The Social Perception of Autonomous Delivery Vehicles Based on the Stereotype Content Model

Abstract

Innovations like autonomous delivery vehicles (ADV) have the potential to transform last-mile delivery to make it more sustainable and human-centered. Yet only if these technologies are socially acceptable can they live up to their potential. Using the Stereotype Content Model (SCM), we assessed how different social groups were perceived when they used autonomous delivery vehicles. Based on the two dimensions of the SCM, warmth and competence, we found combinations of group stereotypes and use of ADV that impact social acceptability, and we identified systematic effects of gender and vehicle usage on social acceptability. Our results highlight the importance of social perception for the acceptance of autonomous vehicles and the relevance of the intersections of gender, vehicle usage, and social group for an accurate and comprehensive evaluation of the social acceptability of autonomous delivery vehicles.

1. Introduction

Emerging technologies diffuse into all areas of life, and technology use is increasingly taking place in social settings—whether at work, in private, or in public spaces—making the social acceptability of technological devices crucial for adoption [1]. In 1994, Nielsen named social acceptability as an essential part of system acceptability [2]. For a long time, however, research in human–computer interaction focused mainly on subjects like usability and user experience [1], and conceptualizations of the social acceptability of technology are still rare (see [3] for a review, also [1,4]). Insufficient social acceptability of a technological device can have a number of devastating effects on users, including negative self- and external image [5], the risk of stigmatization, misperception, and negative judgment from others [3,4,6,7,8,9,10]. Especially in public spaces, the social acceptability of technological devices thus needs to be considered. Studies have investigated the social acceptability of technology usage in a number of different contexts and situations, for example for the use of mobile devices [4,11,12], data glasses [7,13], or head-mounted displays [9,10,14,15] and for specific interaction modes such as gestures (for an overview see [3]). A considerable number of these studies found a link between the social acceptability of technologies and the perception of different user groups, e.g., that devices are more accepted if they assist people with a disability [14]. The use of different technologies significantly changes the social perception of different social groups, and moreover, devices themselves are perceived in stereotypical terms [4]. Prior research that assessed the perceptions of different technological devices like tablets, VR headsets, or quadcopters, in conjunction with different user groups, used the Stereotype Content Model (SCM) [16,17] to explain the social acceptability of devices for different user groups. Theoretically grounded in the Stereotype Content Model, the present study extends existing research in two aspects. Firstly, we assess the social acceptability of electric autonomous delivery vehicles (ADV), a technology for last-mile delivery promising to improve efficacy regarding delivery times, cost, and environmental footprint through CO2 reduction and efficient energy profiles [18]. Secondly, we investigate the effect that the assumed use of autonomous delivery vehicles has on the stereotypical perception of a range of different user groups, while considering subgroups in their intersection with gender. Although gender is a fundamental source of stereotyping [19,20,21,22] and often leads to very distinctive perceptions on subgroup levels [20,23], it has often been neglected in previous studies on perceptions of technologies [4,24].

Most of the work done regarding acceptance of autonomous vehicles has primarily researched the transport of people in cars or buses (see e.g., [25,26] for reviews). Autonomous electric vehicles for the transport of goods have the potential to substantially alter and improve last-mile delivery [27], yet little research has been done on the acceptance of these autonomous delivery vehicles [28,29,30]. As the SCM has been found useful in an array of different fields and areas, including understanding the social acceptability of a range of mobile devices [4,31], we refer to the SCM to develop a deeper comprehension of the underlying principles of the social acceptability of autonomous delivery vehicles and their effect on the stereotypical perceptions of different user groups and vice versa. Moreover, as the SCM draws on two basic dimensions of human perception, its application may offer insights, which are transferable to related scenarios in the use of autonomous vehicles.

2. Background

The following section investigates different models describing the mechanisms of social acceptability of technology and technological devices. In the following sections, we draw on previous research related to the social perception of autonomous vehicles. We start by describing the different models of measuring the social acceptability of technologies. We then denote the influence that different social user groups have on the perception of technological devices and how stereotypical notions need to be considered as general principles of social perceptions when investigating the acceptability of technological devices. Moreover, we refer to the influence of different social user groups on the stereotypical perception of technological devices in terms of the Stereotype Content Model [16,22]. As the present study aims to extend findings related to the stereotypical perception of technologies by focusing on the acceptability of autonomous vehicles and considering gender as a source of stereotyping, previous research on the perception of autonomous delivery vehicles (ADV) is reported and our research questions are derived.

2.1. Social Acceptability of Technological Devices

Generally speaking, social acceptability is defined as “the absence of social disapproval” by the APA Dictionary of Psychology [32]. This definition by negation is also adapted by Kelly and Gilbert, who developed the WEAR Scale to measure the social acceptability of a wearable device and suppose that “a socially acceptable wearable is most notably marked by an absence of negative reactions or judgments from others”. This absence of fear needs to be accompanied by the fulfillment of aspirational desires to account for the social acceptability of a wearable device [12].

Moreover, social acceptability is construed as a reciprocal and contextual rather than an isolated process. According to Goffman’s theory of impression management, asserting that all public action is a performance, people cultivate their identity not only for themselves but present themselves to others by adapting the roles of performers for the spectators [5]. This duality of social acceptance was described by Montero et al. [33], who defined social acceptance on two dimensions: (a) the user’s social acceptance, an internal effect of the interaction that defines the user’s subjective impression; and (b) the spectator’s social acceptance, an external effect of the user’s interactions. In their study on the social acceptance of gestures, the spectators’ perceptions of others depended on their ability to interpret the performers’ device manipulations. This duality between the users (performers) and others in their vicinity (spectators) is present throughout a large number of previous works [7,34,35,36]. These findings underline the importance of social, reciprocal perceptions regarding the social acceptability of technologies and devices.

The widely used Technology Acceptance Model (TAM) describes the adoption of new technologies by individuals as influenced through two main factors: perceived usefulness (PU) and perceived ease-of-use (PEOU) [37]. The growing awareness of the importance of social factors when considering technology acceptance led to the creation of derivate models of the TAM (e.g., UTAUT), featuring subjective norms and social influence [38,39]. However, they only account for positive social influences, i.e., feedback encouraging the interaction because “an individual perceives that important others believe he or she should use the new system” ([40], p. 451), and the TAM/UTAUT does not take the perception of others who use a technology into consideration. Yet the perception of others who use a technology is crucial to understanding not only the acceptance, but the acceptability of a technology, which as a reciprocal process relies on the anticipation of other people’s approval or disapproval. As such, social acceptability of technology has been shown to highly correlate with ratings of the dimensions of the Stereotype Content Model [4,11] and, thus, the SCM has been validated as a method to assess social acceptability in technologies [31].

2.2. Social Perception of Devices and Their Connection to Different User Groups

A substantial number of studies indicate that the perceived social acceptability of a device also depends on who is using it. Profita et al. assessed the perceived social acceptability of head-mounted displays and discovered that the use was considered more socially acceptable if the device was being used to assist a person with a disability [14]. Rico and Brewster conducted a study on the social acceptability of gestures for mobile interfaces [41]. The authors, featuring strangers, colleagues, friends, family, and partners as potential users, suggested that the social acceptability increases with the familiarity of the audience. Schwind et al. showed that the acceptance of consumer-grade VR glasses depended on the situation and the number of people in the context of the user [10]. Moreover, bystanders, who express privacy concerns, reduce the social acceptance of devices with built-in cameras, such as life-logging cameras or smart glasses, as shown in works by Koelle et al. and Wolf et al. [7,13]. Subsequentially, the presence and reactions of others can have substantial effects on the users themselves, who then may feel uncomfortable and embarrassed and less willing to use the device [42,43].

Altogether, the perception and social acceptability of devices and interaction techniques depends on the perception of the user groups and their social contexts. Research has shown that social acceptability is higher if a technology supports a person with physical disabilities [14], indicating that the perception of certain groups that use a technology and the perception of this technology are intertwined. Accordingly, as perceptions of other people are influenced by stereotypes, it needs to be considered how these stereotypes may change the social perceptions of a device systematically.

2.3. Stereotypes

Our perception of other people usually happens intuitively and largely automatically, i.e., without conscious effort [44,45,46]. These perceptions are—at least partially—informed by stereotypes. A stereotype is a simplification of a complex social reality; it comprises knowledge and assumptions about the attributes and behavior of people [47]. Stereotypical notions can be activated by implicit associations [47]. This way, stereotypes can influence our social perception and our behavior, even without our awareness [48,49]. While stereotypes are socio-cognitive knowledge structures that each person carries within themselves, they are also socially shared, created, and recreated through social situations. This dual structure of stereotypes leads to a continuous validation through consensus, leading to an asserted knowledge of a social group that is perpetuated in every act of repetition [19].

2.4. The Stereotype Content Model

Regarding the content of stereotypes, there are two basic dimensions that systematically organize people’s perception of others and thus help them navigate the social world: the assessment of someone’s warmth and competence [50,51]. The Stereotype Content Model (SCM) is a central framework in social psychology, linking these two dimensions to the human disposition to assess someone’s intent to either harm or help them [17,21,22]. The dimension “warmth” refers to how benign and likeable members of a certain group are perceived, thus evaluating what the person’s intent might be and whether they have positive or negative intentions toward oneself. The dimension “competence” describes how able and independent someone is considered, thus evaluating how effectively a person can pursue their intentions and how capable they are to either benefit or harm oneself. The SCM offers a framework to explain how the perception regarding these two dimensions elicits distinct patterns of affective and behavioral responses [21,50]. High ratings on both the warmth and the competence dimensions are usually reserved for one’s ingroup (i.e., the group oneself feels part of) and groups that are generally highly admired in a society. Notably, perception of outgroups is frequently not generally negative, but ambivalent, with high values on one dimension and low values on the other dimension [16,22]. One example for such ambivalent stereotypical perceptions are “rich people” or “business women”, which are perceived as cold but competent and are consequently envied, whereas other groups such as housewives, the elderly, and people with disabilities are perceived as warm, but incompetent and are pitied [22,50]. Cold and incompetent groups (e.g., homeless people) are despised, and warm and competent groups (usually one’s own ingroup) are admired. This positive assessment of the ingroup is called “ingroup favoritism”. Typically, groups that are classified similarly on these dimensions, i.e., are from the same cluster, elicit similar emotional and behavioral responses [16]. To sum it up, according to the SCM, the stereotypical status of a social group is predicted by four competence–warmth combinations: paternalistic stereotypes (high warmth, low competence), admirable stereotypes (high warmth, high competence), contemptuous stereotypes (low warmth, low competence), or envious stereotypes (low warmth, high competence); see also Figure 1.

This pattern of uni- and ambivalent perception has been found to be remarkably stable across 19 different countries [50]. The SCM has been applied to various fields and found useful for understanding the mechanisms of racism [52], ageism [53], and sexism [54].

2.5. Gender as a Source of Stereotypes

Gender has been found to be a strong and prevalent source of stereotyping. In terms of the content of gender stereotypes, women are stereotypically perceived as warmer than men, but less competent [19,20,22,55,56,57,58]—a phenomenon that has been found to be interculturally stable [59]. Gender stereotypes tend to be constant over time [57,60,61]; however, recent research shows that the difference between men and women is getting smaller regarding the competence dimension and women are progressively seen as having the same level of competence as men [23,62,63]. Eagly and colleagues [64] refer to the analyses of public opinion polls in the US from 1946 to 2018, to show that women are increasingly seen as equally—or occasionally even more—competent than men. The different perception regarding the warmth dimension remains unchanged: women are still considered more communal and less agentic than men.

While there are overarching stereotypes for men and women, people distinguish between the stereotypic traits of several male and female subgroups [54,65,66]. For example, Wade and Brewer explored subgroups of women and found the “homemakers” and “business women” to be highly divergent in their stereotypical perception [67]. The participants of a recent study in Germany rated unemployed people as low in warmth and competence, thus denoting contemptuous stereotypes; pensioners were rated as medium in competence and warmth; and physicians received high ratings in both warmth and competence [68]. However, the perceptions of gender subgroups were not considered in this study. Various other studies referred to the SCM to explore occupational stereotypes (e.g., [69,70,71,72]), but only some of them consider gender subgroups. Yet when the gender of a person or group member is not explicitly specified or marked as “female”, usually a masculine mental representation of those group members is activated when reading a group name, in line with the findings of linguistic psychology (for overview see [73]). Thus, instead of measuring the perception of a seemingly “gender-neutral” group, participants are actually predominantly rating the group members of one gender only—which is the male one in most cases. Likewise, according to the model of intersectional invisibility [74], due to the ideologies of androcentrism, ethnocentrism, and heterosexism, the prototypical human is a man, the prototypical citizen (in a Western context) is white, and heterosexuality is prototypical of human sexuality. Members of subgroups are thus rendered invisible when they belong to two subordinate groups (e.g., ethnic minority women or ethnic minority gay men). These findings stress the importance of differentiating between subgroups when studying stereotypical group perception.

The perception of professions and gender are linked: several studies demonstrate (e.g., [75,76]) that healthcare professions such as social workers, psychologists, and nurses and jobs in education such as school and kindergarten teachers are linked to warmth traits, as women are considered to have more “social qualities” [19], whereas managerial and technical professions are not considered to require such traits. Men, on the other hand, are stereotypically perceived to possess better reasoning and technical skills than women; thus, industrial engineers or electrical apprentice are perceived as “masculine” [75,76,77].

Thus, the perception of these subgroups often reflects the combination of gender with other social roles such as parental or occupational status [56,58] or the intersection of gender with other social categories such as race/ethnicity and sexual orientation [78,79,80]. Bye [23] assessed and analyzed the stereotypical perceptions of men and women across gender subgroups. She discovered that even in contemporary Norway, women as well as men in care-giving roles are depicted as stereotypically warm but low in competence. On the other hand, women and men in traditional male roles (leaders, politicians, or business people) are perceived stereotypically competent and low in warmth. Women were usually viewed as warmer on a subgroup level, whereas competence ratings were inconsistent across different subgroups. These findings are in line with social role theory [58] and furthermore underline the importance of considering the intersection of different subgroups, to gain insights on how groups, in other words, people, are perceived.

Notably, gender stereotypes do not only have descriptive aspects (i.e., how members of a group are) but also have prescriptive aspects, i.e., what members of a group should be like [19]. Deviations of these prescriptive stereotypes can lead to rejection or punishments. Such social and economic sanctions for counter-stereotypical behavior are known as the backlash effect [81]. Referring to the SCM, this is also evident in the stereotypical perception of subgroups, e.g., that business women are perceived as more competent than the “average woman”, but cold(er), whereas businessmen are perceived as competent, but without the loss of warmth that business women receive [22]. Another example refers to working women, who, when they become mothers, usually trade in a gain in warmth for a loss in competence while this does not affect men who become fathers—the latter even gain in perceived warmth [20].

2.6. The SCM and the Perception of Technological Devices

As the SCM refers to deeply rooted mechanisms of human perception and inter-social behavior, it has been applied widely and considered helpful in understanding a range of phenomena in different fields, including the perception of nonhuman entities, such as brands and mobile devices. Previous studies demonstrate that consumers perceive brands in a similar way to how they perceive people, and they also feel and behave towards brands in ways that are comparable to their behaviors towards other people [82,83]. Even national stereotypes about a product’s country of origin can influence consumers’ expectations and willingness to buy a product [84]. The stereotypic perception of the “typical users” of a brand also seems to have an effect on people’s desire to buy and own a brand [85].

Frischknecht [24] refers to the SCM in her study on the general perception of autonomous technical systems. She suggests that increasing levels of technical autonomy change the attribution of competence and agency, while attributions of warmth and experience do not change under conditions of increasing technical autonomy.

A study by Hohenberger and colleagues [86] also found also a difference between women and men regarding their willingness to use autonomous vehicles related to their emotional responses to “automated cars”. Men were more likely to associate positive emotions, while women were more likely to associate negative emotions towards automated cars. However, their study was not related to ADV, but only to autonomous vehicles in general.

Beyond the influence of user groups, contextual factors of device use have also found to be influential regarding the perception of wearable devices in terms of the SCM. People differentiate between the placement of the device, particularly devices in the user’s hand, and activities in which the device can contextually be misused [31].

Schwind et al. [4] investigated if the principles of the Stereotype Content Model (SCM) can be applied to the perception of mobile devices. In a first study, they assessed various combinations of mobile devices and different group stereotypes and showed that the anticipated use of different mobile devices had a systematic effect on the perceptions of stereotypical warmth and competence of a range of social groups. Some devices, such as LED glasses or a life-logging camera, located nearly all groups into the contemptuous quadrant (low-warmth and low-competence). On the other hand, medical devices generally increased perceived warmth. However, not only did the strength of influence vary across devices (e.g., LED glasses, life-logging camera, and blood glucose sensors had stronger effects than VR or EEG headsets), but the effects also varied across different social groups. Senior citizens using devices that are usually seen as very competitive when associated with other groups (VR headsets or quadcopters) showed a higher influence of the device on perceived competence than on warmth.

The authors backed their assumption that different devices have their own location in the SCM, i.e., that a mobile device is recognized as a social object and perceived stereotypically by itself, in a second study, which measured the stereotypical perceptions of the different devices independently of a specific user group. Thus, Schwind et al.’s [4] combined findings support the notion that a systematic shift in the perception of warmth and competence occurs when people use a certain technological device as described by the SCM. Moreover, their results indicate that stereotypical perceptions of devices are not independent of the people that are its users (and vice versa) and the social acceptability of said devices depends on the assumed abilities of the user’s social group.

However, their study only focused on the perception of mobile devices (a specific, albeit heterogenous group) as social objects. Also, they did not differentiate between group members according to their gender when assessing the influence of different potential user groups. As pointed out above, gender is a highly prevalent source of stereotyping, and the perceptions of members of the same social or professional group often differ according to gender-related subgroup stereotypes: hence, gender may elicit very dissimilar, sometimes ambivalent stereotypes. Moreover, the present study aims to extend the previous findings regarding the stereotypical perception of devices to the perception of autonomous and self-steering vehicles and varying potential user groups within the framework of the SCM.

2.7. The Perception of Autonomous Delivery Vehicles

Autonomous and self-steering vehicles have been used in pilot projects for years (e.g., [30,87]) and they are likely to become common in real-world settings [88]. Their potential to reduce emissions and increase safety in urban areas is considered to be tremendous, but their use is also controversial, e.g., because of ethical concerns [89,90,91,92]. Most previous studies so far focused on the acceptance on AVs in general, often for human passengers, e.g., autonomous cars or shuttles (see [26] for a review on shuttles); see also [93,94,95]. However, it needs to be considered that the perceptions of vehicles used for the transportation of people may differ from the perception of those used for the transportation of goods, as the involvement of people as passengers could influence the perceptions of the vehicles (e.g., safety concerns), and ways of use are different (e.g., shared use in a neighborhood vs. private use). Likewise, the context and the assumed activities may influence the perceptions ([31]; see also Section 2.6).

Autonomous delivery vehicles (ADV) are electric and self-driving ground vehicles, which drive on streets or sidewalks with a limited speed of 5–10 km/h and are able to manage all driving tasks by themselves without human intervention in a mixed traffic environment. [30]. They are equipped with space for the transportation of various sensors, cameras, and GPS tracking and can be used to deliver goods such as groceries and parcels directly to the doorstep (last-mile logistics). Unlike conventional delivery methods, the interaction is based on a human–technology interface (i.e., mobile, so no human–human interaction is involved during the final drop-off [30]. They are still in the trial phase and mainly used in pilot projects, e.g., in Germany/Heilbronn [30], Düsseldorf, and Hamburg. ADV are believed to have the potential to revolutionize last-mile delivery in a way that is more sustainable and customer-focused, but only if they are widely accepted [28]. However, the research on the acceptance of ADV is still in its beginnings [29,30,96].

Marsden et al. [30] assessed the attitudes towards ADV and found that ADV were generally seen as rather “innovative”, “environmentally friendly”, and “interesting”, but men in general showed slightly more positive attitudes (and slightly less negatives ones) towards ADV and also indicated a stronger wish to use them than women. Moreover, some safety concerns were expressed; however, participants rather disagreed with the suggestion of “dangerous”, “uncanny”, and “senseless” ADV. Additionally, ADV were considered to be rather well-suited for people with physical constraints and least suited for older people.

Kapser and Abdelraham [97] applied the Unified Theory of Acceptance and Use of Technology (UTAUT) to investigate the user acceptance of ADV in the context of last-mile delivery. In their study, performance expectancy has been found to be the most important variable, followed by social influence, whereas no effect could be found for effort expectancy. In the following, Kapser and Abdelraham [29] employed an extended Unified Theory of Acceptance and Use of Technology (UTAUT2) and adapted it to the context of ADV in last-mile delivery including risk perceptions. They found price sensitivity is the strongest predictor of behavioral intention (i.e., user acceptance), followed by performance expectancy, hedonic motivation, perceived risk, social influence, and facilitating conditions, whereas no effect could be found for effort expectancy. Further on, they extended the model (UTAUT2) in another study which served to empirically verify previous assumptions of their models and included gender as a moderator variable [28]. Their results indicate that trust in technology, price sensitivity, innovativeness, performance expectancy, hedonic motivation, social influence, and perceived risk determine behavioral intention. However, no significant effect of social influence, hedonic motivation, and perceived risk on behavioral intention could be found for men—these were only relevant for women. Notably, on average, the participants showed a rather “neutral” attitude towards ADV, which indicated that they have not yet formed a strong opinion on ADV. Pani and colleagues [96] conducted a study on the public acceptance of autonomous delivery robots (which are, in essence, similar to ADV) during the COVID-19 epidemic. They offer a detailed analysis of consumer preferences, trust, attitudes, and willingness to pay (WTP); however, their study focused on marketing aspects and different buying habits and attitudes.

While these findings offer many insights on the attitudes towards autonomous vehicles and their social acceptability, the results are sometimes fragmentary or inconsistent and integrated in several different frameworks. Some of the mentioned concepts, such as “trust” in autonomous vehicles [98], also relate to the SCM. Gender differences have been argued to play a crucial role in technology acceptance research [39,40], but were often not considered. Many of the mentioned studies also did not consider the influence of the social context, e.g., (other) users and their perceived stereotypes, including gender, on the perceptions and acceptance of a technological device.

2.8. Research Questions and Hypotheses

The present study aims to close the aforementioned research gap by assessing the social perception of autonomous and self-steering delivery vehicles in terms of the SCM and how the association with ADV affects the stereotypical perception of a person based on their social group, considering the intersection of vehicle use, gender, and specific group membership. Thus, we focus on the following research questions:

How does the usage of autonomous delivery vehicles (ADV) affect the perception of warmth and competence of members of different social groups?

Hypotheses:

H1.: 

Female social groups are perceived warmer than male social groups when they are using ADV.

(Null hypothesis (H0) of H1: Female and male social groups are perceived comparably warm when they are using ADV.)

H2.: 

Social groups, who use are using ADV are perceived as less warm than social groups not using ADV/basic versions (main effect).

(H0 of H2: Social groups who use ADV are perceived as equally warm as social groups not using ADV.)

H2.1.: 

The decrease of warmth for groups using ADV is stronger for female social groups than for their male counterparts (interaction).

(H0 of H2.1: The decrease of warmth for groups using ADV is similar for female and male social groups.)

H3.: 

Social group members that are using ADV are perceived as more competent than groups not using ADV/their basic versions (main effect).

(H0 of H3: Social group members that are using ADV are perceived as competent as groups not using ADV/their basic versions.)

How is the stereotypical perception of different social groups influenced by the participants’ characteristics, i.e., their own group memberships (e.g., gender, age…)?

H4.: 

Participants show ingroup favoritism regarding the warmth and competence ratings of social groups of their own gender.

(H0 of H4: Participants rate the warmth and competence of social groups of their own gender similarly to social group members of the other gender.)

How does the influence of using autonomous and self-steering vehicles on competence and warmth perceptions affect different subgroups?

Referring to previous studies on the stereotypical perception of devices, we generally expect increases of competence, and we expect groups that are subject to paternalistic stereotypes lose comparably less warmth [4,14].

H5.: 

Groups that are subject to paternalistic stereotypes (high warmth/low competence) lose less warmth when they are using ADV compared to groups who elicit envious stereotypes (high competence/low warmth).

(H0 of H5: Groups that are subject to paternalistic stereotypes (high warmth/low competence) lose the same amount of warmth when they are using ADV compared to groups who elicit envious stereotypes (high competence/low warmth).

3. Method

The main goal of the study was to assess how the usage of autonomous delivery systems is perceived in terms of the SCM and how this association affects the stereotypical perception of a range of social groups.

3.1. Design

Our study is based on eight social groups (covering professions, social status, or other social attributes), each presented as either female or male. German is a grammatical gender language [99]; therefore, person denotations in German usually come in a male and a female version. Hence, we have a total of 16 social groups: male/female physicians; male/female homemakers; male/female students; business men/women; unemployed men/women; men/women with handicaps; retired men/women; and men/women (see Appendix A.1 for a list with all German role names).

Each of these 16 groups was either presented in its “basic version” (i.e., not using ADV) or additionally described as “using autonomous and self-steering delivery vehicles” (using ADV). The treatment therefore had 8 × 2 × 2, i.e., 32 conditions, as each social instance group has four possible versions (conditions), e.g., business men; business women; business men using autonomous delivery vehicles; business women using autonomous delivery vehicles. Adding the condition of participant gender, we account for 2 × 32 (8 × 2 × 2 × 2) conditions, i.e., 64 conditions.

The assignment to the treatment conditions was randomized, i.e., half of the members of the social group were presented as female, the other half as male, and within those groups, one half was additionally described as “using autonomous delivery vehicles” while the other half not (basic version/not using ADV). The assignment to the conditions was balanced regarding the frequency of the conditions: each participant rated two group members presented as male “using autonomous and self-steering delivery vehicles”, two group members presented as male in their “basic” version, two group members presented as female “using autonomous and self-steering delivery vehicles”, and two group members presented as female in their “basic” version. Participants were never exposed to female and male members of the same social group, nor to the same group in the basic and in the “autonomous delivery vehicles” version. Each participant rated only one of the four possible versions of each of the eight social groups.

The conditions were treated as between-subject factors as participants were not exposed to female and male members of the same group to avoid spilling-over effects and early cancellation.

3.2. Materials

We used a shortened version of the SCM questionnaire [22,50] in accordance with Asbrock [100,101] measuring the dimensions “warmth” and “competence” with three items each (see Appendix A.2). Participants rated each of the eight social subgroups. Several criteria were considered for selecting the different social groups for the study. All groups should have already been used in previous studies referring to the SCM, at least in one (gender-)version. Our variety of groups should feature groups that are roughly distributed equally on all four quadrants: ambivalent groups eliciting pity (high warmth/low competence), i.e., elderly/retired people and housewives; negatively perceived groups eliciting contempt (low warmth/low competence), i.e., unemployed people; ambivalent groups that elicit envy and are perceived as cold (low warmth/high competence), i.e., business women; and groups that elicit admiration, i.e., physicians (e.g., [16,22]; see also Figure 1 in Section 2.4). Perceptions for “Men” and “Women” were also included, as they serve to provide a “baseline” effect without explicit occupational stereotypes and are also well researched [16,101]. We used “warmherzig” (warm), “freundlich” (good-natured), and “sympathisch (likeable) for assessing the warmth dimension; and we used “kompetent” (competent), “eigenständig” (independent), and “konkurrenzfähig” (competitive) to measure the competence dimension. We made one minor change to Asbrock’s version [100]: we translated “good-natured” into “freundlich” (i.e., “friendly”), because Asbrock’s German translation as “gutmütig” implies a certain naivety and gullibility [102], which, from our perspective, the English original does not imply and we therefore found problematic in this context. All items of the SCM questionnaire used a 5-Point Likert Scale.

3.3. Participants

The study was conducted from April to October 2019 at the biennial federal fair for gardening and landscape architecture in Germany. In total, 416 participants took part by filling out a digital questionnaire on tablet computers. Due to a high number of missing values, six datasets had to be excluded during analysis, leaving to a sample size of 409 participants (female 47.8%, male 47.8%, diverse 1.2%, not indicated 3.1%).

Participants ranged in age from 18–29 years (20.4%), to 30–39 years (13.7%), 40–45 years (11.8%), 50–64 years (33.9%), 65 years and older (17.5%), and 2.6% did not indicate their age. Other demographical data regarded citizenship (86.5% reported a German citizenship, 7.7% indicated a German plus another citizenship, 2.9% indicated another citizenship, and 2.9% did not indicate any citizenship).

Additionally, we asked a general question about the participants’ attitude towards autonomous and self-steering vehicles, using a 5-Point Likert Scale, ranging from positive (1) to negative (5).

3.4. Survey Procedure

The participants were either in an exhibition room featuring a model of a vehicle for autonomous delivery of goods or on the premises close to the exhibition room where such a vehicle was driving around. The vehicles were characterized by a number of locked, albeit accessible, boxes, where deliveries could be either received or sent out by different users, all controlled via an app.

Filling out a digital questionnaire on a tablet computer (using the survey software keyingress), each participant rated one version of each of the eight social groups. As pointed out above, the assignment to the different setup of groups was randomized but featured a balanced probability factor within the programming of the questionnaire, so that with a total of about 400 participants, each specific subgroup would be rated approximately equally often.

Participants rated each member of the eight social subgroups on the dimensions “competence” and “warmth” using a German version of the SCM scale by Cuddy and Fiske [16,50] (see Appendix A.2). Demographical data were gathered at the end.

4. Results

We took the mean of the three items to calculate the value for the dimensions “warmth” and “competence”. The values for both dimensions for all 32 groups are presented in

Table 1. Means and standard deviations (in brackets) of warmth and competence for all social groups.

Groups	Competence	Warmth
Men using ADV	3.67 (0.78)	3.24 (0.75)
Men	3.65 (0.74)	3.22 (0.73)
Women using ADV	3.79 (0.80)	3.52 (0.77)
Women	3.68 (0.74)	3.89 (0.61)
Male physicians using ADV	3.91 (0.80)	3.35 (0.74)
Male physicians	3.78 (0.75)	3.32 (0.61)
Female physicians using ADV	3.90 (0.81)	3.44 (0.76)
Female physicians	3.93 (0.77)	3.53 (0.71)
Female homemakers using ADV	3.60 (0.84)	3.46 (0.78)
Female homemakers	3.36 (0.79)	3.73 (0.79)
Male homemakers using ADV	3.50 (0.83)	3.31 (0.71)
Male homemakers	3.41 (0.80)	3.72 (0.72)
Male college students using ADV	3.79 (0.84)	3.46 (0.68)
Male college students	3.64 (0.73)	3.66 (0.71)
Female college students using ADV	3.86 (0.78)	3.64 (0.70)
Female college students using ADV	3.83 (0.66)	3.71 (0.80)
Business men using ADV	3.97 (0.66)	3.13 (0.67)
Business men	3.88 (0.71)	2.79 (0.72)
Business women using ADV	4.01 (0.75)	3.31 (0.79)
Business women	3.89 (0.70)	3.24 (0.63)
Unemployed women using ADV	3.02 (0.90)	3.22 (0.82)
Unemployed women	2.85 (0.77)	3.24 (0.74)
Unemployed men using ADV	2.91 (0.86)	2.92 (0.71)
Unemployed men	2.56 (0.67)	2.89 (0.58)
Men with handicaps using ADV	3.43 (0.83)	3.53 (0.73)
Men with handicaps	3.16 (0.90)	3.58 (0.78)
Women with handicaps using ADV	3.44 (0.79)	3.62 (0.63)
Women with handicaps	3.14 (0.80)	3.71 (0.80)
Retired men using ADV	3.36 (0.85)	3.48 (0.73)
Retired men	3.05 (0.68)	3.54 (0.67)
Retired women using ADV	3.37 (0.87)	3.40 (0.75)
Retired women	3.08 (0.84)	3.62 (0.77)

.

4.1. Stereotypical Rating of Different Social Groups

The ratings for warmth and competence of all basic groups are visualized in Figure 2; the ratings for the groups using ADV are displayed in Figure 3.

4.2. Aggregation of Data

In a second step, the main effects and interactions of group gender and ADV were examined. As each participant rated four male members of subgroups in total (of which two were “using ADV”) and four female members of subgroups (of which two were “using ADV”), the two male groups using ADV were combined into one value and the two male groups not using ADV into another value. The female groups were combined likewise. This resulted in a variable of the stereotypical perceptions, carrying the information for all conditions (gender; group membership; with or without using ADV) for each participant. Participants who did not indicate their gender or indicated “diverse” had to be excluded from analyses that considered participant gender due to methodological reasons; however, they were included in all other analyses. All analyses were done in IBM SPSS 23.

4.3. Analysis of Aggregated Data

4.3.1. Warmth

H1.: 

Female social groups are perceived warmer than male social groups.

H2.: 

Social groups who use autonomous delivery vehicles (groups using ADV) are perceived as less warm than social groups not using ADV (main effect).

H2.1.: 

The decrease in warmth is stronger for female social groups than for their male counterparts (interaction).

The ANOVA (N = 397) with repeated-measures of the variable “warmth” and the within-subjects factors “gender of the user” and “usage of autonomous delivery vehicles (ADV)” and the between-subjects factor “participant gender” yield a main effect of “gender of the user”, F (1;395) = 64.22, p < 0.001, a main effect of “usage of autonomous delivery vehicles”, F(1;395) = 12.03, p < 0.001, and an interaction of these two factors, F(1;395) = 4.29, p < 0.039. (See Figure 4. Participant gender did not yield any effects, F = 0.001, p > 0.974.

• N = sample size
• F = F statistic/F value (degrees of freedom and residual degrees of freedom are in brackets)
• p = probability of error
• M = mean value
• SD = standard deviation

In line with our hypothesis H2.1, female group members using ADV (M = 3.45; SD = 0.62) were considered as less warm than female groups not using ADV (M = 3.58, SD = 0.63; F(1;405) =17.36; p < 0.001. Male group members using ADV (M = 3.30, SD = 0.61) were not considered to be less warm than male group members not using ADV (M = 3.34, SD = 0.58; F(1;406)= 1.00, p > 0.318 (see Figure 4).

Considering all groups using ADV, female group members possessed more warmth than male group members using ADV, F(1;406) = 21.31; p < 0.001. Regarding the group members not using ADV, female group members also were perceived as warmer than male group members, F(1;406) = 48.44, p < 0.001.

Thus, we could confirm our hypotheses H1, H2, and H2.1.

Figure 4. Perceptions of warmth for female and male groups using or not using ADV.

Figure 4. Perceptions of warmth for female and male groups using or not using ADV.

4.3.2. Competence

H3.:

Social group members that are described “using autonomous and self-steering delivery vehicles” (using ADV) are perceived as more competent than groups not using ADV/their basic versions (main effect).

The ANOVA (N = 397) with repeated-measures of the variable “competence” and the with-subjects factors “gender of the user“ and “usage of autonomous delivery vehicles (ADV)” and the between-subjects factor “participant gender” yield a main effect of “gender of the user”, F(1; 395)= 4.46, p < 0.035, a main effect of “usage of autonomous delivery vehicles”, F(1; 395) = 28.23, p = 0.000 (see Figure 5). There were no main effects of participant gender, F < 0.975, p > 0.324.

In line with our hypothesis H3, social group members gained in perceived competence when described using ADV compared to their version not using ADV/basic version; this was the case for female group members using ADV (M = 3.61, SD = 0.70) compared to female group members not using ADV (M = 3.47, SD = 0.61, F(1;405 = 11.63, p < 0.001) and was also true for male group members using ADV (M = 3.57, SD = 0.70) compared to male group members not using ADV, M = 3.39, SD = 0.64, F(1;406) = 19.32, p < 0.001.

Female group members (M = 3.54, SD = 0.54) were rated as slightly, albeit significantly, more competent than male group members. (M = 3.48, SD = 0.53, F(1;408) = 3.88, p < 0.050.)

4.3.3. Ingroup Favoritism

H4.: 

Participants show ingroup favoritism regarding the warmth and competence ratings of social groups of the same gender.

Warmth:

The aforementioned ANOVA (N = 397) with repeated-measures of the variable “warmth” and the with-subjects factors “gender of the user” and “usage of autonomous delivery vehicles (ADV)” and the between-subjects factor “participant gender” did not yield a significant interaction of “participant gender” and “gender of the user”, F < 2.89, p > 0.09.

Female participants (N = 199) rated the warmth of female group members at M = 3.54, SD = 0.53, and male participants (N = 199) rated female group members at M = 3.50, SD = 0.54. Regarding male group members’ warmth ratings, female participants rated male group members’ warmth at M = 3.30, SD = 0.51, and male participants rated it at M = 3.34, SD = 0.48.

Competence:

The aforementioned ANOVA (N = 397) with repeated-measures of the variable “competence” and the with-subjects factors “gender of the user” and “usage of autonomous delivery vehicles (ADV)” and the between-subjects factor “participant gender” did not show any significant effects, F < 3.35, p = 0.068.

Thus, we cannot confirm H4.

4.4. Analysis of the Influence of Gender and Autonomous Vehicle Use on the Perception of Different Social Groups

The results of the analysis of the influence of gender and the use of autonomous delivery vehicles on the perception of different social groups is described in the following sections.

4.4.1. Social Group 1: Women and Men

The MANOVA (N = 404) of the variables “warmth of social group 1” and “competence of social group 1” and the between-subjects factors “gender of social group members” and “usage of ADV” yield significant effects of “gender of the user”, F(2;399) = 25.04, p < 0.001, Wilk’s Λ = 0.888, partial η² = 0.112, “usage of ADV”, F(2;399) = 5.47, p < 0.005, Wilk’s Λ = 0.973, partial η² = 0.027, and an interaction of these two factors, F(2;399) = 7.79, p < 0.003, Wilk’s Λ = 0.971, partial η² = 0.029.

Warmth:

The between-subject factors “gender of the user” showed a significant effect on warmth ratings, F(1;400) = 45.00, p < 0.001, as did the between-subjects-factor “usage of ADV”, F(1;400) = 6.02, p < 0.015. Furthermore, the interaction between these two factors reached significance, F(1;400) = 7.79, p < 0.006.

Women (M = 3.71, SD = 0.71) were generally considered as warmer than men (M = 3.21, SD = 0.73); this was the case for “women not using ADV” compared to “men not using ADV”, F(1;400)= 45.57, p < 0.001), as well as for the group “women using ADV” compared to their analogue male groups, F(1;400)= 7.60, p < 0.006. These findings are in line with our first hypothesis (H1).

While women were seen as significantly warmer not using ADV (M = 3.89 SD = 0.61) than women described using ADV (M = 3.52 SD = 0.77, F = 13.67, p < 0.001), this difference did not exist for men’s ratings, F < 0.057, p > 0.811; see also Table 1 and Figure 6. These findings support H2.1.

Competence:

The ANOVA (N = 404) of the variable “competence of social group 1” and the between-subject factors “gender of social group members” and “usage of ADV” did not show any significant effects of these factors F < 0.308, p > 0.579.

4.4.2. Social Group 2: Physicians

The MANOVA (N = 404) of the variables “warmth of physicians” and “competence of physicians” and the between-subjects factors “gender of social group members” and “usage of ADV” did not yield any significant effects, F < 2.28, p > 0.103, see also Figure 7.

4.4.3. Social Group 3: Homemakers

The MANOVA (N = 405) of the variables “warmth of homemakers” and “competence of homemakers” and the between-subjects factors “gender of social group members” and “usage of ADV” yield significant effects of “usage of ADV”, F(2;400) = 20.18, p < 0.001, Wilk’s Λ = 0.908, partial η² = 0.092. No other effects were found, F < 0.634, p > 0.531.

Homemakers using ADV (M = 3.39; SD = 0.75) were considered as less warm than homemakers not using ADV (M = 3.73, SD= 0.75, F(1;401) = 20.58, p < 0.001).

Homemakers using ADV (M = 3.55, SD = 0.83) were also viewed as more competent than homemakers not using ADV (M = 3.38, SD = 0.79, F(1;401) = 4.18, p < 0.042), see also Figure 8.

4.4.4. Social Group 4: College Students

The MANOVA (N = 407) of the variables “warmth of college students” and “competence of college students” and the between-subjects factors “gender of social group members” and “usage of ADV” yield a significant effect of “usage of ADV”, F(2;402) = 4.81, p < 0.009, Wilk’s Λ = 0.977, partial η² = 0.023. No other effects were found, F < 1.73, p > 0.178.

However, regarding the effects of “usage of ADV” on warmth and competence perceptions, no significant effects were found, F < 3.71, p > 0.055, see also Figure 9.

4.4.5. Social Group 5: Business Women/Men

The MANOVA (N = 406) of the variables “warmth of business women/men” and “competence of business women/men” and the between-subjects factors “gender of social group members” and “usage of ADV” yield significant effects of “gender of the user”, F(2;401) = 10.21, p < 0.001, Wilk’s Λ = 0.952, partial η² = 0.048, “usage of ADV”, F(2;401) = 4.61, p < 0.010, Wilk’s Λ = 0.978, partial η² =0.022. No other effects were found, F < 1.99, p > 0.138.

Business women in general (M = 3.27, SD = 0.71) were perceived as warmer than business men in general (M = 2.95, SD = 0.71, F(1;402) = 20.15, p < 0.001.

Business men and women using ADV (M = 3.22, SD = 0.74) were considered as warmer than those not using ADV (M = 3.01, SD = 0.71, F(1;402) = 8.45, p = 0.004).

No effects of these factors on “competence of business women/men” were found, F < 2.04, p > 0.154, see also Figure 10.

4.4.6. Social Group 6: Unemployed Women/Men

The MANOVA (N = 403) of the variables “warmth of unemployed women/men” and “competence of unemployed women/men” and the between-subjects factors “gender of social group members” and “usage of ADV” yield significant effects of “gender of the user”, F(2;398)= 10.27, p < 0.001, Wilk’s Λ = 0.951, partial η² = 0.049 and “usage of ADV”, F(2;398) = 7.14, p < 0.001, Wilk’s Λ = 0.965, partial η² =0.035. No other effects were found, F < 6.75, p > 0.510.

Unemployed women in general (M = 3.23, SD = 0.78) were considered as warmer than unemployed men in general (M = 2.90, SD = 0.65, F(1;399) = 20.57, p < 0.001.

Unemployed women in general (M = 2.94, SD = 0.84) were seen as more competent than unemployed men in general, M = 2.73, SD = 0.79, F(1;399) = 6.55, p < 0.011.

Unemployed group members/people using ADV (M = 2.97, SD = 0.88) were seen as more competent than unemployed group members not using ADV (M = 2.70, SD = 0.74, F (1;399) = 10.59, p = 0.001), see also Figure 11.

4.4.7. Social Group 7: Women/Men with Handicaps

The MANOVA (N = 406) of the variables “warmth of women/men with handicaps” and “competence of women/men with handicaps” and the between-subjects factors “gender of social group members” and “usage of ADV” yield significant effects of “usage of ADV”, F(2;401) = 8.30, p < 0.001, Wilk’s Λ = 0.965, partial η² = 0.035. No other effects were found, F < 6.75, p > 0.510.

Persons with handicaps using ADV (M = 3.43, SD = 0.81) were seen as more competent than persons with handicaps not using ADV (M = 3.14, SD = 0.85, F(1;402) = 12.45, p < 0.001. No effects on warmth perceptions were found, F < 0.848, p > 0.358, see also Figure 12.

4.4.8. Social Group 8: Retired Women/Men

The MANOVA (N = 407) of the variables “warmth of retired women/men” and “competence of retired women/men” and the between-subjects factors “gender of social group members” and “usage of ADV” yield significant effects of “usage of ADV”, F(2;402) = 12.34, p < 0.001, Wilk’s Λ = 0.942, partial η² = 0.058. No other effects were found, F < 0.63, p > 0.534.

Retired men/women using ADV (M = 3.44, SD = 0.74) were considered less warm than retired men/women not using ADV (M = 3.58, SD = 0.72, F(1;403) = 3.78, p = 0.053).

Retired men/women using ADV (M = 3.36, SD = 0.86) were perceived as more competent than retired men/women not using ADV (M = 3.07, SD = 0.77, Welch (1;398.583) = 13.67, p < 0.001), see also Figure 13.

4.5. General Attitude towards Autonomous Delivery Vehicles

Additionally, we asked participants for their general attitudes on autonomous electrical self-steering vehicles for delivery using a five-point Likert scale. In total, 35.5% indicated a “positive”, 30.1% a “rather positive”, 23.5% a “mixed”, 8.1% a “rather negative”, and 1.7% a “negative” (1% missing) view of ADV.

5. Discussion

The main goal of the present study was to assess how usage of autonomous delivery systems is perceived in terms of the Stereotype Content Model (SCM) and how this association affects the stereotypical perception of a range of social groups. Therefore, we measured the stereotypical perceptions of eight social groups, each in a female and male version, described either with or without usage of autonomous and self-steering delivery vehicles (ADV). In doing so, our goal was twofold. First, we intended to determine the effects of the association that the use of ADV has on the perception of competence and warmth of the people who use them, whether these effects are different for female or male subgroups and whether the results were affected by the participants’ own gender. We used aggregated data to gain a broader understanding of stereotypical perceptions. Secondly, we assessed the effects of the usage of ADV for a range of different social groups in terms of the SCM, focusing on the intersection of group membership, gender, and ADV use.

We tested several hypotheses regarding gender differences and the effects of the association with autonomous delivery vehicles on the stereotypical perceptions of female and male members of social groups within the framework of the SCM.

We could confirm our first hypothesis and found that on a general level, female members of social groups are perceived as warmer than male members of social groups. Furthermore, and in line with our second hypothesis, members of social groups that are described using ADV are perceived as less warm than members of social groups in their basic versions (not using ADV). However, this effect was modified, due to the interaction of group gender and usage of ADV, as the decrease in warmth was only significant for the total of female members of social groups and not for the total of male members of social groups. None of the effects was linked to participant gender.

Regarding the competence dimension, we could confirm our third hypothesis and found that social group members described as “using autonomous and self-steering delivery vehicles” (using ADV) are generally perceived as more competent than group members not using ADV. Additionally, we found female group members received slightly higher competence ratings in general than their male counterparts. Effects of usage of ADV on the perception of competence were again not linked to participant gender. As participants did not show any ingroup favoritism regarding warmth and competence ratings of social group members of the same gender, we could not confirm our fourth hypothesis. A possible explanation for this phenomenon could be that gender was not salient enough to provoke ingroup effects as participants identified more with a certain occupation, which weakened the effect of gender on the ratings.

The overall higher warmth ratings for female groups replicate a vast body of previous studies ([21,22], including intercultural findings [59]. While these studies usually reported lower competence values for women, our study is mostly in line with findings of more recent studies, which also report a shift in competence ratings, as women are increasingly seen at least as competent as men [23,62,64] and in our case, as even more competent. However, though “using ADV” led to increased levels of supposed competence for males as well as females, this gain in competence was linked to a loss of warmth for female groups, but not for male groups. Thus, while the supposed competence of female groups generally seems to increase, a certain backlash effect occurs ([103,104]; i.e., the gain of competence is paired with a loss of warmth, as prescribed stereotypes are supposedly violated for female groups. In this case, we assume that autonomous delivery vehicles elicit a combination of agentic, competitive and possibly more “masculine”, technology-associated stereotypes [19,103,105], which are still partly incongruent with the stereotypical perceptions of females.

Whilst the stereotypical perception of people as users of certain devices varies between different groups, as indicated by Schwind et al. [4,11], we extended this principle to assessing different (sub)groups of potential users of autonomous delivery vehicles. Besides using aggregated data, we analyzed the effects of usage of ADV on the perceptions of eight different social groups, each in a female and in a male version of a group member. The findings show that while some effects were (nearly) universal, the different groups were indeed affected in distinctive manners: the ratings of “women” and “men” as distinct social groups replicated the general findings in many aspects, i.e., that women were rated as warmer than men; however, their association using ADV resulted again in a loss of warmth not pertinent for men. Competence ratings did not differ between them. Considering social technology acceptance processes, the alleged loss of warmth may discourage female users from approaching autonomous delivery vehicles. Assuming a bidirectional effect of users’ and devices’ stereotypically on perceptions (see [4]), this may perpetuate the effect, as the lesser use of ADV by females compared to male users may lead to continuous underrepresentation, and the association/congruency between female stereotypes and the stereotypicality of ADV may be weakened further.

Groups who were located approximately in the “admiration” quadrant of the SCM, such as physicians and college students, were more defined by their social groups than by their genders or use of ADV, as none of these two factors had any influence on their ratings. A certain halo effect of their group membership can be assumed [106]. Further, we argue that physicians are expected to use modern technology in their practice on a regular basis and college students are possibly perceived to be tech-savvy (either by participants who are college students or by the largely middle-aged participants); thus, the alleged use of ADV had no effect.

Groups who are usually seen as warm but incompetent and prompt piteous stereotypes—in this case, handicapped men and women and retired people [16,22,50]—both gained competence using ADV. However, unlike women and men with handicaps, retired people still lost warmth when using ADV. These outcomes support the notion that the social acceptability of a device depends on the assumed abilities of the user’s social group [4]. While both groups gain competence, only the group considered as less competitive does not lose warmth as paternalistic stereotypes were strongly activated for that group. Schwind found that devices used for medical purposes were usually viewed as less competitive and the (negative) effect on warmth for highly competitive devices such as LED glasses was often comparably mild for senior citizens (in contrast to more competent groups, such as physicians). These findings also match those of Profita [9,14], who found that smart glasses were more accepted when they were supporting people with disabilities.

Homemakers, a group which is often reduced to the female stereotype of “housewife” and used to have many overlaps with women in general [50], were seen as warm, but rather incompetent. In our study, both genders gained competence, but lost warmth through association using ADV. The role “homemaker” itself seems to provoke the stereotypical perceptions, presumably another halo effect. Also, male homemakers are not prototypical, which also could account for a less differentiated perception of male and female homemakers.

As described in the BIAS map of the SCM, business men and women are also viewed as competent but cold and are subject to envious stereotypes. Business women, however, were still considered to be warmer than business men. Interestingly, the results for using ADV contrast away from the results in the other groups: “using ADV” led to an increase in warmth for both business men and business women.

As expected, unemployed groups were assumed as incompetent and low in warmth [68]. Unemployed women were considered warmer and more competent than unemployed men. Using ADV led to an increase in competence ratings for both unemployed women and men. The lower competence ratings for men could be explained by the stronger societal pressures and expectations for men to have good jobs [107], and presumably the stronger gender role conflicts for men in case of unemployment [108].

We thus argue that autonomous delivery vehicles are seen as rather competitive devices, as their use is linked to an increase in competence in general and nearly throughout all social groups. Only groups already perceived as highly competent did not gain more competence by using ADV. Although there are a lot of parallel findings and tendencies between the aggregated data and the subgroup analyses, the detailed analyses still showed very specific patterns of stereotypical perceptions for the different social (sub)groups, sometimes resulting in inverse effects. Some groups were also affected more profoundly than others.

Based on our results and in line with [4], we assume that the social acceptance of devices, functioning as social objects, depends on the perception of the device itself combined with the stereotypicality of the group that is associated with it. With the present study, we extend this assumption to the perception of autonomous delivery vehicles, which are also stirring stereotypical perceptions that vary systematically with a range of different stereotypical user groups, often in intersection with gender. In terms of the SCM, the ability to pursue a goal depends on the combination of the autonomous vehicle, the social group and in many cases, the gender of the group members. Since the relationship between humans and technological devices are potentially bidirectional, the social acceptability of autonomous delivery vehicles may vary based on the assumed abilities of the user’s social group and assumed intentions.

6. Limitations and Future Work

Our treatment of gender is binary and therefore an oversimplification and exclusive (1.2% indicated diverse). In the present study, female and male participants did not differ in their stereotypical perceptions and ingroup favoritism could not be detected. Still, the limitations of such a categorization need to be considered, also regarding the binary treatment of the gender of the users. Further studies could include a “neutral form” in addition; however, it still needs to be considered that in German, seemingly “neutral” forms such participles as “Studierende” (meaning: people who are studying) usually have a slight male bias from a psycholinguistics perspective [73]. Regarding participants’ own group memberships, we asked for gender, age, level of education, but not their professions, as we consider the number of participants large (and heterogenous) enough to level out any potential small ingroup effects regarding occupations or social group memberships. However, it still needs to be considered that the participants lean towards an older demographic, as about half of the participants were 50 years or older.

The study was conducted in the south-western part of Germany in a region with a very strong automotive industry. This background may positively affect the perceptions of autonomous delivery vehicles in our sample compared to other regions and countries, where cars do not have the same status as in (Southern) Germany. We also need to consider that most likely, participants saw prototypes of autonomous and self-steering vehicles for delivery purposes presented in the context of the fair in the exhibition room and in the perimeter of the room where we conducted the study. Possible pre-selection effects and biases were reduced by actively asking people for participation outside the room. Besides, the theme of the room was not clearly visible from the outside and participants sometimes entered the room for unrelated reasons, e.g., seeking shelter from rain. Still, the context itself might have enhanced positive perceptions of autonomous delivery vehicles in general as well as induced slightly higher competence ratings for ADV users. On the other hand, as the prototypes were displayed, we can assume that all participants had a rather homogenous understanding of the concept “autonomous delivery vehicle”, Following on from this, we did not specify the interaction with autonomous delivery vehicles in our study, but it was clear to participants that the interaction works via an app and access to compartments and that no special input modes, such as gestures, would be necessary.

Furthermore, we note that, as autonomous delivery vehicles have slightly different uses (for delivery of goods, for shuttle-services, or for individual traffic), each one of these uses might influence the stereotypical perception of ADV; presumably an individual use will lead to more envious stereotypes than a transportation scenario, as used in our study.

In this sense, our study mainly focused on the perception of autonomous delivery vehicles as “devices”. While the use of artificially intelligent technology is implicit for their application, we did not explicitly address this dimension. As the use of AI in autonomous delivery vehicle needs to be considered also in terms of ethical dilemmas [109], future studies could take this a step further and address this aspect more, e.g., to investigate how transparency standards affect the perception of autonomous delivery vehicles within the framework of SCM. Further, when different brands enter the consumer market, this may affect the perception of ADV and their users at the intersection of stereotypical brand and technology perceptions.

Further work should also consider the diversity of user groups themselves. While we assessed different age groups, we did not systematically include them in our analyses, as this was not the focus of our study. But users of different generations often have different perspectives of technical devices. Moreover, despite an intercultural robustness of the principles of the SCMs and group perceptions [59], people from different cultural backgrounds may still diverge in their stereotypical perceptions of social (sub)groups and autonomous delivery vehicles. Finally, while positive attitudes are positively related to intentions to use a device [110] and social acceptability can be seen as a certain prerequisite to use a device, it still needs be stressed that there are also other factors that can hinder an interaction with or use of a device.

7. Conclusions

In conclusion, the present study examined how the usage of autonomous delivery vehicles (ADV) is perceived in terms of the Stereotype Content Model (SCM) and how being associated with an ADV affects the stereotypical perception of a range of social groups. Our study found that, on an aggregated level, female groups that are described as using autonomous and self-steering delivery vehicles are perceived as less warm than members of their social groups who are not using ADV. Overall, social group members described as using ADV are generally perceived as more competent than group members not using ADV. Moreover, the study revealed that while gender typically triggered ingroup effects related to warmth perceptions among various social groups, it was not salient enough to induce ingroup effects regarding physicians and college students. This suggests that these groups were defined more by their social professional identities than by their gender or use of ADV. Our results support previous findings on the suitability of the SCM as a framework for specifying the social acceptability of technological devices. These findings have important implications for the technology acceptance processes and suggest that the alleged loss of warmth may discourage female users from approaching autonomous delivery vehicles. Knowledge regarding the way that certain groups (e.g., female groups) are affected by a loss of warmth when using ADV could also to be put into consideration in further projects to foster use of ADV in urban areas, in marketing, and in further campaigns for sustainable mobility. Efforts should be made to promote the use of autonomous delivery vehicles by females, which could lead to a stronger association/congruency between female stereotypes and the use of ADV. In sum, our findings suggest that the combination of gender, vehicle usage, and group-specific stereotypical content needs to be considered to get an accurate and comprehensible evaluation of the social acceptability of innovations like autonomous delivery vehicles.