Social Evaluation of Artificial Agents by Language Varieties

Effects of Language Variety on Personality Perception in Embodied Conversational Agents

“…The videos shown to the participants contain variations in non-verbal behaviour and language variety. The hypotheses are theoretically grounded in earlier findings: (i) East Upper German is perceived by German speakers as extrovert as opposed to Standard German which is perceived as introvert (Gärtig et al 2010); (ii) Viennese dialect is evaluated as more extrovert as Austrian standard (Krenn et al 2012); and (iii) gestural expressivity increases the perception of extroversion (Neff et al 2010).…”

“…1017 participants rated the "typical German", 501 participants rated the "typical Bavarian". Krenn et al (2012) show that language attitudes towards synthesized voices representing local varieties are comparable to language attitudes towards natural voices. 91 subjects of (Austrian) German mother tongue were presented with a semantic differential comprising 19 adjective pairs representing positive and negative extremes on a value dimension, where each pair had to be rated on a 5-point Likert scale.…”

“…Results from a major, recent language attitude study on German varieties (natural voices), for instance, indicate that speakers of Bavarian are perceived as more extrovert (high in spirits) than speakers of the German standard (Gärting et al 2010). In a study on synthetic standard Austrian German and Viennese dialectal varieties, (Krenn et al 2012) demonstrated that: (i) language attitudes towards natural voices transfer to synthesized voices, and (ii) the dialectal Viennese variety is characterized by attributes that also characterize extrovert behaviour. For instance: the dialectal voice compared to the standard Austrian voice is associated with sense of humour, emotionality, self-confidence, open-mindedness and an easy-going nature.…”

Two Techniques for Assessing Virtual Agent Personality

“…There are also cultural considerations that are outside the scope of this current study that should be investigated in future work. Other work shows that the relatively new ability of TTS engines to produce voices with dialectical variations leads to stereotypical attributions to agents with those voices: for example, an agent that speaks with a regional accent has a sense of humor, whereas an agent that speaks with Received Pronunciation is more educated and trustworthy [11].…”

“…Increasingly sophisticated virtual agents are now sometimes designed to mimic these cues in order to induce human observers to make similar attributions about the agent with whom they are interacting. The generation of consistent, human-like patterns of expression based on prescribed personality traits makes interaction with these agents easier for users [5,6] and similarly allows for the beneficial tailoring of agents to specific users or to specific task domains [7][8][9][10][11][12][13][14]. Furthermore, there is evidence that suggests that these cues are most effective when cues are congruent, rather than conflicting [12,15] .…”