Humans have multiple means to express their emotions, verbally through words and tone of voice or non-verbally through gestures, facial expressions and touch. Hertenstein et al. (2009) showed that eight distinct emotions, i.e., anger, fear, disgust, sadness, happiness, love, gratitude, and sympathy, can be communicated via touch. Although affective touch relates to the tactile communication of any emotional and social information (McGlone et al., 2014), pleasantness is one of the strongly elicited emotions that is linked to affective touch (Essick et al., 1999; Ackerley et al., 2014; Pawling et al., 2017). Due to this link, it is even mentioned as pleasant touch by Löken et al. (2009); Essick et al. (2010). Beyond being perceived as pleasant, affective touch enhanced perceived embodiment of a rubber hand (Crucianelli et al., 2013; Van Stralen et al., 2014) in variations of the classical rubber hand illusion experiment (Botvinick and Cohen, 1998). From a neurological perspective, Morrison (2016) claims that affective touch acts as a stress buffer, i.e., calibrates the stress responses of the body. Affective touch also reduces anxiety levels and autonomic responses, e.g., skin conductance and heart rate variability, which indicates a calming effect, when administered to a partner (Mazza et al., 2023). Therefore, including affective touch in human-robot interaction may be beneficial to the human’s experience.

Recently, semi-autonomous systems have been playing an important role in daily and professional life. They serve as a bridge towards fully autonomous systems, aiming to alleviate both physical and cognitive burdens on humans. Besides, semi-autonomous systems can be more favorable than autonomous systems in certain scenarios. Despite significant progress in machine learning and artificial intelligence, human expertise remains indispensable for complex tasks like surgical procedures involving medical robots, potentially enhancing patient trust Moustris et al. (2011). Moreover, human-in-the-loop systems offer flexibility and adaptability, crucial in dynamic and uncertain environments, such as semi-autonomous vehicles navigating through traffic congestion (Trautman et al., 2015; Schwarting et al., 2018).

Human-in-the-loop shared control approaches can also be useful in upper-limb prosthetics given the high dexterity required. Two main challenges in control of upper-limb prostheses are high cognitive burden of prosthesis use (Thomas et al., 2023b) and unnatural grasp movements that are incompatible with the motion of the intact limb (Guo et al., 2023). They reviewed various semi-autonomous control strategies that can reduce the workload of users and nonhuman behavior of prosthesis joints.

We claim that semi-autonomous systems, capable of adjusting their behavior while considering user experience, can enhance both performance and overall user satisfaction. Specifically, we propose the use of affective touch as a means to elevate the user experience during human-robot interaction. Achieving this goal requires a thorough understanding of the psychological factors influencing shared autonomy.

Enhancing the quality of interaction, whether it involves human-human, human-computer, or human-robot interaction, requires thoughtful consideration of the psychological states and needs of individuals involved (Cansev et al., 2021). The study of user experience (UX) in interactions with intelligent systems has gained attention in recent years (Shneiderman et al., 2017; Beckerle et al., 2018; Alenljung et al., 2019; Torricelli et al., 2020). The primary objective is to apprehend how technology can enhance the overall experience for its users. The subsequent challenge is considering the human factors as design criteria for interactive systems (Prati et al., 2021; Hao et al., 2022). By combining these approaches, developing human-centered designs will lead to enhanced UX, such as, enhanced embodiment of interactive systems (Schofield et al., 2021), which in turn fosters the research and advancements on human-centered designs including semi-autonomous systems.

Trust of a user on a semi-autonomous system is one of the most frequently considered psychological factors in discussions about shared autonomy. According to Lee and See (2004), trust is a multidimensional concept influenced by analytic, analogical, and affective interpretations of automated systems. They posit that emotions play a pivotal role, as affective processes significantly influence both analytic and analogical responses, emphasizing that trust is not only a cognitive consideration but also an emotional experience (Fine and Holyfield, 1996; Lee and See, 2004). At this point, we consider the potential impact of affective touch on eliciting emotional responses for trust-building in (semi-)autonomous systems while acknowledging the contribution of task-related analytic and analogical cognitive processes.

Analyzing and generalizing trust in automation is challenging as trust can fluctuate due to many human-related, robot-related, and environmental factors (Hancock et al., 2011; Schaefer et al., 2016). As shown in the review by Kohn et al. (2021), self-report measures along with behavioral measures are preferred over physiological measures (e.g., electrodermal activity, eye gaze tracking, and heart rate change and variability) in interpersonal trust analysis, because they are easy to validate and integrate into interaction. We believe incorporating physiological measures more often in trust analysis is important for avoiding biased results between subjective and objective measures as well as enabling real-time trust analysis. We agree with Hancock et al. (2011) that potential discrepancies between individual’s self-report and behavior can only be monitored when subjective measures are combined with objective measures. We posit that objective physiological measurements must be considered to accurately picture the state of trust. Additionally, we argue that real-time measurement of trust appears essential to create shared autonomy strategies that can adapt the behavior of semi-autonomous systems based on the trust of the user. To the best of our knowledge, the self-reported trust measurement with the most real-time capability assesses trust every 25 s via gamepad buttons (Desai et al., 2013). Such a measurement frequency would not suffice for responsive interaction in dynamically changing conditions. Among physiological measurements, both electrodermal activity and heart rate positively correlated with stress, anxiety, and cognitive workload (Caplan and Jones, 1975; Payne and Rick, 1986; Jacobs et al., 1994), emerge as valuable indicators of trust (Waytz et al., 2014; Akash et al., 2018). Since affective touch has soothing effects, e.g., acting as a stress buffer (Morrison, 2016) and reducing anxiety levels and autonomic responses under certain conditions (Mazza et al., 2023), as explained in Section 1, we believe that it can improve the trust of a person on a semi-autonomous system. For example, a prosthetic arm can autonomously mediate affective touch on the residual limb to soothe the wearer when trust on the prosthesis is decreased. This is why the real-time and accurate estimation of trust is important.

While trust is highlighted as the principal motivator and modulator of shared autonomy in this paper, we anticipate that agency, i.e., the feeling of being in control of an object’s movements (Braun et al., 2018), constitutes another psychological factor, aligning with the task-related analytic cognitive processes as defined by Lee and See (2004), in our perspective. Crucianelli et al. (2013) demonstrated that affective, slow stimuli not only enhances perceived ownership, but also improves the sense of agency, a sub-factor of embodiment (Schettler et al., 2019). The roles of other affect-related factors during semi-autonomous interaction is still to be investigated.

Social touch can induce changes in human behavior to an extent measurable by researchers in “real-world” scenarios. These changes in behavior are referred to as the Midas Touch Effect. Brief, light touches on the hand and/or arm compared to no tactile contact during social exchanges resulted in people tipping more at a restaurant (Crusco and Wetzel, 1984), being more likely to return and lend money (Kleinke, 1977), and spending more time and money inside a store (Hornik, 2014). Social touch not only affects human behavior in an altruistic manner, but it also impacts one’s attitude toward and evaluation of the toucher or environment. Students who were touched on the hand by library clerks when returning the library cards rated the library personnel and facilities more favorably than students who did not receive a touch (Fisher et al., 1976). People who received a light touch from a store employee while shopping or from a waiter in a restaurant rated the environments as more friendly than those who were not (Hornik, 2014). Social touch increased people’s willingness to participate in mall interviews (Hornik and Ellis, 1988). In a classroom setting, students were given the opportunity to write the solution to a statistical problem on the board. During the exercise, the teacher briefly touched a handful of students on the forearm. Results showed that being touched correlated with increased volunteering among the students (Guéguen, 2004). Additionally, bus drivers were more inclined to give passengers a free ride when they were touched on the arm while being asked the question (Guéguen and Fischer-Lokou, 2003). Finally, researchers have shown that people are more willing to give strangers a free cigarette if they were touched together with the request (Joule and Guéguen, 2007). Social touch is understated but plays an important role in our decision making and social behavior.

Current clinical prosthetics provide no cutaneous haptic feedback to the wearer, thus reducing the tactile interaction to a visual or purely kinesthetic stimulus: the person will see the touch or feel the force via the prosthetic socket on the residual limb from the touch. This begs the question, “What happens when one’s sense of touch is missing, such as for an upper-limb prosthetic user? Will the Midas Touch Effect remain if a person receives a touch on a prosthetic limb?” We hypothesize that prosthesis wearers do not experience it in the same manner, but an emotional change may still occur. Therefore, we advocate for researchers to investigate the social haptic needs of upper-limb prosthetics wearers. This supports placing equal emphasis on improving social touch experiences for prosthetics users and can enhance prosthesis acceptance and improve quality of life.

We suggest that researchers investigate if and to what extent social touch enhances upper-limb prosthetics user experience. This may include distributing questionnaires that ask what their social touch experiences are in everyday life. Furthermore, user-studies that explore prosthesis wearer reaction in social touch scenarios, both in a lab and real-world setting, could be conducted. The Trier Social Stress Test (Kirschbaum et al., 1993), and other laboratory psychological experimental paradigms (McCarthy and Elson, 2018) have been used for decades to uncover underlying emotional experiences. For upper-limb prosthetics, the Midas Touch paradigm provides an appropriate framework for this, as previous studies without prostheses have been able to obtain quantitative data describing the effects of social touch.

To relay affective touch to the user, appropriate hardware must be used such as touch sensors on the prosthesis that signal haptic feedback displays to provide the cue to the wearer. Researchers have developed sensory skins specifically for prosthetics (Chortos et al., 2016; Yang et al., 2019; Chen et al., 2021), and sensors for social touch mediation (Eid and Al Osman, 2016; Huisman, 2017). Haptics researchers have created haptic displays that can relay social touch to the wearer (Huisman et al., 2013; Casini et al., 2015; Culbertson et al., 2018), which could be modified for prosthesis users. Furthermore, thermal displays can be incorporated to provide the realistic warmth felt during skin-to-skin contact in social touch studies (Iberite et al., 2023; Muheim et al., 2024).

To address the emotional and social aspects of the multi-faceted field of human-robot interaction, especially in prosthetics research, we suggest taking a holistic human approach and setting design requirements based on human needs. Maslow’s hierarchy of human needs involves social and emotional wellbeing, which are met through social haptic interaction. Recognizing social and psychological needs of a person, in addition to the physical functional requirements needed to lead a healthy life, will shift human-robot interaction research from a narrow focus on device advancement to the overall promotion of human wellbeing.