Culturally Determined Preferences: Automatic Speech Recognition (ASR) Systems vs. Live Help

Introduction

One of the critical success criteria of any technology is its effective use and adoption by large sections of the user population across all cultures. Thus, any systematic cross cultural evaluation of Human-Computer Interaction (HCI) technology is always significant because each culture or environment poses a unique set of challenges that must be dealt with in order for a particular technology to gain a foothold and be successfully used. Subtle properties or changes in the design can have major implications in the culture-context in which the technology is deployed. For example, Russo and Boor (1993) discuss many cases of cross-cultural blunders in product launch, two of which will suffice: When the British manufacturer Rolls Royce introduced the Silver Mist in Germany the adverts proved disastrous as the word ‘mist’ in German means manure. Also, when the Italian car maker Fiat introduced Uno in Finland it proved equally embarrassing as the word ‘Uno’ in Finnish translates as garbage. Thus, mere language translation without proper understanding of the culture-context can prove costly for cross-cultural usability. In general, language is regarded as a major vehicle of culture (Marcus and Gould, 2000; Kaplan, 1966; Kluckholn, 1950; Triandis et al, 1988; Ting Toomey, 1999). Therefore, Automatic Speech Recognition (ASR) systems whereby humans use language to interact with computers offer an excellent opportunity for studying cross cultural issues. Increasingly, human interaction with computer systems involving ASR is becoming globally pervasive. Even in remote parts of the developing world such as India and Africa, the phenomenon of humans interacting with computer systems using speech is gaining in popularity. This has been attributed to the penetration of the ubiquitous mobile phone which allows various non-governmental organizations (NGOs) and technology service providers to offer telephone-based automated services and information to the populace (e.g., check mobile phone minutes used, check rolling electricity blackout times, etc.).

However, while the promise of using our own voice to interact with computer systems is being realized, the impact of culture on such systems is something that is generally acknowledged but not systematically studied (Nielson, 1990; Chakraborty et al, 2008; Stewart & Chakraborty, 2008; Stewart et al, 2009). The study of cross-cultural issues in technology becomes more important especially in light of the fact that ASR systems are fraught with many usability problems including speech recognition errors, the cognitive burden on users having to quickly respond to the system (which is exacerbated in the event of an error), constraints on what the system can understand, etc., all of which increase the difficulty of using such systems, and underscores the need to investigate how different cultures perceive their use and/or usefulness in the context of HCI. Several theoretical models for the study of cross cultural variables in communication abound (Aykin, 2005; Chakraborty et al, 2008; Chakraborty, 2009; Hofstede, 1991; Marcus & Gould, 2000; Nielson, 1990; Yeo, 1996). Unfortunately, there is very little empirical data to validate any of the models in ASR. This is the issue we address in this chapter by focusing on the broad and foundational question of whether a framework for cross-cultural research like the one proposed by Hofstede (1991) can be applied directly to ASR. More specifically, in order to investigate culturally determined user preferences for the use of ASR systems, we examine the applicability of Hofstede’s proposed questionnaire based on one major cultural variable: individualism versus collectivism. We address the following related questions: (a) Does individualism affect or influence how people interact with ASR systems? (b) Does collectivism in any way or form affect or influence how people interact with ASR systems?