Stress in manual and autonomous modes of collaboration with a cobot

Highlights

• Secondary stress appraisal is higher in the manual cobot mode than in the autonomous mode.

• Heart rate is higher in the autonomous cobot mode than in the manual mode.

• Control when working with the cobot improves the outcomes of the human-cobot collaboration.

Abstract

Working with collaborative robots (cobots) can be a potential source of stress for their operators. However, research on specific factors that affect users’ stress levels when working with a cobot is still scarce. This study is the first to investigate the levels of psychological (primary and secondary stress appraisal) and physiological (heart rate) stress in human operators working in two different cobot modes (i.e., manual and autonomous). We applied an experimental within-subject repeated-measures design to 45 healthy adults (26 women, 19 men). The results show that the levels of secondary stress appraisal were lower and the heart rate levels were higher in the autonomous cobot mode. The results suggest that, when working with a cobot, control plays a key role in the emotional, cognitive, and physiological reactions during the human-robot collaboration. Implications for organizational practice are discussed.

Introduction

Collaborative robots (cobots) are increasingly being used at work. These systems work side by side with humans, often share their workspaces and tasks, and will be an indispensable tool in different future organizations (Fast-Berglund, Palmkvist, Nyqvist, Ekered, & Åkerman, 2016; Kildal, Tellaeche, Fernández, & Maurtua, 2018; Simões, Lucas Soares, & Barros, 2019), thus transforming work processes in the industry (Seriani, Gallina, Scalera, & Lughi, 2018). Due to their safety-centered design, cobots can be appealing companions for tedious, repetitive, or physically straining tasks because they help to improve ergonomics, reduce risk factors, and offload demands on robotic agents (Kildal et al., 2018; Safeea, Neto, & Béarée, 2019). They also create an opportunity for increased performance because they allow their human operators to focus on more value-added tasks (Prewett, Johnson, Saboe, Elliott, & Coovert, 2010; Safeea et al., 2019; Shah, Wiken, Williams, & Breazeal, 2011).

Although human-robot collaboration has been shown to be fruitful, achieving fluency and greater work safety in these teams has created some challenges (Baraglia, Cakmak, Nagai, Rao, & Asada, 2016). Because the human teammate plays a central role in the human-cobot partnership (Kildal et al., 2018), robots have to easily integrate with their human operators (Nikolaidis, Ramakrishnan, Gu, & Shah, 2015), and it is essential for the individual to achieve optimal psychological states through this collaboration. Unfortunately, pan-European research indicates that attitudes towards autonomous robotic systems, including robots assisting workers, have declined over a five-year period (Gnambs & Appel, 2019). Along these lines, initial research suggests that the interaction with a cobot might be a source of negative cognitive-emotional reactions or mental strain for their human operators (Arai, Kato, & Fujita, 2010; Gombolay, Bair, Huang, & Shah, 2017; Hoffman & Breazeal, 2007). However, research on specific factors that can contribute to operators’ stress is still in its infancy.

In this experimental study, we show that different cobot modes (i.e., autonomous, manual) can have a differential effect on their operators’ levels of psychological and physiological stress. To our knowledge, this is the first study to focus on both the psychological (primary and secondary appraisal) and physiological (heart rate, HR) stress response to different cobot modes in people operating with a cobot.

Stress is a potential barrier to ensuring an optimal human-cobot collaboration because it can have detrimental consequences for the person (Wallace, Edwards, Arnold, Frazier, & Finch, 2009) and the organization (Podsakoff, LePine, & LePine, 2007). According to the transactional approach to stress, stress can be defined as “a particular relationship between the person and the environment that is appraised by the person as taxing or exceeding his or her resources and endangering his or her well-being” (Lazarus & Folkman, 1984, p. 19), which means that the cognitive appraisal of a situation is essential in determining the stress experience. Lazarus and Folkman (1984) distinguish between two concurrent appraisals: primary appraisal and secondary appraisal. During primary appraisal, the demands are categorized as sources of threat, challenge (Lazarus & Folkman, 1984), or both (Folkman, 1997; Kozusznik, Peiró, Lloret, & Rodriguez, 2016). The appraisal of threat is defined as the perception that one might experience harm, whereas the appraisal of challenge is focused on potential gain or growth and accompanied by eagerness or excitement. The focus on these two types of appraisal reflects the interest in research on positive aspects of the stress process, in addition to the negative ones (Folkman & Moskowitz, 2000). During the concurrent secondary appraisal, a complex evaluative process of “what might and can be done” (Lazarus & Folkman, 1984, p. 35) about the demanding situation takes place. Secondary appraisal is influenced by the person's beliefs about control (e.g., over environmental circumstances), related to feelings of confidence and mastery of the situation. It can be operationalized as two factors referring to the “self-concept of one's own abilities” and “control expectancies” (Gaab, Rohleder, Nater, & Ehlert, 2005) when faced with the stressor.

Within this context, it is important to note that the use of solely subjective measures to assess stress may provide limited insight into the underlying psychobiological mechanisms involved. The Autonomic Nervous System (ANS) operates somewhat independently from subjective experiences and could provide crucial information about the actual activation of the stress system and the process of stress regulation (Allen, Kennedy, Cryan, Dinan, & Clarke, 2014). The ANS is rapidly activated by stressful situations, and one of the most important biomarkers is the cardiac response, measured with the heart rate (HR) (Allen et al., 2014; Pulopulos, Hidalgo, Puig-Perez, & Salvador, 2018). Importantly, previous research has shown that a stress-induced ANS response in general, and cardiac reactivity in particular, is associated with individual differences in cognitive stress appraisal (e.g., Pulopulos, Baeken, & De Raedt, 2020; Quigley, Barrett, & Weinstein, 2002; Zandara, Garcia-Lluch, Villada, Hidalgo, & Salvador, 2018). Therefore, in addition to psychological measures, investigating cardiovascular activity as a marker of ANS may provide crucial evidence about the stress response during interactions with cobots.

In human-cobot interactions, stress factors may include the changes in the nature of the work (from physical to mental activities; Argote, Goodman, & Schkade, 1983), the proximity of the robot to the human operator and the robot's movement (Arai et al., 2010), or the loss of control that can stem from the automation of robotic agents (Gaudiello, Zibetti, Lefort, Chetouani, & Ivaldi, 2015; Stein, Liebold, & Ohler, 2019). The first attempt to describe the subjective characteristics of working with a robot took place in the initial period of factory automation. In their research, Argote et al. (1983) showed that the implementation of robots in a workplace causes a shift in the work itself, from primarily manual to primarily mental activities, resulting in higher levels of employee stress. Furthermore, Elizur (1970 in Argote et al., 1983) and Mann and Hoffman (1956) revealed a difference between employees of automated and non-automated enterprises, in terms of their level of control and sense of responsibility. The introduction of autonomous robots caused their operators to experience less freedom in carrying out their tasks and perceive the results of their work as not depending on them. More recently, stress at work with a robot was analyzed in the context of legal restrictions, work safety, and well-being (Prewett et al., 2010; Złotowski, Yogeeswaran, & Bartneck, 2017). These authors conclude that stress is an important indicator of the human operator's well-being that stems from an adequate level of technical (e.g., automation and display) and social resources (Prewett et al., 2010) and is especially relevant in the domain of working with autonomous robots because they represent a realistic and symbolic threat to their operators’ safety and well-being (Złotowski et al., 2017).

Although the research on stress at work with cobots is limited, a noteworthy exception is the work by Arai et al. (2010) who assessed physiological (i.e., skin conductance response) and psychological strain (conceptualized as a state of fear, surprise, and discomfort) in cobot users. They showed that less distance between a cobot and its human operator causes physiological strain, and that a greater motion speed of the cobot induces both physiological and psychological strain (i.e., fear and surprise) (Arai et al., 2010). Overall, this research indicates that working with a cobot may increase its operator's perception of stress, a factor that may have lasting negative health and economic consequences. However, more research is still needed to understand which factors in the work with cobots are critical to provoking stress in their human operators.

One of the factors that could explain why working with a cobot can be especially stressful is the cobot's operating mode. A cobot can work in either of two modes, i.e., autonomous (also called automatic) and manual (Fast-Berglund et al., 2016), which both result in a synchronous human-robot joint action (Gombolay and Shah, 2014). Choosing a specific mode depends on the need to eliminate nuisance and insecurity during a specific task (Cherubini, Passama, Crosnier, Lasnier, & Fraisse, 2016), and it determines the level of autonomy while working with a robot (Harriott, 2015), the operator's capacity to exert control over the robot's performance (Gombolay and Shah, 2014), and the constructive engagement of the human operator (Heyer, 2010).

In the autonomous mode, the cobot controls all the operations by itself. After initiating the cobot, the human operator performs his or her tasks synchronously with it (Gombolay & Shah, 2014; Shi, Jimmerson, Pearson, & Menassa, 2012). Successive sequences are implemented repetitively, until the cycle is interrupted when the sensor systems detect an intrusion into the cobot's work space. Although the operator of a cobot working in the autonomous cobot mode is able to disengage the system and take over its job, this only occurs in certain situations, such as the failure of the autonomous robotic system (Prewett et al., 2010).

In manual mode, each step in the human-cobot collaboration is initiated by the operator (Kruger, Lien, & Verl, 2009) by touching a button on the controller or one of the sensors installed on the cobot's arm. Thus, the user operates the cobot (Shi et al., 2012) on tasks that require the flexibility and adaptability of the human operator (Charalambous, Fletcher, & Webb, 2016). A crucial question is whether the two different modes provoke different stress responses. In this domain, the cobot's autonomy and, thus, the human's limited possibility to exert control may be an important factor in explaining individual stress levels during the human-cobot interaction.

Following the occupational stress theory of demand-control (Karasek, 1979), limited control at work means a reduction in a job resource that employees need in order to deal with job demands. As a result, and in line with the principles of reactance theory (Brehm, 1966), limited control at work can have negative consequences for well-being (Kozusznik, Maricutoiu, Peiró, Virga, Soriano, Mateo-Cecilia, 2019; McCoy & Evans, 2005). Furthermore, in their Model of Autonomous Technology Threat, Stein et al. (2019) suggest that autonomous technology can be perceived as a source of stress for its users because it contributes to a general experience of threat to one's control, safety, and identity. These researchers specifically describe this as a continuum of threat, ranging from physical harm to one's realization of the loss of human uniqueness. Following Gaudiello et al. (2015), autonomous robotic agents with decision-making capabilities can be perceived as having more intrusive consequences in our lives than other non-autonomous technologies. Accordingly, taking the decision-making authority away from the human teammate and giving it to the robotic counterpart may lead to workers' negative emotional and cognitive reactions (Gombolay & Shah, 2014). Indeed, research shows that humans are not always willing to cede their control to the robots because they would rather be the decision-makers (Nikolaidis et al., 2015). This idea is consistent with the findings of Złotowski et al. (2017), who show that people perceive autonomous robots to be significantly more threatening than non-autonomous agents. The researchers (Złotowski et al., 2017) explain user reactance to the robot based on the notion of the importance of power in social interactions. According to the authors, people share a general opinion that robots should be helpful and obedient. Therefore, when they meet a decisive autonomous robotic agent, they feel less certain about the outcome of their interaction and perceive a threat to their control. Despite the undeniable value of these results, they are based on the reports of participants who watched videos of robots, and the effect of a real collaboration with a cobot on its human user remains to be seen.

Taking all of the above into consideration, in this study we expect that the manual mode will be associated with less psychological and physiological stress in cobot operators compared to the autonomous mode. Specifically, we formulate the following hypotheses:

Hypothesis 1

The levels of primary appraisal will be lower in the manual cobot mode than in the autonomous cobot mode.

Hypothesis 2

The levels of secondary appraisal will be higher in the manual cobot mode than in the autonomous cobot mode.

Hypothesis 3

HR levels will be lower in the manual cobot mode than in the autonomous cobot mode.

Addressing these hypotheses will make a contribution to the literature on the impact of cobot use on individuals in three ways. First, we conceptualize psychological stress as including both primary and secondary stress appraisal that explain the potential differences in stress levels when collaborating with a cobot. Second, we include both psychological and physiological stress measures to assess the effects of different cobot modes on the human users' outcomes. Finally, we show the robustness of the effects by applying a within-subject repeated-measures experimental design to a large sample of individuals collaborating with a cobot. Understanding the cobot mode-related stress relationship is necessary in order to improve the human-cobot interaction and prevent stress-related disorders in the workers. Gaining insight into the cobot operator's stress responses in different cobot modes has clear, practical implications for work design and training in industrial workplaces using cobots.