Isolated by Robotic Co-Workers: The Impact of Verbal Ostracism on Psychological Needs and Human Behavior

This study explores the impact of robotic ostracism by language on fundamental psychological needs and its carryover effect on subsequent human-human interaction. In a laboratory experiment, n = 68 participants experienced ostracism or inclusion by two co-present robots during a teamwork task. During ostracism, the robots communicated in an unknown language. Findings reveal that verbal ostracism by robots negatively influences basic psychological needs and fosters prosocial behavior towards other humans in later interactions. Implications for mixed human-robot teams in the workplace are discussed.


INTRODUCTION
With continuing technological development, robots are increasingly prevalent in various social contexts [2], including free time activities (e.g., restaurant visits) and work environments [7].
As humans are about to work with social robots in mixed teams, there is a growing need to explore whether such interactions have psychological consequences for humans.Furthermore, it is crucial to investigate whether experiences with these artificial partners can influence subsequent interactions with other human co-workers.
It has already been established that social psychological phenomena, such as social exclusion or ostracism also apply to HRI (e.g., [10,11]).Ostracism is defined as a situation where a person is ignored in their presence, which often occurs without an explanation or explicit negative attention [23].Baumeister and Leary [1995] established that humans have an urge to establish and maintain a minimum level of lasting, positive, and meaningful interpersonal relationships: the Need to Belong.This need is threatened by ostracism, making it an exceptionally aversive experience.This can be explained by the fact that exclusion from a group represented a life-threatening risk during human evolution that persists to this day [25].
Various ways exist to induce ostracism: In psychological research, the most popular paradigm is "Cyberball" because it is easy to regulate and displays very strong effects [27].In "Cyberball" exclusion takes place during a ball-tossing game, where one individual is excluded from receiving the ball.A sense of exclusion can also be readily triggered by the mere thought of anticipating a future life without fulfilling relationships [20,21] or when one is exposed to other people conversing in a foreign language in their presence, we call this "verbal ostracism" in the following [8,12].While it is unnecessary for robots to use natural language to communicate with each other, the rise of social robots has led to the widespread adoption of natural language in HRI.Nevertheless, when we think of mixed human-robot teams, situations may arise in which robots exchange information with other robots in the presence of humans without the use of natural language (because of efficiency).Most of the time this communication is silent and the person present is not directly aware of it [8].But what if humans can perceive the exchange of information yet struggle to comprehend the conveyed message?Could this lead to humans feeling excluded by their robotic co-workers?

Social Ostracism & Carryover Effects
Only subtle cues are needed to evoke feelings of ostracism, such as being left out of a ball-tossing game [25].Moreover, it has been demonstrated that feelings of ostracism can also be triggered when people converse in a language that a third person cannot understand [8].Experiencing ostracism has significant psychological consequences on an affective, cognitive and behavioral level: as it contributes to lowered mood and a diminished sense of overall well-being [22], along with a reduction in cognitive performance [5].It can also affect individuals' behavior, prompting both antisocial [13,14,19] and prosocial tendencies [3,25,26].According to the Reconnection Hypothesis [16] excluded people want to protect themselves from further exclusion, but at the same time strive towards new social relationships.The reaction to social ostracism, thus depends on the extent to which the excluded individual is likely to consider a social reconnection to others [3,17].If another person is considered a realistic source of social integration, showing prosocial behavior towards that person is more likely [16].These behavioral reactions to ostracism are not confined to the immediate interaction but can also impact subsequent interactions, known as "carryover effects" (e.g., [1,15]).Considering a future in which humans work in teams with robots, the question arises as to whether these effects are also observable.

Social Ostracism & Carryover Effects in HRI
Outside interpersonal encounters it has been shown that artifacts such as computers robots can cause feelings of ostracism too: Zadro et al. [2004] found that being excluded in "Cyberball" had a negative effect on basic needs satisfaction even when participants assumed that they were playing with a computer.Similar, Erel et al. [2022] showed that experiencing ostracism by two non-humanoid robots during "Cyberball" had a significant negative effect on participants' mood and basic psychological needs, such as belonging and control.Regarding verbal ostracism, Rosenthal-von der Pütten and Bock [2023] recently demonstrated that observing two robots conversing in a non-understandable way (i.e., beeps) in a videorecording resulted in feelings of ostracism and lowered trust, although participants were not involved in an actual interaction with the robots.Additionally, Erel et al. [2022] reported a carryover effect of being excluded by robots to a subsequent interaction with a human, where excluded participants showed increased proximity and prosocial behavior toward the experimenter.
Against this background, ostracism seems to be relevant in HRI, but, to our knowledge, only one video-based study has explored the role of verbal ostracism so far [18].Therefore, the present study examines verbal ostracism during live HRI within the context of a human-robot-robot team task.According to the recent state of research, we hypothesize: Participants who were verbally ostracized by robots report a higher need threat (H1) and show more prosocial behavior towards another human in a subsequent humanhuman interaction (H2) than people who were not verbally ostracized. 1

Experimental Design
We conducted a laboratory between-subjects experiment in which participants were asked to complete a teamwork task together with two humanoid robots (model: Pepper, Figure 1).The robots either verbally excluded (Ostracism condition) or included (Inclusion condition) the participant during the task.The assignment to the conditions was random while balancing gender between the groups.In the Ostracism condition the robots communicated with each other in an unknown language.Based on a pretest we chose Maltese text-to-speech which was not understandable to the participants.In the Inclusion condition, the robots talked to each other in German which all participants spoke fluently.For both conditions the robots' language was pre-scripted using the robots' own text-to-speech.Both robots were controlled by a person from a separate room (WOZ design).

Dependent Variables
Basic Needs.Since the Need to Belong [4] is threatened by ostracism, the effect of verbal ostracism on the satisfaction of the four basic needs (control, belonging, meaningful existence, selfesteem) was assessed with the Need Threat Scale (NTS) [22].This questionnaire consists of 20 items that measure the costs of ostracism and the benefits of inclusion on a 7-point Likert-type scale from "fully disagree" to "fully agree".
Prosocial behavior was operationalized by means of the number of questionnaires participants chose out of five to fill out voluntarily.This happened after the main study as a favor to the researcher.Given that the participants were not expected to get a reward, this can be seen as a purely prosocial act towards the researcher, which is known from related research on prosocial behavior (e.g., [9]).

Procedure
The study protocol was approved by the local IRB.When entering the laboratory, participants were first informed about the sequence before they signed informed consent.After completing a pre-interaction questionnaire, participants were led to the two robots who introduced themselves and the task.Meanwhile, the experimenter left the room to enable an undisturbed experience.The teamwork lasted for about 10 minutes and consisted of an allocation task among the three team members (two robots, one participant, Figure 1).The team had to plan the Christmas party of the company by assigning a list of tasks (blue cards, Figure 1) to the team members.A task is assigned when all members agreed with the allocation.During this phase of the experiment, the robots either talked in German (Inclusion condition) or in Maltese (Ostracism condition) with each other and in German with the participant (see Table 1 for an example dialog).
When the task was finished, participants were asked to fill out the Need Threat Scale.Then, the researcher led participants to another room for a brief interview.Afterwards, the experimenter thanked the participant for their attendance, and asked if they would be willing to fill out additional questionnaires for another project.It was emphasized that these questionnaires were unrelated to the present study, but that the participant's assistance in filling them out would help the researcher.The exact number of questionnaires that participants wanted to fill out as a favor for the researcher was documented.Finally, the researcher debriefed the participants and made sure they left with an overall positive experience.[Ostracism] Trid tieħu fuq il-kompitu?
[Inclusion] You can do the job.I think you're good at it, Pepper.

Pepper (turns back to participant):
We agree that I can take over the task.
Table 1.Example Dialog

Participants
In total, n = 68 persons (42 female), aged from 18 to 62 years (M = 26.18,SD = 8.23) participated in the study.Their education was in general quite high: The majority (35.3%) had a bachelor's degree or higher, 29.4% indicated a university entrance qualification, 29.4% had completed vocational training and 5.9% indicated a school graduation.Only seven participants had previous experience with humanoid robots, mostly in the context of other psychological studies or fares.

RESULTS
The analyses were conducted using SPSS Version 29.0.1.1.The reliability of the scales was determined using Cronbach's α.Values above .7 indicate a high internal consistency.

Need threat after verbal ostracism
To test our first hypothesis, mean scores were initially calculated for the subscales of the NTS (Belonging, α = .91,Control, α= .95;Meaningful Existence, α= .96;Self-esteem, α= .83).Afterwards, we conducted a one-factorial MANOVA with the two groups (Ostracism/ Inclusion) as fixed factor and the four subscales of the NTS as dependent variables.Overall, the MANOVA revealed a statistically significant difference between the experimental groups for the combined dependent variables, F(4, 63) = 30.25,p < .001,partial η² = .66,Wilk's Λ = .34.

Prosocial behavior after verbal ostracism
To test our second hypothesis, we conducted a linear regression analysis to see whether verbal ostracism predicts prosocial behavior towards the researcher after the actual interaction.For this purpose, the predictor was dummy-coded (0 = ostracism, 1 = inclusion).The analysis revealed a significant [F(1, 66) = 4.65, p < .05.] negative relationship between verbal ostracism and prosocial behavior (β = -.44):ostracized individuals showed a higher willingness to fill in additional questionnaires, in line with H2 (Table 2).However, only 6,6% of the variance in prosocial behavior was explained by verbal ostracism (R² = .066).

DISCUSSION
Given the notable increase of robots in work environments [7] and previous research on ostracism by robots [9,10,18], we were wondering whether humans might feel excluded by robotic coworkers, if they have to collaborate with more than one robot.In contrast to earlier work that investigated the mere observation of verbal exclusion [18], we test the live experiencing of verbal ostracism during HRI.Therefore, two robotic co-workers verbally ostracized a human participant by using an unknown language when conversing with each other in a laboratory study.The results demonstrate that robots can cause the feeling of ostracism and that verbal ostracism can furthermore lead to a carryover effect to a subsequent interaction with other humans in line with previous findings from HRI studies [9,10,18].

Psychological consequences for the individual
Initially, we confirmed that a short period of verbal ostracism by robots was sufficient to exert a strong negative impact on all four fundamental psychological needs (belonging, control, self-esteem, and meaningful existence) for the individual (H1).Experiencing exclusion during a collaborative task appeared to elicit intense aversive feelings, even though exclusion happened verbally, and participants were aware that they were interacting with robots.In addition, this result emphasizes that humans are highly susceptible to social exclusion, and this susceptibility may manifest in negative feelings due to need threats by robots in the workplace.

Behavioral consequences in subsequent interactions
Secondly, an increase in prosocial behavior towards the researcher following verbal ostracism by robotic co-workers was observed (H2), which underlines that robotic ostracism not only affects the ostracized individual but also carries over to subsequent interactions.In our experiment, participants tended to seek new social relationships after ostracism through showing prosocial behavior towards a person who is considered a realistic source of social integration (the researcher), in line with the Reconnection Hypothesis [16].This reveals that interactions with robots can subconsciously change the way humans interact with each other.Our observations are consistent with findings on the carryover effect in interpersonal ostracism research [3,25,26] and also recent work in HRI [9].While prosocial behavior has numerous positive attributes, it may not always be advantageous.When prosocial tendencies become strongly linked with the desire for social acceptance, they can make people more vulnerable to social manipulation and persuasive efforts [6,24].
As more and more robots become integrated into human workforces, it becomes important to consider not only the potential consequences for individuals, but also broader effects on the entire work environment.In other words, the impact that HRI with multiple robots might have on other interpersonal encounters at work must be considered, especially as antisocial behavior may also follow ostracism [13,14,19].In a positive light, robots in the workplace offer the opportunity to be designed in a way that strives to include all team members.This, in turn, can potentially have positive carryover effects on interpersonal integration.The latter should be investigated in future research.

Limitations and future directions
For the purpose of the present study, we chose the planning of a Christmas party as a simple allocation task participants are familiar with, and which relies mainly on communication.The ecological validity of such scenario for human-robot-robot teams in the near future can of course be questioned.However, we believe that the impact of verbal exclusion by robots is independent of the teamwork task and should thus be replicable in other conversational interaction scenarios.
With respect to our measure of prosocial behavior we cannot rule out the possibility that some participants were just curious about the content of the questionnaires and therefore indicated a high willingness to fill out many, regardless of prior ostracism.Nevertheless, this cannot explain the significant difference in prosocial behavior between the groups.Though, future studies can address this limitation by introducing a boring request, such as asking participants to file questionnaires.In addition, the limited explained variance in our analysis suggests that there may be other unconsidered variables that influence variations in prosocial behavior.These factors should be disentangled in future research.

CONCLUSION
Overall, the results from our experimental study demonstrate that verbal ostracism by robotic co-workers is psychologically meaningful and consequential to humans.Moreover, our findings highlight the importance of carefully designing even the communication between robots when these are intended to collaborate with humans.This raises new questions regarding when robot-to-robot communication should be transparent and how such a communication can be designed in detail.

Figure 1 .
Figure 1.Participant discussing the allocation of tasks with the robots Salt (l.) and Pepper (r.) in the laboratory