ch 4 revisions

karulf-thesis/thesis-results.tex

@@ -5,8 +5,7 @@ \section{Results}
 
 \subsection{Effects of Prior Experience} % (fold)
 \label{sub:effects_of_prior_experience}
-In general, we found that prior experience with video games or with controlling a robot affected a subject’s performance on the search task.
-The single factor that showed the greatest influence on performance was video game use (Table~\ref{tab:prior-vg}). Subjects who did not regularly play video games took almost twice as long, on average, to complete the search task than subjects who did regularly play video games. An even more marked difference was seen in total neglect and idle times. Non-gamers had, on average, approximately three times longer neglect and idle times, compared to regular gamers.
+In general, we found that prior experience with video games or with controlling a robot affected a subject’s performance on the search task. The single factor that showed the greatest influence on performance was video game use (Table~\ref{tab:prior-vg}). Subjects who did not regularly play video games took almost twice as long, on average, to complete the search task than subjects who did regularly play video games. An even more marked difference was seen in total neglect and idle times. Non-gamers had, on average, approximately three times longer neglect and idle times, compared to regular gamers.
 
 
 \begin{table}[ht]
@@ -71,7 +70,7 @@ \subsection{Use of Supervisory Mode} % (fold)
 \label{sub:use_of_supervisory_mode}
 
 Subjects spent, on average, more than 69\% of their time in supervisory mode $(t = 2.4944, p < 0.01)$, and less than 0.3\% of their time in first-person mode $(t = -2.747, p < 0.01)$.
-Completion time, total neglect time, and total idle time were significantly affected by the percentage of total time spend in supervisory mode. The median time percentage of time spent by subjects in supervisory was 94.19\%. Subjects that spent more than this median percentage of time in supervisory mode completed the search task, on average, twice as fast as subjects who spent less than median time in supervisory mode (Table~\ref{tab:supervisory}). The effects on total neglect time and total idle time were similar.
+Completion time, total neglect time, and total idle time were significantly affected by the percentage of total time spent in supervisory mode. The median time percentage of time spent by subjects in supervisory was 94.19\%. Subjects that spent more than this median percentage of time in supervisory mode completed the search task, on average, twice as fast as subjects who spent less than median time in supervisory mode (Table~\ref{tab:supervisory}). The effects on total neglect time and total idle time were similar.
 
 
 \begin{table}[ht]

karulf-thesis/thesis-userstudy.tex

@@ -1,21 +1,21 @@
 %!TEX root = karulf-thesis.tex
 \chapter{User Study}
-We conducted a formal user study to evaluate the effectiveness of the RIDE interface and test the validity of our thesis. We designed an hour-long user study with the goal of testing the asynchronous task and notification system.
+We conducted a formal user study to evaluate the effectiveness of the RIDE interface and to test the validity of our thesis, specifically the notification system. We designed an hour-long user study with the goal of testing the asynchronous task and notification system. Our experiment had two-conditions; in one condition the notification system was displayed, and in the other condition, the notification systems were not enabled. An example notification is shown in Figure~\ref{fig:ride-notification}.
 
-We conducted a two-condition experiment, one with notification displayed and one with notifications hidden. An example notification is shown in Figure~\ref{fig:ride-notification}. 
+The user study was advertised via flyers posted publicly on the Washington University Campus. Our user pool consisted of 22 adult participants, 6 female and 16 male ($M=23.5, SD=5.45$). The median age was 21.
 
-The user study was advertised via flyers posted publicly on the Washington University Campus. Our user pool consisted of 22 adult participants, 6 female and 16 male ($M=23.5, SD=5.45$).
+Using two simulated robots in a small house, subjects were asked to perform a search task. Subjects were directed to locate three boxes in the house, as quickly as possible. The subjects were told the boxes were not shown on the map and that the boxes would be detected by the laser range-finder sensor. To confirm that the subjects had found the box, they were asked to point it out to the experimenter, once the box had been found. Subjects were informed they could use any of the features of the interface they wanted. In all runs of the experiment, the starting positions of the robots, and box locations, were the same. The simulated robots modeled Videre Design Erratic ERAs with a Hokuyo URG laser range finder. 
 
-We ran each participant individually through the user study. A user study session consisted of a pre-experiment questionnaire, a short set of practice sessions, two search and rescue experiments, and a post-experiment questionnaire. We created two separate experiment conditions, one without notifications and one with notifications. We developed a program to execute both conditions in a random order during the user study.
+Each subject was individually run through the user study. A user study session consisted of a pre-experiment questionnaire, a short set of practice sessions, two search and rescue experiments, and a post-experiment questionnaire. We created two separate experiment conditions, one without notifications and one with notifications, and created a program to execute both in a random order during the user study.
 
 \section{Experimental Design}
-The primary goal of our user studies was to determine if the RIDE user interface was an effective tool for control groups of robots. Our secondary goal was to test the individual interface elements unique to RIDE to determine if they enhanced the user experience. We looked through prior work to find activities that a group of robots could accomplish more effectively if controlled properly and would not require background experience. We decided upon ``search and rescue'' as it is a generally understood activity that can be accomplished faster through effective coordination of robots. 
+The primary goal of our user studies was to determine if the RIDE user interface was an effective tool for control groups of robots. Our secondary goal was to test the individual interface elements unique to RIDE and to determine if they enhanced the user experience. We examined prior work to find activities that a group of robots could effectively accomplish, if controlled properly, and would not require additional background experience of the subjects. We decided upon the ``search and rescue'' modality, as, generally, it is an easily understood activity that can be accomplished quickly, through effective coordination of robots.
 
-We designed a scenario where the robots were used to find boxes hidden throughout a house. The boxes would not be visible on the on-screen map directly, but they would display through the sensors on the robots. The house can be seen, without boxes, in Figure~\ref{fig:test-environment}. 
+As discussed, we designed a scenario where the robots were used to find boxes hidden throughout a house. The boxes would not be visible on the on-screen map directly, but they would display through the sensors on the robots. The house can be viewed, without boxes, in Figure~\ref{fig:test-environment}. 
 
-The robots were configured to simulate Erratic ERA-MOBI robots. The simulation engine, as described in Section~\ref{sec:ride_code}, was programmed to emulate a laser sensor and an odometry sensor. These two sensors allowed the robot to detect the boxes, walls, and other obstacles through the laser readings or by picking up a stalled engine through the odometry sensor.
+The robots were configured to simulate Erratic ERA-MOBI robots. The simulation engine, as described in Section~\ref{sec:ride_code}, was programmed to emulate a laser sensor and an odometry sensor. These two sensors allowed the robot to detect the boxes, walls, and other obstacles, through the laser readings or by picking up a stalled engine through the odometry sensor.
 
-We designed three separate runs for a user study session: a training run, a run with notifications disabled, and a run with notifications enabled. We decided to assume that subjects would have a small amount of training before they were allowed to control robots. In order to meet this assumption, we provided each subject with the same practice run to ensure a common training set.
+We designed three separate runs for a user study session: a training run, a run with notifications disabled, and a run with notifications enabled. We elected to assume that subjects would require a small amount of training before they were allowed to control robots. In order to meet this assumption, we provided each subject with the same five-minute practice run to ensure a common training set.
 
 In the following subsections I will describe, in order, the components of a user study session. This research protocol was approved by the Washington University Institutional Review Board (IRB). The full documentation submitted to the IRB is documented in Appendix~\ref{app:IRB}.
 
@@ -30,7 +30,7 @@ \subsection{Pre-Experiment} % (fold)
 
 \subsection{Training Run} % (fold)
 \label{sub:training_run}
-The training run takes place inside the virtual house pictured in Figure~\ref{fig:test-environment}. Unlike the experimental runs, there are no boxes present within the house during the training run. Instead, the subject is asked to demonstrate basic control over the interface by accomplishing a series of tasks. Once the user completed the training activities, she was given an opportunity to continue using the user interface until the user was comfortable proceeding with the experiments.
+The training run took place inside the virtual house pictured in Figure~\ref{fig:test-environment}. Unlike the experimental runs, there were no boxes present within the house during the training run. Instead, the subject was asked to demonstrate basic control over the interface by accomplishing a series of tasks. Once the user completed the training activities, he/she was given an opportunity to continue using the user interface until the user was comfortable proceeding with the experiments.
 % subsection training_run (end)
 
 \subsection{RIDE UI Experiments} % (fold)