Measuring social acceptability of mobile voice user interfaces

Sammanfattning: Even though mobile voice user interfaces have been available for several years, their use in the public space is still low. Voice user interfaces (VUI) introduce a new and highly visible behaviour, which raises questions if or where their use is socially acceptable. The social acceptability of mobile voice user interfaces and interactions has not been defined before, and there is a lack of knowledge about how to measure and evaluate it. This thesis aims to increase the understanding of the social acceptability and social acceptance of mobile voice user interfaces, to gather more knowledge on how the social acceptability of mobile VUIs can be measured, and to reflect on the suitability of evaluation methods that could be used for such measurement. After a literature review, we defined and explained essential aspects of the social acceptability of VUIs. We followed four out of eight steps of DeVellis’s scale development methodology and suggested a multi-item scale draft for the measurement of social acceptability of voice user interfaces. We also presented a summary of evaluation methods used for the measurement of social acceptability, and we designed our own low-fi pictogram video prototype. A version of our multi-item measurement scale (draft V3) was evaluated on the example of Siri voice assistant with two evaluation methods (in situ and with pictogram videos). Evaluations resulted in recommendations for the removal or revision of the existing scale items and the inclusion of new items. Partly controlled in-situ evaluation is our preferred method for evaluating the social acceptability of VUIs. Some degree of control of the setting (location, time of the day) would likely improve the consistency of the surrounding and ensure better condition for the analysis of the data. Our pictogram style video prototype was good in preventing bias and creating a situation where participants would imagine reactions of others instead of observing actors’ reactions. Evaluation based on a video prototype suffered from several issues: lower empathy, lower clarity and differentiation between statements, underestimation of comfort, misjudgement of the effect of the audience, and overall – incorrectly imagined reactions. We conclude that the pictogram video method did not give a close estimation of the real situation.