MusicSim: Integrating Audio Analysis and User Feedback in an Interactive Music Browsing UI

Related comments:

Summary:
Two researchers from the University of Munich, Ya-Xi Chen and Andreas Butz, conducted research in the area of music information retrieval (MIR) about integrating audio content analysis with metadata-based interfaces to create a program called MusicSim.  In their paper, they address users' increasing need to browse music in more diverse ways and the difficulty that non-experts have in navigating many current music browsing interfaces.  The researchers advocate active user control and feedback of the system to improve performance by integrating audio analysis in the UI and providing visual assistance to music browsing.  MusicSim presents songs clustered by content similarity and is controlled by the user through mouse operations; it supports a text-search feature, playlist generation, album-art and color-coded genre visualization.  The user can input feedback into the system by splitting or merging album clusters or adjusting a slider in the recluster control panel.  MusicSim implements very low-level audio analysis tools like jAudio to compute similarity between songs.

MusicSim was tested by 36 study participants, over half of which found the visualization UI very useful.  However, the system did have a few problems, such as a confusing graph view and the fact that genre understanding varies wildly from user to user and should be used as additional information rather than the basis for clustering.

Discussion:
I think MusicSim is a great idea that provides a good starting point for future work in music collection visualization and management.  In fact, a system similar to MusicSim was proposed by my senior capstone group for our semester-long project not too long ago.  We felt that users wanted as much direct interaction with their libraries as possible, and drafted a multitouch system that would utilize a large screen (thereby solving MusicSim's difficulty with the cramped graph view).  The clustering functionality of MusicSim is really useful for large libraries because it can help users locate what they want more quickly, and though using album art to represent music items is a no-brainer (and not exactly the most original idea in the world) it is nevertheless demonstrably effective and easy for users to identify with.

Passages Through Time: Chronicling Users' Information Interaction History by Recording When and What They Read

Related comments:


Summary:
In this paper, Karl Gyllstrom of the University of North Carolina Computer Science Department outlines his research in building interaction history for users to improve information retrieval and characterize document activity.  His system Passages collects data from text-based desktop activities like web-page content, emails, and other files, as well as precise timing information about these items' visibility.  These data combine to provide a very detailed record of what content was viewed and when.  This record can then be used to answer user questions about their content such as, "when did I read this paper," or "which documents have I spent the most time composing?"


The system is comprised of two subsystems: the tracing subsystem for recording event streams in the GUI and filesystem, and the retrieval subsystem for handling artifact requests ("when was the last time I read this document?") and temporal retrieval ("what files did i work on the most during this time period?").  The author goes on to detail the algorithms behind these subsystems and compare it to existing approaches, which do not currently consider how long users view material and give as much weight to glanced-over pages as ones that have been thoroughly pored over.  The author's user study examined 15 participants' use of Passages over a total of 14.27 hours of activity; the results indicated that the system was well-suited to adding nuance and precision to document history requests.


Discussion:
For the super-organized, this probably a dream come true.  Being able to know which documents you were reading during a certain period of time or finding out how much time you spent reading a paper could potentially be very useful.  I found a couple of the author's assumptions about the algorithm to be somewhat ill-founded; for starters, it is based on whether content is visible as the active window.  I have a 27" monitor and frequently have multiple documents open simultaneously, with one or neither set as the active window.  I may be reading either of them but not having that time or content logged into the system because of the way Passages is set up.  This obviously isn't going to be a problem for everyone, but I feel like the productivity-hawks who would use something like this would probably have large or multiple displays, thus rendering the system less effective.

More than Face-to-Face: Empathy Effects of Video Framing

Related comments:

Summary:

In this paper, David T. Nguyen of Accenture Technology Labs and John Canny of UC Berkeley's Institute of Design present their research about the effectiveness of different videoconferencing techniques in the context of empathic interaction.  The authors first cover some of the basic benefits of videoconferencing (time and money-saving for business interactions) as well as some of its shortcomings (lower degree of trust, decreased measure of non-verbal cues, disparities in gaze matching).  They mentioned several previous bodies of work (mainly pertaining to gaze preservation) and established their central hypothesis -- that the correct framing of the subjects in a video conference could reduce or eliminate any disparities between it and a face-to-face meeting. A basic hierarchy was established: 

• participants (or dyads) in face-to-face meetings exhibit the highest level of empathy
• dyads in upper-body-framed video meetings exhibit the next highest level of empathy
• dyads in head-only-framed video meetings exhibit the lowest level of empathy

Based on the findings of their study (which consisted of 62 test sessions with the various types of meeting/framing), the authors then presented some design guidelines for video systems that would ensure the highest level of dyad empathy in video conferences.  These basically consisted of measures that framed the participants' upper bodies and allowed the greatest amount of non-verbal body-language cues to be detected.

Discussion:

The old sci-fi B-movie staple of the "videophone" has become rather commonplace now with the inclusion of integrated webcams and chat software in most commercially-available laptops.  Users can place video calls to people across the planet (via Skype or other video conferencing client) and have empathic interactions in real-time.  Based on my own personal experience with video conferencing, I can attest that framing which favors the transmission of non-verbal cues are the most effective; much of the authors' research provided unsurprising results in this regard.  The most intriguing thing to me was the concept of gaze-matching.  This is often the one part of video conferencing that I (and high-powered business-types around the world) find the most trouble with -- dyads can never really make "eye contact" in a video conference, and some of the systems they discussed presented some cool solutions to this problem.

Learning from IKEA Hacking: "I'm Not One to Decoupage a Tabletop and Call It a Day."

Related comments: Jill's blog.

Summary:
Daniela Rosner and Johnathan Bean, two UC Berkeley students, present their findings about the activities of the DIY community, specifically in the area of IKEA furniture "hacking."  These hackers take pieces of IKEA furniture kits and cobble them together in new and interesting ways as a method of artistic expression, practical modification, or just to see what they can create.  The challenge in IKEA hacking is to create something new that has no instructions, though participants have increasingly reached out to other hackers via online communities like  instructables.com and ikeahacker.com.  The paper gives an overview of IKEA hacking and delves a little into the rationale for it -- DIY-types enjoy the change of pace provided by working with their hands to create something physical, in contrast with their daily jobs; the presence of IKEA products in homes and workplaces has become so prevalent that "hacking" provides participants a way to personalize their furniture and distinguish it from others; and displaying creativity and ingenuity in hacking can be very rewarding to the participant.

Discussion:
Being a participant in (or at least a follower of) one particular DIY community -- music -- I can understand and appreciate the appeal of creating something new from readily-available parts and seeking to put together something totally individual without any kind of manual or assistance.  One thing the paper hit on very well was the online community aspect of hacking.  Collaboration is a huge part of any DIY community -- participants can inspire and be inspired by the work of others, share ideas, show off their creations, and get help if necessary.  This kind of collaborative, supportive environment really fuels participation.  One of the most interesting points of this paper was the idea that IKEA has no style; that it has become so ubiquitous that any style it held has disappeared.  This "lack of style" really gives DIY-types a wide-open playing field to create something that would never hit a store or showroom floor.

The Inmates Are Running the Asylum (Chapters 1-7)

My initial reaction to Alan Cooper's The Inmates Are Running the Asylum was almost the same as to Donald Norman's The Design of Everyday Things -- these are the irrelevant rants of a cranky old professor who has nothing better to do than complain and would sooner send mail via carrier pigeon than learn to use "one of these newfangled machines."  I'm sure this was due in part to his less-than-hospitable treatment of programmers and his unapologetic tone, and part due to my tendency to not take seriously as a technology authority an individual who can't open a Word document.

As I read on, however, my critical attitude softened somewhat.  Some of his arguments (particularly concerning "dancing bearware" and "cognitive friction") seemed apropos to the modern discussion of software design.  A lot of what he argued was really a paraphrasing of Norman's points in TDoET, applied almost exclusively to interface design.  My annoyance (and sometimes anger) at his vitriol aimed toward programmers turned to pity over the course of the book as I slowly started to put myself in his place -- and time -- and realized that most of the mistakes he rails against simply aren't happening anymore.  
The software of the turn of the century was, as Cooper says, written and designed primarily by programmers who were as much concerned with finding discrete, creative ways to blame users for the problems with their programs (giving them an out for responsibility) as they were with making money from the stuff they continued to shovel onto hapless computer owners.  Obviously, this pattern revealed itself when the dot-com bubble burst and a large contingent of these shovelware developers disappeared as quickly as they had come.  No doubt much of Cooper's readily apparent bitterness stems from his time working with such companies.

The developers that remained, however, learned their lesson and have started to design in a much more user-centric manner.  Cooper's arguments didn't seem relevant to me because they aren't relevant -- companies are already doing the things he's talking about.  Additionally, I felt that some of his concerns didn't apply to a society where a large part of the population has grown up using computers (though he would undoubtedly classify our generation as "scarred" to the point of numbness by our experiences with poorly-designed interfaces).  I don't know how large a part his writings and work influenced the industry, but I do know that a lot has changed since this book was published.  Apple, Microsoft, and other developers spend millions of dollars on interface design for every product they make, and though I'm sure Alan Cooper may still think they're far from perfect, the software products of today are much more user-centric than those of the past.

The Application of Forgiveness in a Social System Design

Related comments: Jill's blog.


Summary:
In this paper, Asimina Vasalou, Jens Riegelsberger and Adam Joinson (a joint research team from the University of Bath and Google UK) lay out a framework for designing "reparative social systems" by defining the process of forgiveness in social interaction.  They bring light especially on the case of the online user who may offend unintentionally or accidentally; reparative systems may more easily allow trust to be re-established and the user returned to good standing in the online community.  They define forgiveness as "the victim's prosocial change towards the offender as s/he replaces these initial negative motivations with positive motivations."  These positive motivations can be influenced by a number of factors: offense severity, intent, apology, reparative actions taken, non-verbal expressions, dyadic history (previously rewarding interactions between two users), and history in the community.  Victims will assess all of the above factors when deciding whether to forgive the offender.  They expound further on the definition of forgiveness in the following ways:

• Forgiveness cannot be mandatory (it does not follow automatically after an offender's penance)
• Forgiveness is not unconditional (rather, it follows the offender's acknowledgement of responsibility and amends)
• Forgiveness does not necessarily repair trust or remove accountability

Forgiveness has its benefits -- offenders can relieve guilt or shame through apology or reparative action, victims may reduce or release their anger toward the offender, and it can empower an online community to learn by example and move toward self-moderation.  


From the definitions presented, the researchers arrived at the following provisions for reparative design in social systems:

• Respect the dyadic nature of forgiveness (overcome the asynchronous nature of online forum communication by notifying offenders and providing a grace period for response)
• Support the motivating factors of forgiveness (provide systems that allow the victim and community to measure an offender's previous and current actions, intent, apology, etc. and gives the offender tools to provide an adequate apology)
• Increase public awareness (make the offense public to educate the community)
• Build flexibility within and around forgiveness (allow victims to retract decisions of trust and accountability if desired)
• Design interventions to lower attributions (assess an offender's previous history to prevent victims from jumping to conclusions about an offender's motivation)

Discussion:
The research presented in this paper certainly gives a thorough definition of forgiveness as well as methods of enacting it; the problem with putting such systems in place in an online community is that the selection of communities willing to self-moderate and enact these kind of reparative systems are probably few and far between.  The researchers mention "trolls" all too briefly at the beginning of the paper and don't seem to realize that a troll can simply create another account on a forum if s/he got banned for offensive or abusive conduct.  The systems they mention apply most directly to buyer/seller sites like Amazon, eBay or Craigslist where buyer/seller trust is paramount, and they do a good job of mentioning their system's applications in this respect.  However, they seem to view these systems as being more widely applicable than they really are.

“Pimp My Roomba”: Designing for Personalization

Related comments: Zach's blog

Summary:

In this paper, JaYoung Sung, Rebecca E. Grinter and Henrik I. Christensen of Georgia Institute of Technology dispense the results of their research on the effects of personalization on user interaction with Roombas.  They postulated that people will naturally personalize their Roombas (or any other object, really) if encouraged by the design of the object or given a set of tools to do so; they also wanted to see if people's experiences with the device were positively impacted by their personalization.  The researchers did a study of 30 households in the Atlanta area, each of which were given a Roomba.  15 of the households were given "personalization toolkits" to use on their Roombas which included stickers, lettering sets, coupons for Roomba skins, and a booklet showing how other users had customized their Roombas in the past.  The other 15 households were given no toolkit or any other indication that customization was possible or desirable.

It turns out that 10 of the 15 households with toolkits used them or went online to order additional skins with the included coupons, but of those, only 6 actually customized the Roomba.  There were three primary motivations behind this customization: expressing the Roomba's identity, show its value in the household, and to make it either stand out or blend in with its home environment.  For example, some families gave their Roomba a name and decorated it differently according to its gender to give it some personality.  Other families decorated the Roomba as a "reward" or out of "gratitude" for the services it provided to show its worth to the family.  Finally, some of the households decorated Roomba to make it stand out from the carpet or decorations, and others tried to make Roomba fit into the aesthetic of their home.  The end result of the customization was a feeling of connectedness to Roomba, with householders seeing it as "their" robot instead of "a" robot.  Conversely, none of the 15 households who did not receive toolkits decorated Roomba.

Discussion:
Personalization is ingrained in our society.  Think about it -- what's the first thing most users do with a new computer?  More likely than not they will change the desktop display image to something they want, to make the computer more theirs.  Customizable skins are available for every device imaginable, from laptops to cell phones and game consoles.  The paper shows that users will customize a device to make it more personal, as long as easy methods to do so exist.  Some of the families talked about ordering more skins but found the process too complicated.  The families without kits, obviously, didn't know that customization was an option, and therefore didn't personalize Roomba.  In fact, after reading this paper, I decided I want a Roomba.  This Roomba.  If you don't think he's cute you don't have a heart.