true nonetheless.


Part 2:
Summary:
In this paper, the authors try to see if human social cues will be applied to computers. They attempt five different tests: politeness, self and other, voice self and other, gender, and programmer v. computer. They found that people do in fact apply politeness and gender roles to the computer based upon its voice.

Discussion:
To me, this paper is basically a repeat of the previous paper. However, it is interesting to see a repeat of the validity of the experiment, even though I think it is quite obvious that people place human qualities on computers.


Part 3:
Summary:
In this paper, the authors demonstrate that personality can be given to a computer without any special artificial intelligence. They showed that changing the way computers gave information not only made them seem more dominant or submissive, it also made people like or dislike them more based on their own personalities.

Discussion:
Again, no surprises here. I have applied personalities to people I have never met when reading a book, so finding it out that people apply it to a machine isn't incredibly exciting.

Paper Reading #21 - Automatically Identifying Targets

Comments:
Comment 1
Comment 2

References:
Title: Automatically Identifying Targets Users Interact with During Real World Tasks
Authors: Amy Hurst, Scott E. Hudson, and Jennifer Mankoff
Venue: IUI 2010, Feb. 7-10 2010

Summary:
In this paper, the authors describe a method of gathering user click data in an accurate, device-agnostic way. They do this by using a hybrid method of kernel-based tracking combined with image identification.

Their user data gatherer, called CRUMBS, works in two levels. At the lower level, a series of different data gatherers reports on what they each think is what the user clicked on. For example, some of the low-level gatherers include Microsoft's Accessibility API, an image difference checker, as shown above, and a template checker. Then, at the high level, a machine learning interface decides based on all the data gathered from the low-level gatherers to make a final decision on what the user clicked on.

With their method, they reported a 92% correct click identification rate, which they mention is higher than using only the accessibility API. Furthermore, they mention if they captured a larger portion of the screen on a click (they currently grab only a 300x300 space), they could get an even larger portion of clicks correct.

Discussion:
I think that CRUMBS could be a very useful tool to use when testing how users make use of your program. If the collection data from other sources is as bad as they say in the paper, then it is very difficult to gather real usage information for programs, and this could help.

One thing I am curious about though is if this information gatherer must be turned on and off manually or if it does so automatically. Otherwise, it might gather usage information in a clicking space where it isn't needed, such as in a video game.

(Image courtesy of: this paper)

Book Reading #6 - Things That Make us Smart

References:
Title: Things that Make us Smart
Author: Donald Norman
Editor:Bill Patrick, 1993

Summary:
In this book, Norman talks about how technology can assist or hinder us in our daily lives based on how well it is designed. He begins by discussing how most design today is technology-centered, which is why it's so confusing to use. Instead, if the designs were human-centered, they would fit much better into our lives.

He then discusses two states of the mind, experiential and reflective, and discusses how good designs place us in the right state for a task, while bad designs do not. He then discusses the importance of using the correct designs when displaying data as well as using the correct tools when working on a job. Doing so will keep you in the correct mindset for the task.

Discussion:
I feel like for learning how to design, this book is not as important as the last two we have read, as it talks less about the design of products and instead looks at their effects on the user. However, as an examination of the effects of technology on us it is quite successful. I am definitely going to think more about how I use objects in relation to the tasks I am trying to perform from now on.

(Image courtesy of: PBS)

Paper Reading #20 - Data-Centric Physiology

Comments:
Comment 1
Comment 2

References:
Title: Addressing the Problems of Data-Centric Physiology-Affect Relations Modeling
Authors: Roberto Legaspi, Ken-ichi Fukui, Koichi Moriyama, Satoshi Kurihara, Masayuki Numao, and Merlin Suarez
Venue: IUI 2010, Feb. 7-10 2010

Summary:
In this paper, the authors describe a new method of analyzing the emotions -- or affect -- of people. They describe some of the problems with current emotion modeling solutions like the time it takes to analyze them, and how they believe it can be improved by changing how the data is analyzed.

They mention that analyzing the data continuously along the entire spectrum of data for a user will produce better results than the current method of discretely analyzing emotions. To prove this, they analyzed the emotion changes of two subjects by using the sensors shown in the pictures above and playing music that affected them emotionally.

Then, they describe in detail the algorithms behind their continuous analysis, and show that it is as fast as discrete analysis and should provide better results in certain situations.

Discussion:
To be perfectly honest, this paper was so difficult to read that I'm not exactly sure that I got the correct analysis out of it. It took me half of the paper to figure out what they were trying to do with the emotion readings, and I am still not exactly sure what the point was.

Additionally, I am curious what the benefits are behind sensing emotions of users, especially if it requires the elaborate equipment shown in the picture. I have seen some cool little games that used the user's emotions to modify the game, but no other real applications.

(Image courtesy of: this paper)