Smart devices like mobile phones, tablets, and smart watches are designed under the assumption that they will be used by a single user. In contrast, many other devices such as smart thermostats and smart speakers are inherently sharable. This paper presents preliminary results (based on a data set of 15 participants) from an ongoing multi-methods study (using diary study and semi-structured interviews) that aims to gain a nuanced understanding of motivators and constraints of sharing such smart devices, which are cumulatively referred to as Internet of Things. Specifically, this paper illuminates the purposes and practices of sharing smart speakers and discusses two influential factors that shape these practices. Finally, we discuss the implications of our findings and provide guidelines for the design of future smart conversational speakers.
• Human-centered computing → User studies; Empirical studies in HCI.
Today, due to the increase of low-cost embedded sensors with networking capabilities multiple everyday objects have become smart devices that are collectively referred to as Internet of Things (IoT) [3]. Therefore, IoT in this paper refer to a collection of smart devices that automatically sense data about users or their environment, assist them in automating their spaces or activities, or help the users in gaining knowledge about themselves [5]. Examples of IoT, thus, can include smart speakers with voice assistant (referred to as smart speakers in this paper), wearables (e.g., activity trackers and smart watches), automated home systems (e.g., smart security systems and thermostats), or more traditional devices (e.g., phones, tablets, laptops, or desktop computers).
Prior work (e.g., [2, 8, 11, 12]) has shown that people share their
devices (e.g., phones, tablets, and laptops) and online accounts (e.g.,
streaming accounts). But we still lack a systematic understanding
IUI Workshops’19, March 20, 2019, Los Angeles, USA
Copyright © 2019 for the individual papers by the papers’ authors. Copying permitted
for private and academic purposes. This volume is published and copyrighted by its
editors.
of how constant connected Internet of Things (IoT) are shared. This
paper discusses the sharing practices of smart speakers, which were
obtained as part of a larger study that investigates families’ sharing
practices of IoT in general. The selection of smart speakers as the
focus of this paper is intentional and is based on the facts that (
Our work is influenced by two broad streams of research as discussed below: 1.1
There has never been an incentive for the industry to allow people to share devices [1]. However, previous work has shown that people do share many of their devices, and researchers have studied families’ sharing practices of personal computer (e.g., [2]), smart phones (e.g., [11, 14]), or tablets (e.g., [13]). While all of these studies have focussed on sharing of a single device, notable exceptions [8, 12, 16] investigated sharing practices of multiple devices and accounts amongst household members. Jacobs et al. discussed intentional and unintentional patterns in technology sharing of cohabiting couples [8], in a study on sharing of smart devices Matthews et al. identified borrowing, mutual use, setup, helping, broadcasting, and accidental as the primary reasons of mutual use [12], and Sun et al. focused on identifying specific challenges of watching YouTube videos together on devices including mobile phones, computers, tablets, and TVs [16]. Despite the presence of this extensive body of literature on shared use of devices such as phones, tablets, and computers, there is no published research on shared use of other IoT and voice-connected devices at this time. 1.2
Technology is pervasive among American families and even younger children are active technology users. In fact, 75% of children under 1https://www.nielsen.com/us/en/press-room/2018/nielsen-launches-new-mediatechtrender-survey-to-uncover-consumer-sentiment-on-emerging-technology.html four own their own dedicated mobile device [9]. Previous work on parental mediation (the practice of overseeing a child’s exposure to and use of technology) has categorized into three types: active mediation; where parents and children discuss and negotiate on usage, restrictive mediation; where parents set limits on use, and co-engagement; where parents and children consume content or interact with devices as a joint activity [17]. Parental mediation is a very well-studied topic in the HCI community (e.g., [7]). However, family practices that are established for using devices that are designed to be shared devices (e.g., smart speaker) and have voice input system, which creates the potential for young children to use features and access content that previously required the ability to read and write have not been studied before. Therefore, this paper also aims to explore how parents regulate and support children’s shared use of smart speakers. 2
Data for this paper emerged from a multi-methods study (comprising of diary study and semi-structured interviews) to investigate why and how people share IoT. For logging daily instances participants used the application named PACO [4]. PACO was configured to send three randomly generated reminders between 8 a.m. and 8 p.m daily for 8 weeks. Participants could also record any relevant event as soon as they observed or participated in one. In the diary study, participants were asked to log atleast 3 daily in-situ sharing instances in form of answers to the following questions for two weeks: Name of the device shared, location where the device was shared, relationship to the person with whom device was shared, purpose of sharing the device, and detailed log of the interaction with the device. After the diary study, in the interviews, we asked a list of open-ended questions such as what motivates them to share their smart devices, what are the challenges and constraints that they face, and how they coordinate shared use of devices. The interviews took 45 minutes on an average and were audio recorded and transcribed for the purpose of analysis. This study was approved by the Syracuse University Institutional Review Board.
In Phase 1, researchers individually developed the basic codes corresponding to the entries of diary study and interview transcripts of two participants. After that both authors worked together discussing and refining basic codes to reach high agreement (Cohen’s Kappa 0.85). Basic codes were grouped together to identify categories using afinity diagrams. In second phase, each author coded half of the remaining diary study logs and interview transcripts using the code book developed in Phase 1. Authors met after coding data from each participant to discuss and refine the coding scheme. The data was analyzed on an ongoing basis, which not only improved our understanding of user preferences and sharing habits, but also helped us improve the probes that we used in future interviews. The results include data collected from 15 participants who shared smart assistants. Participants were compensated 30$ for their time. 3
This section presents preliminary results comprising of overview of
sharing and two influential factors that shape the sharing practices
of smart speakers.
3.0.1 Overview of Sharing. In total 201 in-situ instances were logged
by 15 participants (age: avg. 35, min. 30, max. 60; gender: 6 male, 9
female) with a median of 12 sharing instances over the period of
14 days. Participants reported following diferent types of sharing
instances that they participated in during the diary study: (
All participants shared their devices with their children (age: avg. 13, min. 5, max. 16; 61 sharing instances) and spouses/partners (70 sharing instances). Eight participants also shared devices with visitors (e.g., co-workers (24 sharing instances), friends (36 sharing instances), relatives (44 sharing instances)) to their house. It is important to note when participants shared/co-used the device with more than one person (e.g., children and spouse; relatives and spouse), we counted that instance with respect to each sharers/couser involved in the event. Finally, 70 of the logged sharing events comprised of sharing the device in short bursts (e.g., to ask trivia question, listen to a song) that lasted between 30 seconds - 3 minutes, majority of the events comprised of sharing interactions that on an average lasted 21 minutes (min: 5 minutes, max: 30 minutes).
Participants used and shared smart speakers for entertainment purposes (e.g., listening to music, playing games, checking whether report, or asking trivia) or for assist functions (e.g., intercom-calling to other rooms in the house, hands-free phone calls and messages when hands were engaged in something else, controlling other devices, mnemonic for remembering things to be purchased, or accessing calendars of sharers). The various functionalities that smart speakers supported and their voice-enabled connectivity made sharing of smart speaker diferent than other household devices that are usually employed for one specific function (e.g., TVs, smart thermostat, smart lights). Furthermore, our findings reveal that user’s understanding of device’s capabilities or functionalities and their judgement of device’s relevance in daily lives change with time, which in turn afects user’s reasons and practices of sharing smart devices. For example, eight participants who had been using the device for more than six months, at the time when they participated in the study, used the device more because of the convenience that assist functions lended to their lives than using it for mere entertainment purposes.
Surprisingly many of our participants (7 out of 15) had multiple
smart speakers and many of these participants (4 out of 7) had
speakers from diferent companies. Such participants had clear
demarcations and preferences in terms of the activities they performed
on these devices. They reported that Google’s smart speakers are
more eficient in finding answers to complex questions as Google
has a rich database of information, and Amazon’s smart speakers
are appropriate for accessing media (e.g., music or books).
3.0.2 Specifics of Sharing Devices with Children. Majority of our
participants (10 out of 15) explained that, for children, interacting
with smart speakers is mostly enjoyable, but is (
Previous work by Sciuto et al. [15] has shown that while children start interacting with conversational agents at a very young age, children’s verbal intonation and cadences make interacting with these agents a skill to be mastered. However, all the participants (5 out of 15) with 5-7 years old children, mentioned that despite these challenges their children are interested to interact with the devices because they perceive the speakers to be person-like to the point of befriending them. Diary logs explaining children’s interactions with smart speakers personified the device using the name of the assistant (Google Now, Alexa, Siri) or person pronouns (e.g., she). But diary entries illustrating others’ interactions with the device did not personify the device and used name of the device (Echo, Google Home) or object pronouns (e.g., it). When participants were pointed to this during the interviews, they believed that this was because the device sounded like a person to their children (e.g., device addressed everybody with their first names, the tone of the speaker was perceived to be that of a human) but for elder members of the family or visitors ‘it was just a device.’
Therefore, four participants portrayed interactions with the speaker to be a ‘game with a friend’ to then teach their kids pronunciation of new words, spellings, or use the device to complete their homework (e.g., unit conversations or basic calculations). Participants also logged that they co-used smart speaker to play various kinds of educational games (e.g., strategy games, quizzes) with their children. During the interviews parents shared that they did so not only as a co-engaging activity but also expected the kid to develop various skills (e.g., deductive ability, ability to strategize) in the process.
Our study also revealed that parents modified their own or
restricted other’s interactions with the devices because of the
apprehension or realization of being copied by children. For example,
ifve of our participants recalled that they particularly do not use
their smart speakers to order anything online because voice-based
interaction makes it into a public activity, and children could learn,
imitate, and even order things that parents would not approve of.
Furthermore, 10 participants forbade every adult user of the device
to access explicit content through speakers as children could hear
and/or try to then access the content themselves.
3.0.3 The Impact of Not Understanding Smart Devices’ Behavior on
Shared Use. During the diary study, almost all participants indicated
that they are not fully aware of the capabilities of smart speakers.
This was primarily driven by their insuficient knowledge regarding
(
Participants pointed out that the absence of a visual representation2 of available functionalities (physical afordance) or inadequate feedback and information in terms of device’s limits/capabilities (cognitive afordance) led them to not use some of the device’s functionalities. Due to the lack of appropriate physical afordances many participants, for example, were unaware about the possibility to link multiple profiles/accounts to the device or securing functionality of purchasing with a passcode. While the former lead to the frustration of not being able to access personalized content as multiple people accessed content with the same account the latter lead to not using the functionality at all due to the fear of unwanted purchases by others (specifically children).
Five participants expressed that there is a lack of a possibility of granting granular control/ access levels (functional afordance) to their co-users/sharers (specially visitors and children) of their smart speakers. It is important to note that Amazon and Google provide detailed information on how to enable diferent access levels to multiple users on its support website. For example, with respect to Google Home one can link and restrict the content based on the voice of a user3. As a matter of fact, many of our participants did maintain diferent profiles or granted distinct access levels to their co-users, specifically children. However, the problem lies in the fact that a user has to purposely search for the information or read the documentation to understand any particular functionality, and the device explicitly or automatically does not convey such instructions. 4
Even though the results presented in this paper are preliminary and therefore have certain limitations our analysis highlighted following design implications:
(
As voice-connected devices are becoming increasingly common in our daily lives it is important to develop the understanding of how people use and share these devices in their daily lives. To this end we conducted a diary study for a period of 2 weeks. Our paper shows that smart speakers are frequently shared with strong and weak ties in various forms and under varying level of constraints. Based on our findings we propose design implications in terms of improving adaptability of such devices to better support children’s use and providing better cognitive afordances. These findings have implications for the design of future IoT devices, which need to meet the needs of users more closely, thereby enabling long-term
We thank all the participants for providing crucial insights into their sharing practices of Smart Speakers. Authors would also like to thank School of Information Studies at Syracuse University for use. 6 funding this work.