PhD Defense: The Effects of Varying Levels of Reality Based Interaction Styles on a Subjects Ability to Perform a 3D Construction Task

August 12, 2008
1:30 pm - 3:30 pm
Halligan 127
Speaker: Michael Poor, Tufts University


Abstract The nature and quality of Human Computer Interaction (HCI) continues to be of interest to computer science researchers. While most advancement in this realm has come from practical experience, there had not been a great deal of theoretical or experimental research in advancing basic understanding. A new HCI conceptual framework focused on reality-based interactions has been advanced by Dr. Robert Jacob. The core problem that the reality-based interaction (RBI) framework addresses is the need for users to more easily use an interface. The core proposition of the framework is that the more reality-based interactions an interface incorporates, the easier and quicker a user will be able to use the interface. It was apparent from the review of the literature that there is little scientific evidence that supports or refutes this proposition and the framework.

The purpose of this study was to investigate whether a userís ability to interact with an interface improves when you increase the amount of reality-based interaction in that interface. More specifically, this research examined delivering instructions to a user for a three-dimensional construction task through various reality based interactions.

A review of the experimental design creation will be covered. Outlining the entire development process of all conditions, creation of the conditions, and task selection and refinement. Once the preliminary development is covered, the experiment will be outlined in more detail.

The experimental design of this study employed four groups to obtain the necessary data to test the hypotheses. The groups were tested on various levels of reality-based interactions. Those treatment groups included: (1) a simple graphical user interface (GUI) with a simple slide show of instructions, (2) a three-dimensional representation of a person delivering instructions on a two- dimensional monitor (referred to as the 3D condition), (3) a three- dimensional representation of a person delivering instructions in an augmented reality (AR) environment using a head-mounted display, and (4) an animatronic representation of a human being delivering instructions. Each of these interacted with subjects through gestures, speech, or pictures, but they did not interact with the construction task itself.

For this experiment, performance was measured in terms of seven dependent variables. These dependent variables were divided into two groups. The first group included the number of times a step was repeated, the number of times human assistance was requested, and the total number of errors. The second group included the four types of errors that a subject could perform: incorrect connection, incorrect rotation, incorrect piece selection, and incorrect tilt or angle. The higher the occurrence of these dependent variables, the worse a subject's performance.

As was shown in the statistical results, there was a significant difference between interaction styles in all three of the first group's dependent variables. There were also significant differences between interaction styles in two of the dependent variables from the second group, incorrect piece and incorrect angle errors. Therefore, given these results, we are able to reject the null hypothesis and say that there is a significant difference in performance of a user's ability to complete a construction task delivered by four interaction styles with varying levels of reality- based interaction techniques.