Evolving Needs in Iot control and Accountability

1. INTRODUCTION

With the increasing proliferation of Internet of Things (IoT) technologies for consumers, the ‘smart home’ is emerging as a domain for significant potential market growth [103]. Enabled by energy-efficient networking technology [73], such as ZigBee or ZWave and reduced hardware costs, new smart home sensors, devices and services continue to emerge, thus constantly increasing the variety of technologies that might be integrated into today’s smart home. To be adaptable to the individual physical and social character of specific homes [40], many smart home systems have adopted a do-it-yourself (DIY) paradigm. The smart home’s inhabitants can set up, configure, and automate system behavior themselves in accordance with their individual (or family) needs, e.g., having lights switch on in case of detected movement in a hallway, or remotely controlling heating and air conditioning before people arrive home.
However, flexibility in choosing hardware and software setups comes with increased complexity. This complexity, we argue, will magnify as systems develop. The DIY smart home is arguably one of the first IoT- systems that also relies on configuration and management by non-programming users. This non-programming user population cannot be expected to be tech-savvy, or even interested in technology [13,30]. In this vein, Brush et al. have identified poor manageability as a key barrier to the successful adoption of smart homes [13]. Users need clear and unequivocal methods for understanding what is going on with their systems and for making decisions about how they might best fulfil their needs. Such methods, for the reasons we have outlined, have to have a relevance to various user needs and need to service the provision of useful data monitoring devices.
In this paper, we therefore argue for adopting an end user development approach as a means of overcoming some of these challenges. End user development (EUD) has emerged as a way of enabling non-programming users to handle complex systems. Due to its emphasis on self-configuration, EUD is particularly applicable for the DIY ‘smart home’ [48]. We will discuss how intuitive and simplified user interfaces can be constructed so that they make much of the opacity of current smart home systems more transparent. Moreover, the flexibility of the approach provides for more individual management and, associated with this, greater possibilities in respect of reconfiguration over time. Some progress has been made in this regard. Trigger-action programming [94,95], for instance, and rule-based systems have both been studied extensively [12,48,52]. In addition, flexible visualizations as a means of providing users with individually tailored feedback about what is being sensed in the smart home have been considered [7,63]. In a long-term living lab study to understand the demand and potential for EUD in smart homes, Castelli et al. [16] found that, aside from understanding what is going on in the home, participants also wanted feedback on the system’s status. Their findings underscore the presence of the research gap addressed in this paper. For some of the more prominent use cases regarding smart homes (such as can often be seen in advertising), there are already very active research communities. Thus, designing energy consumption feedback has received a lot of attention [1,22,37,79], including from an EUD perspective [47]. Designing for ambient assisted living has also generated a lot of interest [18,43,49].
Conveying system status, however, is not typically oriented towards informing a user about her/his performance, but rather towards making a system’s behavior more transparent. Such system awareness has been shown to be an important feature in end-user configuration, especially for embedded systems [6,64]. In an argument that echoes Dourish’s discussion of system accountability [27], Lim et al., have also shown that limited system awareness may hamper user acceptance of context-aware systems [58]. Another user acceptance barrier is that people do not trust that smart home technology will be safe, secure, and privacy-preserving [17]. In this paper, we argue that system awareness helps users to build an understanding of data disclosure by helping them to learn about a systems’ behavior and operations. Abu-Salma et al. [2], for instance, have shown that a lack of user understanding, compatibility issues, and lack of motivation rooted in a lack of understanding together

contribute to a reluctance to use secure systems. Ruoti et al. have explored the reasons for this reluctance and have subsequently shown [76] how various trade-offs determine user policy.

However, so far, little research has focused on investigating smart home system awareness and how it evolves from a user’s point of view. Woo et al. [97] have investigated ways of fostering understanding of rule hierarchies, suggesting ambient feedback as a possible solution. Yang and Newman have demonstrated how users demand Nest thermostats to provide incidental intelligibility in terms of their behavior and options [101]. The problem of exploring options and possible commands is particularly likely to increase with the onset of voice assistants. Mennicken et al. [60] have provided users in professionally maintained smart homes with system feedback embedded in a calendar. Despite these endeavors, supporting users in making sense of current and past system states remains an under-investigated topic. Moreover, with the notable exception of Oulasvirta et al., who investigated surveillance by smart devices in the home [66], there is a gap in research relating to the longer term appropriation of (not only smart home) technology [62]. Evolving practices, changing levels of expertise and establishing routines around the use of smart home systems over time [69], all make designing for system awareness in DIY smart homes an ongoing challenge. The Living Lab approach we used constitutes one possible way of dealing with this lacuna.
Using a living lab approach [34,36], we investigated evolving user demands regarding smart home system feedback for supporting system maintenance over time. Through a process of iteratively co-designing custom visualizations, we found that information demands and practices shifted significantly once the smart home system became embedded into the participants’ everyday lives. By describing and analyzing these shifts, this paper makes a number of contributions:

• 
  We report on and analyze perceived user needs for system awareness in DIY smart homes by examining the use cases pursued by participants and their respective information requirements.
  
• 
  We identify event data structures and aggregations that participants relied on when trying to find information that could improve their smart home awareness and we demonstrate their potential application in smart home awareness interfaces.
  
• 
  We examine an observed shift in information seeking practices towards “management by exception” in the later phases of smart home use and discuss opportunities for supporting this gradual shift through adaptive visualizations and interfaces.
  

Our findings can inform the design of DIY smart home systems to support both novice and experienced non- programming users in maintaining their smart homes. We conclude that this should involve the use of awareness mechanisms that consider how the relationship between users and their home technology evolves over time.

2.