Giving Computers Personality? Personality in Computers is in the Eye of the User - Coverbal Synchrony in Human-Machine Interaction - page 55

In summary: How the users approached the system differed signifi cantly.
Most subjects avoided any personal pronouns when addressing the
system, some employed the German 'Sie' (formal German version of
'you') and only very seldom was the informal German 'du' used.
Politeness particles
From the total of N = 130 subjects N 1 = 67 used one of the politeness
particles 'bitte' or 'danke' at least once within the packing/unpacking
phase. The maximum number of uses is 34, with a mean of 7.57, std
of 7.89, and median of 4.0. If we do neglect those subjects at and
below the median as only occasional users of these particles, we do
obtain N 2 = 32 subjects that used these particles much more frequently.
Intersecting the group of subjects with at least one occurrence of 'Sie'
(cf. above) with the users of the politeness particles 'bitte' or 'danke,'
resulted in a subgroup of N 3 = 19 subjects. Their combined numbers
of occurrences were ranging from 2 to 32 with mean: 8.60, std: 8.38,
and median: 4.00. In other words, most users of 'Sie' were also users
of the politeness particles.
5.3 Discussion and remarks
Linguistic structures
The use of elliptical structures is a typical aspect of efficient
communication in naturally occurring dialogs (Jurafsky and Martin,
2008). Thus, the dominance of elliptical structures within the users'
contributions in the LAST MINUTE corpus could be seen as a clear
indication that most subjects had experienced their dialogs with
the system in a way that licensed their natural dialog behavior. The
empirical analysis of the structures regarding user utterances was
fed into the implementation of an experimental system as well, thus
allowing to replace the wizard with an automated system based on
the Nuance commercial speech recognizer.
Rapport
Given the figures about the users' employing specific verbs used by the
system, we had to conclude that lexical rapport—i.e. users mirroring
the lexical items of the system—was the exception rather than the
rule. Nevertheless it would seem worthwhile to explore whether—and
how—the subgroup of subjects that did so, is differing from those
subjects that did not. This finely grained analysis is now on the agenda.