Paper Reading #5: A Framework for Robust and Flexible Handling of Inputs with Uncertainty

• Title:
• A Framework for Robust and Flexible Handling of Inputs with Uncertainty
• Reference Information:
• Julia Schwarz, Scott Hudson, Jennifer Mankoff, and Andrew D. Wilson. 2010. A framework for robust and flexible handling of inputs with uncertainty.  In <em>Proceedings of the 23nd annual ACM symposium on User interface software and technology</em> (UIST '10). ACM, New York, NY, USA,  47-56. DOI=10.1145/1866029.1866039 http://doi.acm.org/10.1145/1866029.1866039
• UIST 2010 New York, New York.
• Author Bios:
• Julia Schwarz is a Ph.D. student at Carnegie Mellon University.  Schwarz had the opportunity to work with Andrew Wilson at Microsoft Research, for this paper.  Schwarz enjoys developing apps for the Windows Phone.
• Scott Hudson is a professor at Carnegie Mellon University, where he was the founding director of the HCII PhD program.  Hudson has published more than 100 technical paper and has served as the Program Chair for events such as UIST.
• Jennifer Mankoff is a professor at the Human Computer Interaction Institute at Carnegie Mellon University.  Mankoff earned her PhD at the Georgia Institute of Technology and researches interactive technologies that empower people.
• Andrew (Andy) Wilson is a senior researcher at Microsoft Research.  Wilson received his Ph.D. at the MIT Media Laboratory and researches new gesture-related input techniques.
• Summary
• Hypothesis:
• The researchers hypothesized that if the uncertainty of a gesture is utilized as part of the decision making process when determining an appropriate action to take, the accuracy of gesture input would increase.
• Methods
• Researchers developed several examples to test their framework of event selection which included interactors and mediators. Their system fed the input to all possible recipients of the gesture, along with information about the gesture, to create a probability representing how likely it was that that action was the intended one by the user.  The system utilizes a lazy evaluation scheme, putting off actual determination as long as possible.  Once a finalization request is made, such as when the user lifts their finger off of the screen, the mediator is invoked to select the appropriate action.  One of their examples was speech text entry, in which a "2" and a "q" can sound the same.  In their program the input would be fed to all possible input boxes which could handle such inputs until either more information is given which distinguishes the information or the user finishes entering text and then must simply select the location the text is supposed to go.  Part of that system is combined with their smart text delivery which allows users to enter text without specifying a text box.  Both used very minimal amount of code to link their test example with their input system.
• Results
• All of their tests resulted in affirmation that their system improved the accuracy of input systems that contained uncertainty.  Particularly, a chart was included in the paper that displayed the percentage of incorrect selection of events from motor-impaired users.  For all four users the conventional system got between 5 and 20% of their intended inputs incorrect.  With the probabilistic input system, less than 3% of the inputs entered were misinterpreted, with two users not receiving any incorrect events.  Researchers believe that with a minimal more amount of time becoming accustomed to the system that the users would all receive near 100% accuracy. 
• Contents
• Rather than dispose of the uncertainty of inputs immediately when the event happens, researchers propose that this uncertainty should be saved throughout the decision making process of how to handle a given input.  This allows for the introduction of probabilities to help determine the most likely recipient of an event.  For example, if there are two buttons immediately next to one another, but one button is inactive, then a touch gesture covering both buttons is likely not meant for the inactive button.  Therefore, in this simple example, the probability that it was for the inactive button becomes a 0 and the probability that it was for the active button (assuming the touch gesture does not interact with any other item) becomes a 1 and that action is selected.  
• Discussion
• The researchers effectively demonstrated their point.  This is an interesting model because it has not been introduced into the computer human interaction world yet, even though the types of allowable inputs for computational devices have been changing for quite some time.  Consider the amount of smart phones out on the market and many various users of them. As such, their interactions may not be as perfect as developers would hope for.  Changing the input handling model is appropriate to handle these imperfect interactions.  A stronger underlying system has the potential to greatly improve the effectiveness and accuracy of gesture input which will help morph computer interaction and the computer experience.
  
  
  

  
  

Picture Source: "A Framework for Robust and Flexible Handling of Inputs with Uncertainty"