27.03.2014 Views

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

In the following, we present the results of the<br />

experiments run using the tools that we developed.<br />

II. EXPERIMENTS<br />

A.Pilot test using the EasyTheatre tool<br />

First we ran a pilot test using the EasyTheatre<br />

application [26]. EasyTheatre is an e-commerce<br />

application for booking theatre tickets. One of the key<br />

characteristics of the application is that it includes a<br />

usability dashboard. Users can use the dashboard to enable<br />

and disable functions corresponding to specific usability<br />

mechanisms.<br />

Apart from correctly capturing data, the goal of this<br />

pilot study was to establish the soundness of the<br />

application, as well as the type of data to be collected and<br />

the reliability of the experimentation process.<br />

We ran two experiments in which we tried to measure<br />

the impact of global undo, help, system status feedback<br />

and warning mechanisms on the efficiency and satisfaction<br />

attributes.<br />

The experiments were run with 2011 Universidad<br />

Politécnica de Madrid Master in Software Engineering<br />

students and 2010 Universidad Simón Bolívar Master in<br />

<strong>Systems</strong> Engineering students. A total of 24 subjects<br />

participated in the experiment. They were divided into two<br />

groups: a control group (CG) that did not use usability<br />

mechanisms and a test group (TG) that used built-in<br />

usability mechanisms. Users were given use scenarios and<br />

administered questionnaires used to collect the data in<br />

writing. Users had to use the application to execute the<br />

respective scenarios and then answer the survey questions<br />

about the execution. The only difference between the<br />

scenarios given to students belonging to the control and<br />

test groups was that the control group members were not<br />

asked to enable the mechanisms to be measured (global<br />

undo, help, system status feedback and warning<br />

mechanisms), whereas the test group members were. The<br />

questionnaires were the same for both groups. Divided into<br />

quantitative data: task time in minutes reported by users in<br />

writing and qualitative data: responses on a 1-to-5 Likert<br />

scale for each mechanism graded from Never to Always<br />

[10][24][27].<br />

Note that we use the difference (increase) in the data for<br />

the test group compared with the control group in order to<br />

measure the impact.<br />

Data analysis<br />

The datasets were analyzed depending on the type of<br />

data they contained. The quantitative data were analyzed<br />

using the Microsoft Excel . and SPSS 16 (Statistical<br />

Package for the Social Sciences) [28] software tools,<br />

whereas the qualitative data were analyzed using SPSS 16<br />

only. SPSS was used to perform the non-parametric Mann-<br />

Whitney U test [29] to check whether the differences<br />

between the variables are statistically significant, at 95%<br />

confidence level.<br />

The aim of the experiment was to determine whether<br />

the mechanisms have an impact on the already mentioned<br />

usability attributes. The Mann-Whitney U test determines<br />

whether there is a statistically significant difference<br />

between the data reported by the CG and the TG. Not until<br />

this statistical significance has been determined can we<br />

infer the effect of each mechanism on the usability<br />

attributes.<br />

Quantitative data analysis<br />

Table I lists the mean times used by the control and the<br />

test group for each mechanism.<br />

The results of the Mann-Whitney U test applied to the<br />

gathered quantitative data show that the difference between<br />

the mean times to complete a task in the test and control<br />

groups was not statistically significant in any case. The<br />

reasons for this could be any, all or a combination of the<br />

following:<br />

• Homogeneous group: all users were studying for a<br />

computing-related master’s degree and had a<br />

computing-related background.<br />

• There were fewer than 20 subjects per group.<br />

• Users reported time manually in minutes.<br />

Although the Mann-Whitney U test results are<br />

inconclusive, note that, in all cases except for the global<br />

undo mechanism, there was a positive difference between<br />

the means of the time indicator in minutes in the test and<br />

control groups.<br />

There is an explanation for result with respect to the<br />

global undo mechanism, namely, users cannot undo actions<br />

if the mechanism is not enabled. Control group users<br />

cannot find a button to undo the action, so they<br />

immediately give up. Accordingly, when we analyzed user<br />

responses to the control group survey associated with this<br />

mechanism, we found that all the users of this group<br />

responded that it is never possible to undo actions.<br />

TABLE I.<br />

MEAN TEST DURATION (MINUTES)<br />

MECHANISM CG TG Difference Increase (%)<br />

Global Undo 3.36 3.88 -0.52 -15,4<br />

Help 4.0 2.91 1.09 27,25<br />

Warning 2.41 1.7 0.71 29,46<br />

System Status Feedback 4.3 2.3 2.0 46,51<br />

600

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!