RSVP Interaction Methods Study
An Evaluation of Interaction Methods for Controlling RSVP Displays in Visual Search Tasks by Jamie Waese, Wolfgang Stuerzlinger and Nicholas Provart.
- This video is from an earlier version of the paper, before we ran the second study.
Identifying genes of interest can be a daunting and time-consuming task. To be successful, researchers must find subtle differences between nearly identical electronic fluorescent pictographs (eFP images). A display technique called Rapid Serial Visual Presentation (RSVP) could improve the experience by exploiting our ability to recognize differences between images when they are flashed on a screen in a rapid and serial manner.
RSVP is best known for reading applications such as Spritz.
It has not traditionally been used for search applications, but one of my committee members, Dr. Wolfgang Stuerzlinger, suggested an RSVP feature in ePlant might help users to identify genes of interest by rapidly scanning through a series of eFP images as if they were flipping through a book.
We implemented this feature in the form of a “slide show” with a stop/start button and it was received positively during the 2014 user testing in Vancouver. However, few people had actual experience using RSVP for visual search tasks and there was some question as to whether it provides any measurable benefit. Also, some people found the “slide show” approach difficult to use because it is hard to press stop as soon as a target image has been found.
As part of a group project for my Human Computer Interaction class, we compared three different RSVP interface methods for selecting genes of interest against a traditional point and click interface. We ran a study with 81 student participants randomly assigned to one of four groups, with each group tasked with answering the same ten questions. Each group was given a different user interface to control the speed and order in which electronic fluorescent pictograph (eFP) images are displayed on the screen. The figure below shows the three methods we tested: “Slide Show”, “Hover” and “Velocity Control”. A fourth group, our control group, was given a non-RSVP interface that required them to point and click on the button associated with the image they want to see.
We hypothesized that all three methods of RSVP interfaces will result in faster task completion times than a point-and-click (non-RSVP) interface, and that faster completion times will not compromise accuracy. Participants were given ten guided visual search tasks with the instruction: “Select the image on the following page that has the most red in the area that is highlighted below.” Participants had to select one eFP image from a set of twelve using the user interface they were assigned to. Accuracy and time were measured for each question.
We found the “Hover” method to be fastest, and that “Slide Show” and “Velocity Control” are no better than “Point and Click” (F3, 70 = 4.44, p = 0.0065, β = 0.999).
We did not find significant differences between the groups with regards to accuracy. This supports our hypothesis that accuracy is not compromised when using RSVP for visual search (F3, 76=0.4468, ns, β = 0.705).
These findings justify our inclusion of an RSVP display feature into ePlant, and they informed our decisions on the interface design of the most recent version of the tool. We have implemented the “Hover” feature as the default RSVP interface method but also provide three speeds of “Slide Show” for users who still appreciate this feature.
The project is more fully described in the manuscript “An Evaluation of Interaction Methods for Controlling RSVP Displays in Visual Search Tasks” submitted VIS 2017. We are waiting to hear it will be accepted.