Human factors of ubiquitous computing : ambient cueing in the digital kitchen?

This thesis is concerned with the uses of Ubiquitous Computing (UbiComp) in everyday domestic environments. The concept of UbiComp promises to shift computing away from the desktop into everyday objects and settings. It has the twin goals of providing ̳transparent‘ technologies where the information has been thoroughly embedded into everyday activities and objects (thus making the computer invisible to the user) and also (and more importantly) of seamless integration of these technologies into the activities of their users. However, this raises the challenge of how best to support interaction with a ̳transparent‘ or ̳invisible‘ technology; if the technology is made visible, it will attract the user's attention to it and away from the task at hand, but if it is hidden, then how can the user cope with malfunctions or other problems in the technology? We approach the design of Human-Computer Interaction in the ubiquitous environment through the use of ambient displays, i.e. the use of subtle cueing, embedded in the environment which is intended to guide human activity. This thesis draws on the concept of stimulus-response compatibility, which is well known in Human Factors, and applies this to the design ambient display. In addressing this concern, a series of studies are conducted to test combinations of LEDs (1, 2 or 3 LEDs) using different colours (red, blue and green), arranged in either compatible or incompatible combinations. We employ these LEDs to the linear and quadrant cooker-control layout. This was evaluated through questionnaires and experimental testing using a prototype. As expected, human performance is better when there is compatibility between the cues, while incompatible arrangements make performance worse and lead to confusion. We also found that there is no effect of ambient cueing when people use a quadrant layout (because the Stimulus-Response Compatibility is clearly defined by the mapping between burners and controls) and additional cueing is not required. However, there was a potential for the cueing to enhance performance in the linear arrangement, particularly when the cues are compatible. Furthermore, when the number of cues increases, performance time reduces, suggesting a benefit of redundancy in cueing. Finally, when incompatible cues were presented, performance became significantly and adversely affected; response times were slower and more errors occurred, compared to compatible arrangements. This suggests that the risk associated with malfunctioning ambient cueing could be higher than the benefits associated with performance improvements. A small cross-cultural of cueing was also conducted. This used the questionnaire and experimental prototype and results showed little difference in preference or performance with the use of cueing. This is in contrast to preference work which had shown cross-cultural differences and suggests that the use of the cueing could produce a more uniform response across cultures. From these studies, it is now suggested that provision of additional information could enhance consistency of response in the control-burner selection task. Ambient cueing was further explored through Ambient Counter, a mock-up kitchen counter consisting of projected cues, cooking space, utensils and ingredients. The use of ambient cues in the Ambient Counter is to guide users through their cooking activities by using projection to highlight the correct ingredients to use in preparing traditional Malay dishes. A comparison of three forms of interaction was conducted; recipe book, ambient interface and smart chalk interface. Experimental evaluation shows that using a pointing device or a recipe book resulted in poorer performance than when the cueing was related to picking up specific ingredients. Thus, integrating the cueing with the actions leads to better performance. We also compared expert and non-expert cooks following the recipes and our analysis shows that both groups benefited from the ambient cueing. Finally, the thesis considered the consequences on human performance when there are redundant cues, ambient display failure or interruption in the use of ambient cueing. This was intended to explore the potential impact of system malfunction on performance. Experiments show that distraction slows performance significantly and that participants get confused. This suggests that participants tend either to rely on their own expert knowledge and ignore any displayed information (as they are familiar with the recipes and ingredients) or tend to follow, believe and trust the information given by the system. The analysis also shows that if ambient displays fail, this leads to significant impairment of user performance. Furthermore, users tend to follow the information given by the system even though that information is wrong. This thesis emphasizes the need to understand the users‘ perspectives and responses in any particular approach that has been proposed. Therefore, the main contributions of this thesis focus on approaches to improve human performance in the ubiquitous environment through ambient display. The proposed methods consider the nature of users‘ behaviours and the limitations of the technologies when they go wrong. Since the effect of the ambient display failure could be catastrophic, the research undertaken in this thesis is important to alleviate the problem that could affect both the user and the technologies.