While regulatory frameworks call for the implementation of AI certifications, empirical knowledge about how such certifications affect interactions is still scarce. In this work, we examined how AI certifications affect users’ trust and reliance. In addition, we examined whether certifications elevate user expectations and whether unmet expectations subsequently reduce trust. In a 2 (certification vs no certification) x 2 (reliability: high vs low) between-subjects online study, N = 644 participants had to identify bacterial infestation in pictures with the help of an AI. Our results show that, before interacting with the AI, participants trusted the certified system more and showed reduced vigilance. However, these effects disappeared post-interaction, where, instead of the certification, system reliability significantly affected trust and vigilance. Notably, certifications did not raise expectations per se, but instead amplified the impact of system reliability on user trust. Additional exploratory results showed that the certification supported appropriate reliance.

With the deployment of AI systems across multiple domains, understanding how users develop trust has becomecrucial for successful implementation. This study investigates how different AI features influence the decision to use an AI system and which characteristics users prioritise when evaluating them. We focus on whether users prefer systems whose functioning they can understand or whose trustworthiness is certified. We examined whether users favour system transparency through explainability features or rely more on external trust signals, such as AI certification seals, while considering how these preferences interact with technical reliability and fairness. Using conjoint analysis, we systematically compared the influence of four key attributes (transparency by explainability features, technical reliability, external trust signals through AI certifications, and fairness) on user decisions to use an AI system. Through cluster analysis, we identified two groups with opposing preferences and demographic differences. The first group prioritised high explainability and strong AI certification while showing negative preferences for fairness, whereas the second group favoured fairness and reliability while displaying negative attitudes toward explainability and AI certification. These contrasting prioritisation patterns raise important questions about AI systems development, particularly regarding challenges of addressing competing user requirements for trust-related features.

Users often mindlessly treat machines such as voice assistants (VAs) like social beings, speaking politely despite VAs functioning as commandable tools that do not require polite expressions. In fact, unnecessary words can even impair interactions, causing frustration. This contrast between social and functional interactions is especially relevant in family homes with children, where VAs are often integrated long-term. Since children are still learning social communication rules, they might adopt negatively connotated communication patterns with the VA, such as a lack of politeness, commanding phrasing, or verbal abuse. Over 2.5 years, 128 parents therefore completed six online surveys about their own and their children’s communication behaviour towards and relationship with their VA, as well as their own social perception of and satisfaction with the VA. Multilevel analyses revealed that parents considered themselves to be more polite to the VA than their children. Perceived relationship closeness between children and VA positively predicted children’s commanding, politeness, and verbal abuse tendencies towards the VA over time, whereas for the parents, relationship closeness only influenced politeness. The findings highlight the potential impact of social perceptions on long-term communication with VAs and tackle the concern of VA’s negative impact on human-human communication.

People increasingly encounter AI-powered chatbots in their daily lives which often make use of anthropomorphic cues -potentially with the intention to trick users into trusting the system.This study investigates the impact of anthropomorphic design cues on user perception and behavioral intention in human-chatbot interactions focusing on the social bot My AI within Snapchat.In a between-subjects experimental design (N=323) participants interacted with either an anthropomorphic version of the chatbot, featuring human-like cues such as personalized greetings, use of emoticons, and a human-like appearance, or a less anthropomorphic version minimizing these cues.The findings challenge prior assumptions by revealing that anthropomorphic cues neither influenced the intention to use the chatbot nor the trust in it.However, perceived anthropomorphism was found to enhance trust in the chatbot, which in turn positively influenced use intention.Additionally, users who perceived the chatbot as more human-like were more likely to view it as a friend, a perception that further mediated the effect of anthropomorphism on trust.These results suggest that while anthropomorphic design does not universally increase trust and use intention directly, perceived anthropomorphism and perception as a friend plays an important role in user interactions with AI-based chatbots.Future studies will need to clarify whether the design decision to frame the interaction as dialogue might be sufficient to increase perceived anthropomorphism and its related effects.

Users of digital technologies are still held legally responsible for managing their personal data.However, not all users have the same opportunities for privacy management as researchers find systematic sociodemographic differences in, for example, users' privacy literacy and self-efficacy two important predictors of privacy management.The present study argues that one reason for such digital inequalities is how easily people have access to privacy information in their everyday lives (e.g. through their profession or social contacts).Analysing data of a representative sample of German internet users (N = 3978) by means of a Bayesian structural equation model, we find that men and more educated persons report easier access to privacy information which, in turn, positively relates to both privacy literacy and self-efficacy.Older persons feel less confident about privacy protection while men and more highly educated persons have higher privacy literacy levels.In conclusion, the present study reveals that casual access to privacy information can be important in learning and being confident about privacy management, but it cannot fully explain digital inequalities in online privacy self-efficacy and literacy.

Generative AI systems such as ChatGPT are increasingly used to assist with tasks or to receive information. Given that the systems are not perfectly reliable regarding the content they produce, human users need to carefully navigate to which degree they can place trust in the system (calibrated trust). However, based on media equation assumptions, it can be hypothesized that social cues displayed by the system might instill more trust than is warranted. Against this background, the present study investigates in a 2 × 2 between subject design ( N = 617) whether the social cues “typing behavior” and “personalized address” used by ChatGPT increase perceived trust (benevolence, ability, and integrity) in the system and whether this is mediated by perceived similarity and moderated by anthropomorphism inclination. The results show that the social cue “typing behavior” leads to a significant increase in trust in the dimension benevolence. Neither perceived similarity nor anthropomorphism inclination modulates this effect. However, as a side effect, perceived similarity was found to significantly predict trust in ChatGPT.

Trust has been recognized as a central variable to explain the resistance to using automated systems (under-trust) and the overreliance on automated systems (over-trust). To achieve appropriate reliance, users’ trust should be calibrated to reflect a system’s capabilities. Studies from various disciplines have examined different interventions to attain such trust calibration. Based on a literature body of 1000+ papers, we identified 96 relevant publications which aimed to calibrate users’ trust in automated systems. To provide an in-depth overview of the state-of-the-art, we reviewed and summarized measurements of the trust calibration, interventions, and results of these efforts. For the numerous promising calibration interventions, we extract common design choices and structure these into four dimensions of trust calibration interventions to guide future studies. Our findings indicate that the measurement of the trust calibration often limits the interpretation of the effects of different interventions. We suggest future directions for this problem.

Despite the rising prevalence of intelligent voice assistants in people's homes, it remains opaque to most users how they function. With the overall goal to foster informed usage and responsible handling of personal data, the user's understanding increasingly moves into the focus of research. Particularly in the context of black box technologies such as voice assistants, individuals build intuitive mental models (also called folk theories) which can act as a guide for how they understand and interact with such systems. To shed light on individuals' mental models regarding intelligent voice assistants, we applied a visual elicitation method during a citizen science workshop: 26 participants were asked to visualize how they imagine a voice assistant to function. The resulting drawings were categorized based on their theorization complexity level. While basic awareness of the overall functionality (user input -- something happens -- voice assistant output) was found in every drawing, 13 participants revealed mental models on how the input is received (e.g., audio recording and speech processing), six on how the information is then processed (e.g., database requests and data categorization), and five on how the output is generated (e.g., speech synthesis). Overall, the results underline the need to address individuals' gaps in understanding of intelligent technology that is already in widespread use.

Algorithmic news curation determines users’ news exposure in the online environment. Despite its usefulness, it also comes along with the problem of algorithmic opacity. To combat this, explainable algorithmic news curation systems are necessary. One user-centered solution to design these systems can be achieved through the systematic exploration of user folk theories. For this, we conducted twelve in-depth semi-structured interviews to explore (1) the user preferences for explainable system design, and (2) folk theories of algorithmic news curation. By applying qualitative content analysis, we found a psychological trade-off between the desire for transparency and feelings of creepiness, thus a preference for explanations to be hidden. Furthermore, we identified eight assumptions of folk theories. The results are compared to previous folk theories and discussed in terms of the ‘sweet spot’ of system transparency. We conclude that exploring folk theories is a key requirement for designing explainable algorithmic news curation systems.

In this paper, we investigate the human ability to distinguish political social bots from humans on Twitter. Following motivated reasoning theory from social and cognitive psychology, our central hypothesis is that especially those accounts which are opinion-incongruent are perceived as social bot accounts when the account is ambiguous about its nature. We also hypothesize that credibility ratings mediate this relationship. We asked N = 151 participants to evaluate 24 Twitter accounts and decide whether the accounts were humans or social bots. Findings support our motivated reasoning hypothesis for a sub-group of Twitter users (those who are more familiar with Twitter): Accounts that are opinion-incongruent are evaluated as relatively more bot-like than accounts that are opinion-congruent. Moreover, it does not matter whether the account is clearly social bot or human or ambiguous about its nature. This was mediated by perceived credibility in the sense that congruent profiles were evaluated to be more credible resulting in lower perceptions as bots.

Children are raised with technologies that are able to respond to them in natural language. This not only makes it easy to communicate but also to connect with them socially. While communication abilities might have benefits (e.g., for learning), it might also raise concerns among parents as the technologies are not necessarily designed to facilitate the children's social, emotional, and cognitive developments and serve as a model for the construction of a social world among humans. First technologies children can talk to differ in their embodiment (e.g., robots and voice assistants), which could affect central variables, such as social presence, trust, and privacy concerns. The present study aimed to investigate how parents conceptualize socially appropriate interactions between children and technologies. The results underline the parents' emphasis on embodiment and privacy protection. The study underlines the importance of incorporating the parental perspective to meet the expectations of responsible interactions between children and technologies.

Mobile technologies offer new opportunities for the creation and development of systems for crisis communication and management. Vital to the success of using smartphone-based technology for volunteers is to determine whether privacy issues affect user participation. Thus, an online experiment (N = 217) addressing volunteers or people interested in volunteering as well as people with medical qualifications was conducted to investigate the influence of privacy-related data control options on user acceptance of a smartphone-based system to facilitate volunteer engagement in crisis situations. Building on agency models and control theories in the context of information privacy, the results show that higher control options led to a more positive evaluation of the app, which was mediated by perceived privacy concerns. Among people with a high career motivation, higher control also led to a higher willingness to actively participate as volunteer in the system.

During interactions with others, people have expectancies concerning their communication partners' behaviors. Negative violations of these expectancies are known to exert an adverse impact on people's perceptions and attitudes. Likewise, with regard to human-robot interactions, previous research on one-time interactions indicates that people's evaluations of robots can deteriorate equally. To simultaneously investigate possible compensation for this effect, the concept of idiosyncrasy credit was tested for transferability to human-robot interactions. It postulates that compensation can be achieved by an individual's consistent, beneficial behavior over time that results in a social deviation credit. During an experimental medium-term study with a 2x2 between-subjects design, 80 participants interacted with a humanoid, social robot in three time-separated sessions distributed over an average time span of ten days. The robot's acquisition of idiosyncrasy credit was manipulated as well as the occurrence of a negative expectancy violation by systematically varying the robot's polite communication style. Results of repeated measures ANOVAs support the assumption that negative expectancy violations decrease people's evaluations of a robot. However, this effect seems non-persistent over time. Furthermore, no support for the robot's acquisition of idiosyncrasy credit was gained.

It has long been observed that people treat technical entities like computers as social actors. Numerous studies have been conducted on media equation theory, which describes users' social reactions toward devices to be unconscious. These users consider this inappropriate when thought about consciously. However, both long-term trends as well as the reactions of children have been neglected - although children might indeed differ in their reactions, as they are especially prone to the tendency to humanize and anthropomorphize. This study examined the social behavior of 20 children aged six to twelve toward a voice assistant during a five-week field study. In the course of three sessions, interviews were conducted to assess the children's tendency to socially react to the devices and to what extent they perceive such reactions as appropriate. Additionally, the tendency to anthropomorphize was assessed together with qualitative data to gain insight into cues that would trigger anthropomorphization. Results indicate that age and duration have an effect on the tendency to anthropomorphize. However, age and duration have only a partial effect on social reactions and the evaluation of their appropriateness. Furthermore, natural speech appeared to be a rather strong cue in triggering tendencies to anthropomorphize in the children.

Research in social science distinguishes between two types of social influence: informational and normative. Informational social influence is driven by the desire to evaluate ambiguous situations correctly, whereas normative social influence is driven by the desire to be liked and gain social acceptance from another person. Although we know from research that humans can effectively use either of these techniques to persuade other humans, scholars have yet to examine the relative effectiveness of informational versus normative social influence when used by virtual agents. We report a study in which users interact with a system that persuades them either using informational or normative social influence. Furthermore, to compare agents to human interlocutors, users are told that the system is either tele-operated by a human (avatar) or fully-automated (agent). Using this design, we are able to compare the effectiveness of virtual agents (vs humans) in employing informational versus normative social influence. Participants interacted with the system, which employed a Wizard-of-Oz operated virtual agent that tried to persuade the user to agree with its rankings on a "survival task." Controlling for initial divergence in rankings between user and the agent, there was a significant main effect such that informational social influence resulted in greater influence than normative influence. However, this was qualified by an interaction that approached significance; users were, if anything, more persuaded by informational influence when they believe the agent was AI (compared to a human), whereas there was no difference between the agent and avatar in the normative influence condition.

The present study investigated the role of age in the perception of emotional nonverbal behaviors of a virtual assistant in a 2 (seniors vs young participants) x 2 (happy vs sad situations) x 4 (three emotional nonverbal behaviors related to each situation vs neutral behavior) mixed factorial design. In the study, a virtual agent acted as an assistant to review the imaginary monthly schedule of the participant. After uttering each schedule, the agent showed different emotional nonverbal behaviors to express empathy. The stimulus materials were presented to 60 participants (30 elderly people and 30 younger adults) in 25 videos. All participants had to rate the agent's friendliness, intelligence, empathy, trustworthiness, and helpfulness immediately after watching each video. The results indicated that in the presence of the emotional nonverbal behaviors the elderly rated the agent as more empathic and trustworthy compared to the younger adults. The data also revealed that there were differences with respect to rating specific emotional nonverbal behaviors as more empathic than others. Elderly people perceived Dropping the Arms plus Sad Face, Head down, Sad Face, Head Nod plus Smile, and Smile as more empathic, and the nonverbal behaviors perceived as most empathic for the younger adults were Dropping the Arms plus Sad Face, and Head down.

Research has demonstrated promising benefits of applying virtual trainers to promote physical fitness. The current study investigated the value of virtual agents in the context of personal fitness, compared to trainers with greater levels of perceived agency (avatar or live human). We also explored the possibility that the effectiveness of the virtual trainer might depend on the affective tone it uses when trying to motivate users. Accordingly, participants received either positively or negatively valenced motivational messages from a virtual human they believed to be either an agent or an avatar, or they received the messages from a human instructor via skype. Both self-report and physiological data were collected. Like in-person coaches, the live human trainer who used negatively valenced messages were well-regarded; however, when the agent or avatar used negatively valenced messages, participants responded more poorly than when they used positively valenced ones. Perceived agency also affected rapport: compared to the agent, users felt more rapport with the live human trainer or the avatar. Regardless of trainer type, they also felt more rapport - and said they put in more effort - with trainers that used positively valenced messages than those that used negatively valenced ones. However, in reality, they put in more physical effort (as measured by heart rate) when trainers employed the more negatively valenced affective tone. We discuss implications for human--computer interaction.

In this paper we present a system for all and sundry to create robotic applications which is characterized by adaptivity on different levels. First, users are enabled to create their own applications (e.g. play a card game with robot) by the usage of a decision making framework for robot companions. Second, within a created application the robot itself adapts its behavior via user feedback based on a decision making algorithm (e.g. teach the robot a card game based on user feedback). In dependency of the user's expertise, he or she has different possibilities of enhancing the robot's capabilities. Especially for the non-expert user a tool has been developed which provides a graphical user interface to configure applications. The usability of the tool has been evaluated with 5 participants of the age group 40+. The results show that the technical requirements to include non-experts are fulfilled by the framework but also reveal ways to improve the tool, such as the placement of the assistance mechanisms offered to the user.

In open online learning courses such as MOOCs, peer feedback has been regarded as a powerful method to give elaborated feedback on weekly assignments. Yet motivating students to invest effort in peer feedback on top of existing work load is difficult. Students might give insufficient feedback or do not give feedback at all. Students' hesitation to provide feedback might be related to the lack of visibility of spent effort during feedback provision. Alternatively, students might provide less feedback due to lack of perceived benefits. In this study, we investigated the effect of two incentive types on peer feedback provision on weekly assignments. In total, 91 students enrolled in a semi-open online course were announced to receive either (1) a peer rating on their feedback or (2) open access to assignment solutions or (3) no incentive. Results indicate that the incentive type did not affect feedback provision in general, yet it had an impact on the content of the feedback. Students receiving (1) a rating-feedback incentive wrote longer and more specific feedback in comparison to students receiving (2) an information-access incentive or (3) no incentive. Results contribute to findings from peer assessment research that students are more likely to provide detailed feedback if students feel that feedback is attended to. Furthermore, results inform teachers and practitioners on how to encourage students to provide peer feedback in open learning environments.

BinCam is a social persuasive system to motivate reflection and behavioral change in the food waste and recycling habits of young adults. The system replaces an existing kitchen refuse bin and automatically logs disposed of items through digital images captured by a smart phone installed on the underside of the bin lid. Captured images are uploaded to a BinCam application on Facebook where they can be explored by all users of the BinCam system. Engagement with BinCam is designed to fit into the existing structure of users' everyday life, with the intention that reflection on waste and recycling becomes a playful and shared group activity. Results of a user study reveal an increase in both users' awareness of, and reflection about, their waste management and their motivation to improve their waste-related skills. With BinCam, we also explore informational and normative social influences as a source of change (e.g., socially evoked feelings of 'guilt' for non-recycling or food disposal), which has to date been underexplored in persuasive HCI. Design implications for reflection and social persuasion are proposed.

Against the background of the uncanny valley hypothesis [1] and its conceptual shortcomings this study aims at identifying design characteristics which determine the evaluation of robots. We conducted a web-based survey with standardized pictures of 40 robots which were evaluated by 151 participants. A cluster analysis revealed six clusters of robots. The results are discussed with regard to implications for the uncanny valley hypothesis.

Presence is usually assessed via a variety of subjective and objective measures. However, constraints often result in subjective measurements using questionnaires as a key method of data collection. In this paper we present a study of 44 participants of a collaborative augmented reality game known as TimeWarp which used both subjective and objective behavioral measures. Behavior as coded from video recordings of one scene of the game and self-reports about feelings of presence were compared. Our findings indicate that pointing behavior and verbal responses to the virtual content are correlated negatively to sense of presence. We further investigated the influence of subjectively perceived interactivity on perceived presence. We found that the interaction possibilities perceived by the participants predicted their experience of social presence with the virtual characters in the game. Furthermore, playing together with another person did not result in decreased social presence of the virtual characters. Implications for presence research are discussed.

Since most robots are not easily displayable in real-life scenarios, only a few studies investigate users' behavior towards humanoids or androids in a natural environment. We present an observational field study and data on unscripted interactions between humans and the android robot "Geminoid HI-1". First results show that almost half of the subjects mistook Geminoid HI-1 for a human. Even those who recognized the android as a robot rather showed interest than negative emotions and explored the robots capabilities.

The EU project SERA (Social Engagement with Robots and Agents) provided the unique opportunity to collect real field data of people interacting with a robot companion in their homes. In the course of three iterations, altogether six elderly participants took part. Following a multi-methodological approach, the continuous quantitative and qualitative description of user behavior on a very fine-grained level gave us insights into when and how people interacted with the robot companion. Post-trial semi-structured interviews explored how the users perceived the companion and revealed their attitudes. Based on this large data set, conclusions can be drawn on whether people show signs of bonding and how their relation to the robot develops over time. Results indicate large inter-individual differences with regard to interaction behavior and attitudes. Implications for research on companions are discussed.