Coimagining the Future of Voice Assistants with Cultural Sensitivity
Table 4
Statistically significant differences in dialogue theme frequencies by scenario.
Theme
Subtheme
df
S1
S2
S3
Social aspects
Chit-chat
163.06
2
<0.001
93 (95%)
82 (88%)
14 (15%)
Interpersonal connections
63.87
2
<0.001
10 (10%)
12 (13%)
51 (56%)
Advice
73.80
2
<0.001
11 (11%)
10 (11%)
54 (59%)
Sounding board
106.44
2
<0.001
4 (4%)
8 (9%)
57 (63%)
Ideological engagement
38.84
2
<0.001
1 (1%)
30 (32%)
10 (11%)
Harassment
9.61
2
0.008
0
1 (1%)
6 (7%)
VA behaviour
Recommending
28.14
2
<0.001
4 (4%)
3 (3%)
22 (24%)
Giving an opinion
18.57
2
<0.001
6 (6%)
25 (27%)
26 (29%)
Thinking ahead
8.12
2
0.02
7 (7%)
7 (8%)
17 (19%)
Humour
6.36
2
0.04
0
0
3 (3%)
VA knowledge
About the user
6.79
2
0.03
2 (2%)
2 (2%)
8 (9%)
About the environment
97.89
2
<0.001
65 (66%)
14 (15%)
5 (5%)
Technically omnipotent
33.00
2
<0.001
40 (41%)
15 (16%)
7 (8%)
User behaviour
Trusting the VA with complex tasks
111.44
2
<0.001
1 (1%)
10 (11%)
57 (63%)
Giving the VA the lead
8.45
2
0.01
33 (34%)
36 (39%)
49 (54%)
Implicit next
6.40
2
0.04
32 (33%)
41 (44%)
46 (51%)
Discourse features
Gendered language
10.81
2
0.005
1 (1%)
1 (1%)
8 (15%)
Sig. diff. at the level. Sig. diff. at the level. Sig. diff. at the level. S1: scenario 1, about the weather; S2: scenario 2, about the news; S3: scenario 3, asking for advice.