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

References:
A Framework for Robust and Flexible Handling of Inputs with Uncertainty by Julia Schwarz, Scott E. Hudson, Jennifer Mankoff, and Andrew D. Wilson.  Published in the UIST 2010 proceedings in New York, New York.
Author Bios:

Julia Schwarz is a PhD student at Carnegie Mellon University studying Human-Computer Interaction.  Her research interests are natural interaction and handling ambiguous input.  Scott Hudson is currently a professor in the HCI Institute at Carnegie Mellon Unversity.  He holds a PhD in Computer Science from the University of Colorado.  Jennifer Mankoff is an Associate Professor in the HCI Institute at Carnegie Mellon University.  She holds a PhD in Computer Science from the Georgia Institute of Technology.  Andy Wilson holds a BA from Cornell University, and a MS and PhD from the MIT Media Laboratory.  He currently works as a senior researcher for Microsoft, focusing on applying sensing techniques to enable new styles of HCI. 
Summary:
Hypothesis
This paper did not have a hypothesis per se, but aimed to present a framework for handling input with uncertainty in a systematic, extensible, and easy to manipulate fashion.
Methods
To illustrate the framework, the authors present six demonstrations including tiny buttons manipulated by touch, a textbox that handles spoken input, a scrollbar that responds to inexact input, and buttons designed to be easier for those with motor impairments.  The first three demonstrations focused on the ambiguity in determining what the user intends to interact with.  This was shown through window resizing, ambiguous and remote sliders, and tiny buttons for touch input.  Actual determination of intent was based on probability and generalized location of the input.  The next two examples focused how text might be entered into a form through either smart text delivery or speech.  Both examples deliver text to a form with multiple fields and divide the process into three phases.  The first phase, selection, occurs as the text boxes return selection scores based on the match between the incoming characters and its own regular expression that determines what sort of input it is looking for.  The second stage, temporary action, shows the text in gray to indicate that it is not yet finalized.  The third and final stage, finalization, occurs if a users clicks on the textbox to explicitly disambiguate which field should be receiving the text or if the text in the textbox matches some finalization criteria in the regular expression.  The sixth and final example, improved GUI pointing for the motor impaired, aimed to treat the input in a way that handles its true uncertainty well and thereby increase accuracy of user interactions.  This was done by gathering data on real-world interaction of motor-impaired participants and then simulating how the clicks would be interpreted with the uncertainty compensation of the system.
Results
The first set of demonstrations shows that it was easy to adapt to the flexibility of the interface.  There is ambiguity in determining what the user intends to interact with, but the system helps handle the process of deciding.  The authors say that their framework has the potential to enable entirely new forms of interaction, which can adjust their responses based on how likely they are to be pressed.  The second set of demonstrations showed that the framework was very capable of handling multiple interpretations of alternative events with little to no extra development.  Finally, the last demonstration proved that the framework was capable of handling the input robustly, missing only two cases out of over 400.
Contents
This paper begins by expressing the idea that the conventional infrastructure for handling input lacks the ability to manage uncertainty, and goes on to propose a framework that solves this problem.   They describe several experiments designed to prove the robustness and usability of the framework, such as how it can handle touch input over tiny buttons and deal with various text entry modes.  Each of the examples shown prove to be successful demonstrations of the capabilities described.
Discussion:
I feel that the paper presented an interesting idea and, by its own contents, a completely successful resulting product.  I completely agree that there is huge room for development in translating imprecise human input into something that a computer can understand.  However, I was disappointed that this paper made no real note of actual user trials, aside from the data gathered for the motor-impaired GUI.  It is impossible to know if, in fact, they did have normal people test any of the software, but I found its absence in the paper disappointing.  I was also a little bit confused as to why they pursued the extended GUI for the motor-impaired.  While I agree that it is definitely something that would have to be considered, it seemed out of place in the general midst of the rest of the demonstrations.