30.06.2013 Views

Texte intégral / Full text (pdf, 20 MiB) - Infoscience - EPFL

Texte intégral / Full text (pdf, 20 MiB) - Infoscience - EPFL

Texte intégral / Full text (pdf, 20 MiB) - Infoscience - EPFL

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.

7.4. Results<br />

1-2 1-3 1-4 2-3 2-4 3-4<br />

Normality 0.6888 0.579 0.3217 0.8711 0.5527 0.6746<br />

Gaze 0.0001 0.1277 0.0000 0.0126 0.0005 0.0000<br />

Immersion 0.1504 0.5041 0.8614 0.4804 0.7641 0.3252<br />

Anxiety 0.2783 0.6418 0.9561 0.5475 0.3360 0.7010<br />

Friendliness 0.7421 0.8741 0.0122 0.8867 0.0551 0.0456<br />

Environment 0.8470 0.6540 0.6651 0.7965 0.5343 0.3868<br />

Characters 0.1073 0.1222 0.4345 0.9548 0.4963 0.4890<br />

User 0.0179 0.2510 0.0000 0.2185 0.0012 0.0000<br />

Table 7.2: Results to Student’s T-test. 1 is the user-centered mode, 2 is the interest-centered<br />

mode, 3 is the user or interest centered mode and 4 is the random mode.<br />

However, it has to be noted that all participants did not react in the same way towards the<br />

experiment. Some participants focused on other things than expected, such as character size<br />

and the fact that there was no collision detection between them and the user. On the other<br />

hand, other participants set these factors aside and focused on character gaze attention behaviors.<br />

For example, the correlation between perceived immersion and perceived awareness<br />

of characters towards the user increases to 66% if we take out those who complained of lack<br />

of collision avoidance. We thus believe that this factor can greatly account for low scores in<br />

normality, awareness and immersion. This value was above 75% for 10 of the participants<br />

and above 90% for 6 of them.<br />

Regarding the most normal, preferred mode, 14 participants out of the 30 opted for the<br />

user or interest one. This is almost half of the population we have tested. Out of the 16<br />

remaining, 11 chose the user-centered one. A little more than a third of our tested cohort<br />

were therefore possibly biased towards an always attentive population of virtual characters.<br />

Finally, 4 participants chose the interest-centered mode and 1 chose the random one. When<br />

asked why, almost all the participants having chosen the user or interest mode, answered that<br />

it seemed more natural to them because the characters were looking at them but not staring<br />

at them. The characters seemed to be interested in them and aware of them but also of other<br />

characters in the scene. There are thus really two major trends. On one hand, we found a<br />

group of people for whom it seemed more natural of being constantly looked at by the virtual<br />

characters. On the other hand, almost half the tested cohort had a preference for the random<br />

mode in which they were being looked at but not all the time and not by all characters.<br />

It is interesting to note that the correlations are different if we isolate the preferred mode.<br />

For example, the correlation between anxiety and the perceived gaze of characters was 58%<br />

for those who preferred the user-centered mode. On the other hand, it was 29% when considering<br />

those who opted for the user or interest mode. It is quite interesting to note that<br />

for those users who found always attentive characters as most natural, the correlation with<br />

anxiety was almost twice as high as for those who preferred characters who looked at them<br />

but not always. The same tendency was found between perceived normality and awareness<br />

of characters towards the participants, and also between perceived friendliness and perceived<br />

awareness towards the user. The correlations in all these cases are practically twice as high<br />

in the participants who preferred the user-centered mode.<br />

115

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

Saved successfully!

Ooh no, something went wrong!