Cooking and Cleaning: Exploratory Investigation of Users’ Perception of How Two Domestic Service Robots (DSRs) Should Collaborate

Introduction

Domestic service robots (DSRs) are becoming more capable and affordable, with a growing trend of using robots at home (Cavallo et al., 2014). Sales statistics support this trend. In 2023, the household robot market was valued at over $10B, with a compound annual growth rate of 19.3% for 2024 to 2032 (GMI, 2024). Meal preparation and house cleaning are the most frequent and time-consuming domestic chores. Further, since the COVID-19 pandemic, more people have been working from home or alternate working from home and at their workplace, fortifying the need for more support in domestic chores. Bittman et al. (2004) claimed that domestic technology rarely reduces and even, paradoxically, produces some increases in domestic labor.

As more DSRs are introduced into the home, human-robot interaction will not be limited to one-to-one user-robot interactions. Instead, users will interact with multiple robots simultaneously or sequentially to complete different tasks. This requires richer analyses of these reciprocal relationships as well as people’s stances, challenges, and preferences. There are knowledge gaps in making the interaction with multiple robots understandable for nonexpert users and how to design robotic systems that can collaborate and predict others’ states and actions (see Ulfert et al., 2024). Establishing trust is essential for effective human-robot collaboration (Kaplan et al., 2021), as humans need to perceive robotic systems as trustworthy (Chen et al., 2018).

A taxonomy proposed by Tan et al. (2019) for multiple service robot environments highlights the importance of addressing planned and unplanned events like failures or malfunctions. While this taxonomy lays the foundation for interactions in multiple robot environments, it remains unclear to what degree people will accept DSRs to interact without their owner’s event-related consent or acknowledgment or the degree to which users will be willing to delegate authority to one DSR or more (see Olson & Sarter, 2001; Thropp et al., 2018). Specifically, it is essential to learn more about how people expect DSRs to act in situations of failure (Honig & Oron-Gilad, 2018) and how trust affects these relationships when people are at home or remotely.

Direct communication among DSRs can support multitasking, reduce the duration of domestic tasks, and moderate user effort. To illustrate, instead of a user calling a cleaning robot to clean the kitchen, the cooking robot could coordinate directly with it, if permitted. While the potential benefits are clear, we are unfamiliar with research efforts to develop user-centered multi-robot systems for nonexpert users in domestic settings. Table 1 illustrates four challenges.

OPEN IN VIEWER

Table 1.

Challenges in Multiple DSR Systems.

Design requirements	Design requirements for nonexpert users in domestic settings may differ from those of expert users in professional settings. Studies on interactions with multiple robots focus mainly on expert users in professional settings.
Understandability	Domestic robot users need to understand the reasoning process behind their robots’ behavior to collaborate effectively. Interacting with “hard to understand” robots may negatively affect user trust, user experience, task performance, and safety.
Not another “Irony of automation”	DSRs must not become another case for “Ironies of automation” (Bainbridge, 1983)—the more we automate and the more sophisticated the automation becomes, the harder it is to maintain user understanding when something goes wrong.
Level of autonomy and control management.	What happens if a glass breaks in the living room when the cleaning robot accompanies the cooking robot in the kitchen? Who needs to attend? Who is in charge? Who prioritizes what the robot should do? When is human consent required, and is it well-defined?

Figure 1 illustrates the exploration space of the current work. We explore aspects related to the introduction of multiple connected DSRs for the home by looking at (1) Stances, concerns, and preferences regarding DSRs; (2) Differences in perceptions and trust of a novel robot versus a more familiar one when the robots function well or a malfunction occurs; and (3) Preferences regarding the level of autonomy that prospective users of DSRs want in a multi-robot environment when the user is at home or operating remotely, relative to the malfunctioning robot.

OPEN IN VIEWER

Figure 1.

The exploration space of the current study describes the shared human-robot-robot interaction space.

We expect this exploration to raise awareness of the need for design concepts of multiple DSR environments and set the path for mapping the transfer of information options for future works.

Method

We developed and distributed an online questionnaire to collect preferences and stances regarding DSRs.

Results

Findings regarding stances, trust, and preferences regarding the level of autonomy are presented.

Perceptions Regarding the Required Level of Autonomy by Scenario

Recall that there were four possible scenarios varying by which robot malfunctioned (cooking or cleaning) and whether the user was at home or remote. Respondents were asked to choose from 4 levels of automation (do nothing, ask user, update user, then ask the other robot, or ask the other robot).

Figure 2 presents the distribution of user preferences for the level of autonomy in each scenario. The level of autonomy is presented in four levels. Across all four scenarios, respondents were most receptive (84%) to the middle levels of asking the user what to do or updating the user before asking the other robot for assistance.

OPEN IN VIEWER

Figure 2.

Perceptions regarding the DSRs’ level of autonomy per user location (remote or at home) and which robot is the malfunctioning robot (cooking/cleaning).

Chi-Square tests (χ²(9) = 34.69, p = 6.76 × 10⁻⁰⁵) revealed that preferences differed between the cooking and the cleaning robot and whether the user was at home or not. For the cooking robot, user supervision was always the most preferred level of automation (89%). For the cleaning robot, participants were more willing to opt for level (d) of asking for the other robot’s assistance when not at home (see columns 1 and 3 in Figure 2, 3% vs. 22%) were willing to ask for help from the other robot without their intervention.

Discussion

In this exploratory study, we looked at people’s stances towards domestic service robots in general and specifically towards a new cooking robot. Participants’ responses reflect concerns about usability, attitude, and effort. These stances align with the old debate of whether the diffusion of domestic appliance systems into the home reduces the time and effort spent on housework (Bittman et al., 2004). Concerns of competency and control were further emphasized for the unfamiliar cooking robot. Indeed, perceived trust in the robot corresponds with familiarity and understanding of each robot’s competencies and reliability (see Miller et al., 2021).

Regarding DSR environments, respondents were reluctant to allow the robots to interact without their intervention or consent. For a cleaning robot, users are more willing to accept it, asking for assistance from another robot without their intervention when they are not at home. Something they are not willing to allow for a robot they do not fully trust or understand its competencies. These barriers to adopting multiple connected robots in the home reflect the issues raised in Table 1 and impact users’ trust and willingness to adopt connected DSRs. Ulfert et al. (2024) differentiate between three entities towards which trust can develop: (1) the individual robot, (2) the dyadic relationships (e.g., User-Cooking robot, User-Cleaning robot, and Cleaning robot and Cooking robot), and (3) the team. What we see from the reported trust values and Figure 2 is that respondents tend to consider all three entity levels; clearly, they had opinions about each individual robot, about their dyadic relationship with each robot separately, and as noted by the difference in preferences concerning being at home or remote, towards the triadic team.

A major limitation of this study is that it was conducted online on a convenience sample of people who own a cleaning robot. Further, we did not manipulate robot behavior or assess how understandability can change trust.

User needs and concerns need to be mapped and addressed in future studies, including studying interactions with real robots.