Research Synthesis Design Solution

Interaction 03


03 Interaction

Creating an interaction based upon mood was a difficult challenge, since people describe their moods in diverse and poorly-defined terms. Our focus group insisted that moods defy exact specification and that attempting to define a mood based on multiple components would be confusing. We turned to psychological research to find a means of describing mood that would be simple enough to perform by direct manipulation of a tactile interface and also rich enough to be meaningful for users. Our research uncovered the "affect grid" concept, shown below. The energy level of a mood is mapped to the vertical dimension of a two-dimensional grid, while the positive/negative aspect of a mood is mapped to the horizontal dimension. The use of a grid combined well with the physical form concepts from our toy inspirations, and the two aspects are simple enough that we believe they can be determined automatically by analyzing the content of songs algorithmically.

Russell, J. A., Weiss, A., & Mendelsohn, G. A. (1989). Affect grid: A single-item scale of pleasure and arousal. Journal of personality and social psychology, 57, 493-502.

In addition to the automatic rating of songs by computational analysis, we wanted to provide a means for users to provide feedback to the system about their personal impressions of songs. For example, a song which ordinarily would be rated on the negative side of the scale might have associations with a satisfying personal experience, or vice versa. We also wanted to show the user how the system rates an individual song while it is playing. We can achieve both goals by lighting up the grid with a special color in the location corresponding to a song's rating. A user can adjust this rating by touching the indicated area and then touching another location. The song's rating would be updated and the light would move to the new location.

In order to specify a mood, a user presses their hand against a pressure-sensitive surface consisting of many small, moveable rods. As the user creates an impression in the surface, LEDs at the ends of the rods increase in brightness to provide additional feedback. The current state of the system can be determined from a distance based on the different intensity levels of the LED lights at different positions on the grid, where brighter light indicates a greater weighting of that grid area in selecting new songs.

Another feature we wanted to include was the concept of playlists. Our focus group participants all used playlists to create collections of music that represented a specific mood or atmosphere. Existing playlists therefore could be used to seed our system with a set of examples for selecting new songs. We decided that playlist creation would not be a focus of our system, however. We plan to integrate the müz with existing software, such as Apple's iTunes, which can provide richer support for searching and browsing large music collections. In order to accommodate the desire for both playlists containing specific songs and non-specific mood selections, we decided to use graphical representations of music collections that will maintain the mental model of the grid, rather than using textual names or descriptions. Both playlists and mood selections can be represented by the areas of the grid that will be active when songs are being played from them. A small version of the grid will be displayed inside each of a row of buttons on the device. Pressing a button will cause the configuration displayed within it to be transfered to the main grid. A mood selection can be saved by pressing and holding on a button for several seconds, until its display is updated to reflect the grid. Songs can be added to a playlist by pressing the grid where it is lit and then pressing a playlist button. The mood of songs within a playlist can be used to select new songs by pressing a "remix" button.

We also provide basic volume and skip controls via the directional pad at the right of the control panel. Users have a relatively consistent pre-conception that volume control should map to the vertical buttons and skip controls should map to the horizontal buttons. We don't believe that textual labels will be necessary. The central button performs the one remaining common playback function: start/stop. These controls are duplicated on the guest controller, pictured below.

A final consideration for our system was whether to include a screen to display information about the currently playing song. Our design accommodates such a screen, but we leave the ultimate decision for future exploration. Our research indicates that a screen would be unnecessary for the owner of the system, since they typically know enough about their own music collection to find reminders unnecessary. Including a screen might detract from the clean aesthetic of the main grid surface. However, it is useful to display the title and artist of songs for the benefit of guests, who may want to know more about the music they are hearing. We recommend that if a display is included, it should be relatively small and use a monochrome, backlit screen that is consistent in color with the rest of the system.


IID 2005 . Human-Computer Interaction Institute . Carnegie Mellon University