Older adults have increasingly turned to conversational AI as a source of emotional support. However, little is known about how emotionally supportive interactions are experienced in everyday use, particularly when AI systems limit, redirect, or intervene during these interactions. We interviewed 18 older adults about their experiences using conversational AI for emotional support, examining when they turn to AI, how they engage during emotionally vulnerable moments, and how they respond when support feels disrupted. Our findings show that older adults often rely on AI when other forms of social support feel inaccessible. However, current safety-related interventions can redirect interactions in ways that participants experience as interruptions to emotional engagement or as shifts in control away from them. Such disruptions can undermine older adults’ ability to remain emotionally engaged and, in some cases, contribute to emotional distress. We discussed design implications for emotionally supportive conversational AI, emphasizing the need for safety interventions that are enacted within older adults’ social contexts, align with users’ emotional pacing, and preserve their sense of agency.

Intergenerational conversations often break down when differences in tone, language, or expectations lead participants to feel dismissed or misunderstood. In this work, we explore how people envision AI-driven chatbot interventions for addressing communication problems in text-based intergenerational family chat. We conducted a scenario-based design interview with 10 pairs of family members from different generations, in which participants designed chatbot interventions that varied in intervention target and timing. Our findings show that participants expect chatbots to perform multiple themes of intervention, including mediating understanding, providing emotional support, offering evaluative commentary, and guiding interaction through behavioral suggestions. These expectations varied systematically across intervention contexts, giving rise to distinct chatbot roles such as neutral mediators, message coaches, repair facilitators, and emotion regulators. Across these roles, participants positioned chatbots as moral advisors that evaluate communicative appropriateness and exercise varying degrees of moral authority. Rather than prescribing specific system behaviors, this work offers a conceptual and exploratory account of AI-mediated intervention in intergenerational communication, and articulates key design tensions that arise when chatbots are imagined as socially and morally involved actors in intimate family interactions.

The integration of Conversational Agents (CAs) into daily life offers opportunities to tackle global challenges, leading to the emergence of Conversational AI for Social Good (CAI4SG). This paper examines the advancements of CAI4SG using a role-based framework that categorizes systems according to their AI autonomy and emotional engagement. This framework emphasizes the importance of considering the role of CAs in social good contexts, such as serving as empathetic supporters in mental health or functioning as assistants for accessibility. Additionally, exploring the deployment of CAs in various roles raises unique challenges, including algorithmic bias, data privacy, and potential socio-technical harms. These issues can differ based on the CA's role and level of engagement. This paper provides an overview of the current landscape, offering a role-based understanding that can guide future research and design aimed at the equitable, ethical, and effective development of CAI4SG.

This study investigates how Southeast Asian (SEA) immigrant mothers in Taiwan participate in their children’s home-based learning. Drawing on semi-structured interviews and diary studies, we explore how these mothers navigate sociocultural constraints while fostering engagement and transmitting cultural values. Despite facing diminished agency and structural marginalization, mothers engage creatively in their children’s everyday learning interactions. Guided by a justice-oriented lens, we identify various harms and propose design implications for socio-technical systems that center recognition, reciprocity, and accountability in parent-child learning at the individual, familial, and societal levels. Our contribution lies in foregrounding the role of intersectional identity in parent-child learning and proposing justice-oriented design directions that support the flourishing of immigrant mothers within socio-technical systems.

Recent Large Language Model (LLM) based AI can exhibit recognizable and measurable personality traits during conversations to improve user experience. However, as human understandings of their personality traits can be affected by their interaction partners’ traits, a potential risk is that AI traits may shape and bias users’ self-concept of their own traits. To explore the possibility, we conducted a randomized behavioral experiment. Our results indicate that after conversations about personal topics with an LLM-based AI chatbot using GPT-4o default personality traits, users’ self-concepts aligned with the AI’s measured personality traits. The longer the conversation, the greater the alignment. This alignment led to increased homogeneity in self-concepts among users. We also observed that the degree of self-concept alignment was positively associated with users’ conversation enjoyment. Our findings uncover how AI personality traits can shape users’ self-concepts through human-AI conversation, highlighting both risks and opportunities. We provide important design implications for developing more responsible and ethical AI systems.

Non-native speakers (NNSs) face significant language barriers in multilingual communication with native speakers (NSs). While AI-mediated communication (AIMC) tools offer efficient one-time assistance, they often overlook opportunities for NNSs’ continuous language acquisition. We introduce ChatLearn, an enhanced AIMC system that leverages NNSs’ communication difficulties as learning opportunities. Beyond comprehension and expression assistance, ChatLearn simultaneously captures NNSs’ language challenges, and subsequently provides them with spaced review as the conversation progresses. We conducted a mixed-methods study using a communication task with 43 NNS-NS pairs, after which ChatLearn NNSs recalled significantly more expressions than the baseline group, while there was no substantial decline in communication experience. Our findings highlight the value of contextual learning in NNS-NS communication, providing a new direction for AIMC systems that foster both immediate collaboration and continuous language development.

AI-assisted writing raises concerns about autonomy and ownership when benefiting writers. Personalization has been proposed as an effective solution while also risking writers’ reliance on AI and behavior shifting. For better personalization design, existing studies rely on interaction and information solely within the writing phase; however, few studies have examined how reading behaviors can inform personalized writing. This study investigates the effects of integrating reading highlights for personalization on AI-assisted writing. A between-subjects study with 46 participants revealed that the personalization condition encouraged participants to produce more highlights. However, highlighting unexpectedly shifted from a sense-making strategy to an instrumental act of "feeding the AI," leading to significant reliance on AI and declines in writers’ sense of autonomy, ownership, and self-credit. These findings indicate personalization risks in AI-assisted writing, emphasize the importance of personalization strategies, and provide design implications.

AI-generated influencers are rapidly gaining popularity on Chinese short-video platforms, often adopting kinship-based roles such as “AI grandchildren” to attract older adults. Although this trend has raised public concern, little is known about the design strategies behind these influencers, how older adults experience them, and the benefits and risks involved. In this study, we combined social media analysis with interviews to unpack the above questions. Our findings show that influencers use both visual and conversational cues to enact kinship roles, prompting audiences to engage in kinship-based role-play. Interviews further show that these cues arouse emotional resonance, help fulfill older adults’ informational and emotional needs, while also raising concerns about emotional displacement and unequal emotional investment. We highlight the complex relationship between virtual avatars and real family ties, shaped by broader sociocultural norms, and discuss how AI might strengthen social support for older adults while mitigating risks within cultural contexts.

Mental health chatbots are increasingly deployed as scalable interventions, yet the relational mechanisms underpinning their effectiveness remain unclear. Drawing on prior research on digital therapeutic alliance, we operationalized a preliminary multi-dimensional instrument to capture perceptions of relational and functional dynamics in mental health chatbot interactions and conducted a four-week within-subjects study with 56 participants engaging with Wysa and Youper (two widely used CBT-based mental health chatbots). Through iterative factor refinement and regression modeling, we found that user-chatbot relationship formation is primarily driven by two factors: an affective factor, centered on emotional support, and a goal-oriented factor, centered on practical assistance. Conversational control contributed alongside these interpersonal factors, while trust (privacy, non-judgmentalness) and satisfaction emerged as correlated outcomes of supportive, effective interactions rather than standalone predictors. These findings advance models of the Digital Therapeutic Alliance by clarifying its underlying structure and highlighting design priorities for balancing empathy and efficacy in conversational agents.

Conversational agents are increasingly used in education for learning support. An application is “learning by explaining”, where learners explain their understanding to an agent. However, existing research focuses on single roles, leaving it unclear how different pedagogical roles influence learners’ interaction patterns, learning outcomes and experiences. We conducted a between-subjects study (N=96) comparing agents with three pedagogical roles (Tutee, Peer, Challenger) and a control condition while learning an economics concept. We found that different pedagogical roles shaped learning dynamics, including interaction patterns and experiences. Specifically, the Tutee agent elicited the most cognitive investment but led to high pressure. The Peer agent fostered high absorption and interest through collaborative dialogue. The Challenger agent promoted cognitive and metacognitive acts, enhancing critical thinking with moderate pressure. The findings highlight how agent roles shape different learning dynamics, guiding the design of educational agents tailored to specific pedagogical goals and learning phases.

In difficult workplace email conversations, such as layoffs or resource negotiations, the absence of nonverbal cues can exacerbate negative emotions experienced by recipients. While existing tools support senders in refining tone, there is little support for processing emotionally intensive content from the receivers’ side. This study investigated the use of large language models that added positive or neutral reframings, written in either first or third person, to original emails, with the aim of helping recipients view difficult conversations in a different light. In a controlled study with 132 participants, positive reframing reduced receivers’ negative emotions and was rated as more helpful than neutral reframing, regardless of narrative perspective. Although reframing type did not significantly change conflict management behaviors, positive reframing led to fewer power-related words in interpretations of the email. These findings highlight opportunities and challenges for designing AI as a social buffer to facilitate difficult conversations online.

Existing conversational search systems can synthesize information into responses, but they lack principled ways to adapt response formats to users’ cognitive states. This paper investigates whether aligning format and distance, which involves matching information granularity and media to users’ psychological distance, improves user experience. In a between-subjects experiment (N = 464) on travel planning, we crossed two distance dimensions (temporal/spatial × near/far) with four formats varying in granularity (abstract/concrete) and media (text/image-and-text). The experiment established that format–distance alignment reduced users’ risk perceptions while increasing decision confidence, perceptions of information usefulness, ease of use, enjoyment, and credibility, and adoption intentions. Concrete formats imposed higher cognitive load, but yielded productive effort when matched to near-distance tasks. Images enhanced concrete but not abstract text, suggesting multimedia benefits depend on complementarity. These findings establish format–distance alignment as a distinctive and important design dimension, enabling systems to tailor response formats to users’ psychological distance.

Qualitative coding is a demanding yet crucial research method in the field of Human–Computer Interaction (HCI). While recent studies have shown the capability of Large Language Models (LLMs) to perform qualitative coding within theoretical frameworks, their potential for collaborative human-LLM discovery and generation of new insights beyond initial theory remains underexplored. To bridge this gap, we proposed CHALET , a novel approach that harnesses the power of human-LLM partnership to advance theory-driven qualitative analysis by facilitating iterative coding, disagreement analysis, and conceptualization of qualitative data. We demonstrated CHALET ’s utility by applying it to the qualitative analysis of conversations related to mental-illness stigma, using the attribution model as the theoretical framework. Results highlighted the unique contribution of human-LLM collaboration in uncovering latent themes of stigma across the cognitive, emotional, and behavioral dimensions. We discuss the methodological implications of the human-LLM collaborative approach to theory-based qualitative analysis for the HCI community and beyond.

Mental-health stigma remains a pervasive social problem that hampers treatment-seeking and recovery. Existing resources for training neural models to finely classify such stigma are limited, relying primarily on social-media or synthetic data without theoretical underpinnings. To remedy this gap, we present an expert-annotated, theory-informed corpus of human-chatbot interviews, comprising 4,141 snippets from 684 participants with documented socio-cultural backgrounds. Our experiments benchmark state-of-the-art neural models and empirically unpack the challenges of stigma detection. This dataset can facilitate research on computationally detecting, neutralizing, and counteracting mental-health stigma. Our corpus is openly available at https://github.com/HanMeng2004/Mental-Health-Stigma-Interview-Corpus.

Multi-view visualizations have gained prominence for their ability to simultaneously present multiple perspectives in a single display, aiding users in making informed decisions. Although several tools have been developed to facilitate the design of multi-view visualizations, there is a lack of support for effective color design in these systems. Thoughtful color design can enhance readability and provide cues, while careless color choices may lead to confusion. Moreover, human short-term memory imposes constraints on the number of colors that can be effectively employed in a multi-view visualization. To address these challenges, we introduce MVcolor, a color encoding recommendation system designed to maintain color encoding consistency and ensure adequate color discriminability within the constraints of human short-term memory. Our approach employs a unified color scheme across the entire multi-view visualization and groups views and visual objects based on their visual and semantic similarities. We conducted user studies to evaluate the effectiveness of our system and demonstrate its ability to improve color encoding in multi-view visualizations.

Due to the human-like nature, large language models (LLMs) often express uncertainty in their outputs. This expression, known as ”verbalized uncertainty” , can appear in phrases such as ”I’m sure that [...]” or ”It could be [...]” . However, few studies have explored how this expression impacts human users’ feelings towards AI, including their trust, satisfaction and task performance. Our research aims to fill this gap by exploring how different levels of verbalized uncertainty from the LLM’s outputs affect users’ perceptions and behaviors in AI-assisted decision-making scenarios. To this end, we conducted a between-condition study (N = 156), dividing participants into six groups based on two accuracy conditions and three conditions of verbalized uncertainty. We also used the widely played word guessing game Codenames to simulate the role of LLMs in assisting human decision-making. Our results show that medium verbalized uncertainty in the LLM’s expressions consistently leads to higher user trust, satisfaction, and task performance compared to high and low verbalized uncertainty. Our results also show that participants experience verbalized uncertainty differently based on the accuracy of the LLM. This study offers important implications for the future design of LLMs, suggesting adaptive strategies to express verbalized uncertainty based on the LLM’s accuracy. • Medium verbalized uncertainty leads to higher trust, satisfaction, and performance. • At different accuracy levels, the main factor influencing users’ trust varies. • Users perceive medium verbalized uncertainty as better low verbalized uncertainty. • Future LLMs should express verbalized uncertainty using adaptive strategies.

Chatbots hold promise as a technology to aid in sustained weight management. However, determining the optimal way for chatbots to deliver advice to effectively change user behaviors remains a significant hurdle. This research investigates the effects of different chatbot communication styles and expressions of care on user satisfaction, misinterpretation, and intent to adhere to the advice in weight-related conversations. A mixed method study with 97 participants classified as overweight was conducted, dividing them into four groups based on explicit/implicit communication styles and the presence or absence of caring language. Surprisingly, the study found that most participants in the explicit communication groups viewed the chatbot as non-offensive. These participants also reported higher levels of enjoyment and a greater intention to follow the chatbot's recommendations. Utilizing caring language may diminish users' perception of the chatbot as a marketing tool, thereby increasing their willingness to interact. The article discusses the implications for the design of healthcare chatbots.

Chatbots are increasingly integrated into people's lives and are widely used to help people. Recently, there has also been growing interest in the reverse direction-humans help chatbots-due to a wide range of benefits including better chatbot performance, human well-being, and collaborative outcomes. However, little research has explored the factors that motivate people to help chatbots. To address this gap, we draw on the Computers Are Social Actors (CASA) framework to examine how chatbot anthropomorphism-including human-like identity, emotional expression, and non-verbal expression-influences human empathy toward chatbots and their subsequent prosocial behaviors and intentions. We also explore people's own interpretations of their prosocial behaviors toward chatbots. We conducted an online experiment (N = 244) in which chatbots made mistakes in a collaborative image labeling task and explained the reasons to participants. We then measured participants' prosocial behaviors and intentions toward the chatbots. Our findings revealed that human identity and emotional expression of chatbots increased participants' prosocial behavior and intention toward chatbots, with empathy mediating these effects. Qualitative analysis identified two motivations for participants' prosocial behaviors: empathy for the chatbot and perceiving the chatbot as human-like. We discuss the implications of these results for understanding and promoting human prosocial behaviors toward chatbots.

Non-native speakers (NNSs) often face speaking challenges in real-time multilingual communication, such as struggling to articulate their thoughts. To address this issue, we developed an AI-based speaking assistant (AISA) that provides speaking references for NNSs based on their input queries, task background, and conversation history. To explore NNSs' interaction with AISA and its impact on NNSs' speaking during real-time multilingual communication, we conducted a mixed-method study involving a within-subject experiment and follow-up interviews. In the experiment, two native speakers (NSs) and one NNS formed a team (31 teams in total) and completed two collaborative tasks-one with access to the AISA and one without. Overall, our study revealed four types of AISA input patterns among NNSs, each reflecting different levels of effort and language preferences. Although AISA did not improve NNSs' speaking competence, follow-up interviews revealed that it helped improve the logical flow and depth of their speech. Moreover, the additional multitasking introduced by AISA, such as entering and reviewing system output, potentially elevated NNSs' workload and anxiety. Based on these observations, we discuss the pros and cons of implementing tools to assist NNS in real-time multilingual communication and offer design recommendations.

AI conversational agents have demonstrated efficacy in social contact interventions for stigma reduction at a low cost. However, the underlying mechanisms of how interaction designs contribute to these effects remain unclear. This study investigates how participating in three human-chatbot interactions affects attitudes toward mental illness. We developed three chatbots capable of engaging in either one-way information dissemination from chatbot to a human or two-way cooperation where the chatbot and a human exchange thoughts and work together on a cooperation task. We then conducted a two-week mixed-methods study to investigate variations over time and across different group memberships. The results indicate that human-AI cooperation can effectively reduce stigma toward individuals with mental illness by fostering relationships between humans and AI through social contact. Additionally, compared to a one-way chatbot, interacting with a cooperative chatbot led participants to perceive it as more competent and likable, promoting greater empathy during the conversation. However, despite the success in reducing stigma, inconsistencies between the chatbot's role and the mental health context raised concerns. We discuss the implications of our findings for human-chatbot interaction designs aimed at changing human attitudes.

Multi-agent systems, systems with multiple independent AI agents working together to achieve a common goal, are becoming increasingly prevalent in daily life. Drawing inspiration from the phenomenon of human group social influence, we investigate whether a group of AI agents can create social pressure on users to agree with them, potentially changing their stance on a topic. We conducted a study in which participants discussed social issues with either a single or multiple AI agents, and where the agents either agreed or disagreed with the user's stance on the topic. We found that conversing with multiple agents increased the social pressure felt by participants, and caused a greater shift in opinion towards the agents' stances on the conversation topics. Our study shows the potential advantages of multi-agent systems over single-agent platforms in causing opinion change. We discuss the resulting possibilities for multi-agent systems that promote social good, as well as potential malicious actors using these systems to manipulate public opinion.

Conversational Agents (CAs) have increasingly been integrated into everyday life, sparking significant discussions on social media. While previous research has examined public perceptions of AI in general, there is a notable lack in research focused on CAs, with fewer investigations into cultural variations in CA perceptions. To address this gap, this study used computational methods to analyze about one million social media discussions surrounding CAs and compared people’s discourses and perceptions of CAs in the US and China. We find Chinese participants tended to view CAs hedonically, perceived voice-based and physically embodied CAs as warmer and more competent, and generally expressed positive emotions. In contrat, US participants saw CAs more functionally, with an ambivalent attitude. Warm perception was a key driver of positive emotions toward CAs in both countries. We discussed practical implications for designing contextually sensitive and user-centric CAs to resonate with various users’ preferences and needs.

Chatbots are increasingly being used in mental healthcare - e.g., for assessing mental-health conditions and providing digital counseling - and have been found to have considerable potential for facilitating people's behavioral changes. Nevertheless, little research has examined how specific chatbot designs may help reduce public stigmatization of mental illness. To help fill that gap, this study explores how stigmatizing attitudes toward mental illness may be affected by conversations with chatbots that have 1) varying ways of expressing their interpretations of participants' statements and 2) different styles of self-disclosure. More specifically, we implemented and tested four chatbot designs that varied in terms of whether they interpreted participants' comments as stigmatizing or non-stigmatizing, and whether they provided stigmatizing, non-stigmatizing, or no self-disclosure of chatbot's own views. Over the two-week period of the experiment, all four chatbots' conversations with our participants centered on seven mental-illness vignettes, all featuring the same character. We found that the chatbot featuring non-stigmatizing interpretations and non-stigmatizing self-disclosure performed best at reducing the participants' stigmatizing attitudes, while the one that provided stigmatizing interpretations and stigmatizing self-disclosures had the least beneficial effect. We also discovered side effects of chatbot's self-disclosure: notably, that chatbots were perceived to have inflexible and strong opinions, which undermined their credibility. As such, this paper contributes to knowledge about how chatbot designs shape users' perceptions of the chatbots themselves, and how chatbots' interpretation and self-disclosure may be leveraged to help reduce mental-illness stigma.

This study delves into the digital interactions of stay-at-home moms (SAHMs) in Taiwan, exploring their utilization of LINE messaging app's chat groups as a pivotal means for informational, emotional, and practical support amidst their parenting journeys. Amidst the dominant patriarchal societal norms and potential isolations intrinsic to their role, these SAHMs establish a robust, cooperative digital environment, wherein maternal experiences, knowledge, and emotional backings are collectively shared and curated. Through a two-stage interview study, employing chatbot technology over 14 days with 18 participants, we unveiled how SAHMs engage in both tacit and explicit collaborative work within these chat groups, reclaiming their maternal identity and amplifying their parenting confidence. This exploration not only illuminates the critical role of technology in enhancing collaborative maternal work but also propels the discourse on how SAHMs navigate their distinctive needs within specific socio-cultural contexts through technological means. Consequently, this offers a foundational perspective towards crafting technological designs that are empathetically attuned to the lived experiences and needs of SAHMs in Taiwan.

Chatbots have been designed to provide interventions in mental healthcare. However, how chatbot-based social contact can mitigate social stigma in mental illness remains under-explored. We designed two chatbots that deliver either first-person or third-person narratives about mental illness and evaluated them using a mixed methods study. Compared to a web survey group, participants in both chatbot groups decreased their beliefs that individuals are personally responsible for their mental illnesses, and increased their intentions to help. Additionally, participants in the first-person chatbot group showed a reduced level of fear, and a lower desire for social distance from people with mental illness. Many in the first-person chatbot group also reported a feeling of relationship with the chatbot, and chose to phrase their responses empathetically. Results demonstrated that chatbot-based social contact has promising potential for mitigating mental illness stigma. Implications for designing chatbot-based social contact are discussed.

Chatbots are capable of remembering and referencing previous conversations, but does this enhance user engagement or infringe on privacy? To explore this trade-off, we investigated the format of how a chatbot references previous conversations with a user and its effects on a user’s perceptions and privacy concerns. In a three-week longitudinal between-subjects study, 169 participants talked about their dental flossing habits to a chatbot that either, (1-None): did not explicitly reference previous user utterances, (2-Verbatim): referenced previous utterances verbatim, or (3-Paraphrase): used paraphrases to reference previous utterances. Participants perceived Verbatim and Paraphrase chatbots as more intelligent and engaging. However, the Verbatim chatbot also raised privacy concerns with participants. To gain insights as to why people prefer certain conditions or had privacy concerns, we conducted semi-structured interviews with 15 participants. We discuss implications from our findings that can help designers choose an appropriate format to reference previous user utterances and inform in the design of longitudinal dialogue scripting.

The similarity effect refers to the tendency for people to be more easily influenced by others who resemble them in appearance. This phenomenon has been found to have positive impacts, including on the building of trust, that enrich the quality of communication (e.g., fluency or collaboration performance). While research has shown that the similarity effect occurs in screen-based communication platforms, it remains unclear how this phenomenon impacts user perceptions, especially of others' persuasiveness, in immersive environments such as virtual reality (VR). In this study, we adopted a mixed-methods approach to exploring how interaction with avatars of similar appearance to one's own self-representation influences conversations. Such similarity was operationalized as having three levels: identicality, moderate similarity, and dissimilarity. The study found that avatars of moderate similarity have the greatest persuasiveness; however, in both identicality and moderate similarity conditions, participants felt it was easier to communicate with and lower eeriness rating to avatars than in the dissimilarity condition. Multiple linear regression further revealed that users who had relatively low self-esteem and/or were relatively conscientious were more susceptible to the positive effect of appearance similarity on persuasiveness. We conclude that the similarity effect, especially when the similarity in question is moderate, could be leveraged to support persuasiveness in VR-based communication.

Prior research suggests that affective states of self-regulated learning can be used to improve learners’ cognitive processes and their learning outcomes. However, little research explored the effect of using facial movements to detect learners’ affective states on self-regulated learning. In this work, we designed, implemented, and evaluated Mirror: a self-regulated learning tool that applies facial expression recognition to support learners’ reflections in video-based learning. We conducted two studies to identify user needs (with 12 participants) and to evaluate the tool (with 16 participants). The results show that, after watching a video, participants benefited from using Mirror through different reflection processes, e.g., gaining a deeper understanding of their learning experiences through self-observation and attributing causes for their learning affects through self-judgment. Meanwhile, we also identified several ethical concerns, e.g., users’ agency of handling the uncertainty of AI, reactivity towards outcome-based AI, over-reliance on “positive” AI results, and fairness of AI informed decision-making.

Sexual-minority women (SMWs) in China are often subject to strong stigmatization and tend to have limited opportunities to connect with other SMWs in offline contexts. Although dating apps help them connect and seek social support, little is known about SMWs’ practices of self-disclosure and connection-building through those apps. To address this gap, we interviewed 43 SMW dating-app users in China. We found that these SMWs developed distinctive self-disclosure strategies, such as posting non-facial photos and implicitly disclosing their whereabouts by blending location information into photos that only those in the know could understand, to avoid interference from aggressive acquaintances and other risks of unintentional disclosure of their SMW identities. Moreover, they used dating apps not only to recognize other SMWs offline and build relationships with them, but to exchange emotional support in the process of SMW identity development. Our findings have design implications for supporting SMWs and improving their online dating experiences.

In China, lesbian and bisexual women face intense stigma and difficulties developing relationships with each other. Although prior research has shown that online communities help LGBT people connect and exchange social support, few studies have explored the challenges Chinese lesbian and bisexual women face when initiating, growing, and sustaining such communities, in an atmosphere of platform censorship of LGBT-related content and intense discrimination from non-LGBT people. To address this gap, we interviewed 40 Weibo users in China, four bloggers and 36 followers of their blogs, who self-identified as lesbian or bisexual women. We found that a key technique these bloggers used to initiate their online communities was helping followers publish posts seeking support, sharing personal experiences, and seeking offline relationships. Then, their followers built relationships with bloggers by journaling their daily experiences as lesbian or bisexual women via private-messaging channels. As the communities' members grew more attached to them, bloggers and their followers began to work together to protect themselves from external threats, including Weibo's censorship and non-LGBT+ infiltrators' harassment. However, such attachment to the communities sometimes might lead to conflicts within them, which in turn prompted many members to leave, raising questions about the communities' long-term prospects. Our findings foreground important design considerations for those seeking to help lesbian and bisexual women in China and other discriminatory environments to develop safe online communities.

Chatbots are regarded as a promising technology for delivering guidance. Prior studies show that chatbots have the potential of coaching users to learn different skills; however, several limitations of chatbot-based approaches remain. People may become disengaged from using chatbot-guided systems and fail to follow the guidance for complex tasks. In this paper, we design chatbots with (HC) and without (OC) human support to deliver guidance for people to practice journaling skills. We conducted a mixed-method study with 35 participants to investigate their actual interaction, perceived interaction, and the effects of interacting with the two chatbots. The participants were randomly assigned to use one of the chatbots for four weeks. Our results show that the HC participants followed the guidance more faithfully during journaling practices and perceived a significantly higher level of engagement and trust with the chatbot system than the OC participants. However, after finishing the journaling-skill training session, the OC participants were more willing to keep using the learned skills than the HC participants. Our work provides new insights into the design of integrating human support into chatbot-based interventions for delivering guidance.

Chatbots have great potential to serve as a low-cost, effective tool to support people's self-disclosure. Prior work has shown that reciprocity occurs in human-machine dialog; however, whether reciprocity can be leveraged to promote and sustain deep self-disclosure over time has not been systematically studied. In this work, we design, implement and evaluate a chatbot that has self-disclosure features when it performs small talk with people. We ran a study with 47 participants and divided them into three groups to use different chatting styles of the chatbot for three weeks. We found that chatbot self-disclosure had a reciprocal effect on promoting deeper participant self-disclosure that lasted over the study period, in which the other chat styles without self-disclosure features failed to deliver. Chatbot self-disclosure also had a positive effect on improving participants' perceived intimacy and enjoyment over the study period. Finally, we reflect on the design implications of chatbots where deep self-disclosure is needed over time.

Chatbots (conversational agents) are increasingly receiving attention in mental health domains because they elicit honest self-disclosure about personal experiences and emotions. Although such self-disclosure contents can be useful for gauging mental status, little research has addressed how to automatically assess mental status from self-disclosures to a chatbot. If a chatbot can automatically assess the mental status of users, it can help them improve their mental wellness or facilitate access to mental professionals. In this paper, we examine whether indicators that identify depression from written texts (e.g., social media posts) are also useful for assessing mental status from disclosures to a chatbot. We first ran a study with 30 participants who engaged in daily journaling with a chatbot that prompted them to record their moods and experiences for three weeks. We then divided the participants' self-disclosure data into three groups based on their mental state changes before and after the study: improved vs. deteriorated vs. no change. Comparing the data among the three groups, participants whose mental states deteriorated during the study gradually used fewer positive emotion and concrete words but more negative emotion words when describing their daily experiences and feelings to the chatbot.

Chatbots are becoming increasingly popular. One promising application for chatbots is to elicit people's self-disclosure of their personal experiences, thoughts, and feelings. As receiving one's deep self-disclosure is critical for mental health professionals to understand people's mental status, chatbots show great potential in the mental health domain. However, there is a lack of research addressing if and how people self-disclose sensitive topics to a real mental health professional (MHP) through a chatbot. In this work, we designed, implemented and evaluated a chatbot that offered three chatting styles; we also conducted a study with 47 participants who were randomly assigned into three groups where each group experienced the chatbot's self-disclosure at varying levels respectively. After using the chatbot for a few weeks, participants were introduced to a MHP and were asked if they would like to share their self-disclosed content with the MHP. If they chose to share, the participants had the option of changing (adding, deleting, and editing) the content they self-disclosed to the chatbot. Comparing participants' self-disclosure data the week before and the week after sharing with the MHP, our results showed that, within each group, the depth of participants' self-disclosure to the chatbot remained after sharing with the MHP; participants exhibited deeper self-disclosure to the MHP through a more self-disclosing chatbot; further, through conversation log analysis, we found that some participants made different edits on their self-disclosed content before sharing it with the MHP. Participants' interview and survey feedback suggested an interaction between participants' trust in the chatbot and their trust in the MHP, which further explained participants' self-disclosure behavior.

Digital gifting in live streaming, in which viewers buy digital gifts to reward the streamers, was worth over $200 million in 2018 in China and its growth has been accelerating. This paper explores what motivates people to tip and how it impacts interactions between viewers and streamers. Through a survey, we identified the main categories of viewers' tipping motivations. We found that viewers were motivated by the reciprocal acts of streamers, who would engage in various types of social interactions with tippers during the live streams. The styles of interactions and contents of live stream based on the tipping are differently influenced by the motivations of viewers and streamers. For example, viewers often tip large to attract attentions from the crowd or promote preferred live-streaming content. These findings provide more knowledge on the social interaction in live streaming platforms.

Live streaming is an emergent social medium that allows remote interaction between the streamer and an audience of any size. Major live-streaming platforms in some Asian countries have a digital gift-giving feature that allows viewers to directly reward streamers during live sessions. Streamers can later exchange the digital gifts they have received for money, and this monetary incentive appears likely to influence how they interact with their viewers and generate live-streaming content. However, the precise nature and mechanisms of such impact have not previously been explored. Therefore, this qualitative study with 13 streamer participants examines how digital gifting influences streamers' motivations and the nature of both the content that they generate and their social interactions with their audiences. It reports that the digital-gifting function serves as a major incentive for active streaming, but may also disincentivize some streamers from continuing to contribute, for reasons that will be explored. Moreover, the participants devised strategies for both content generation and social interaction with the specific objective of earning gifts from viewers: practices that, in some cases, appeared to limit the quality of their live-streaming content. It was also noted that the streamers tended to have constrained social relationships with their viewers, in part because such relationships were seen as unequal or one-sided due to gift-giving behavior. The paper concludes with a discussion of design considerations for the incorporation of gift-giving features into live-streaming platforms, and additional recommendations for future research and the design of such platforms.

On existing crowdfunding platforms, the allocation of money is often not regulated, which leads to less-than-ideal distribution of resources. For example, recent donations to hurricane victims through their crowdfunding campaigns often lead to overfunding of certain victims while underfunding others. Inspired by algorithms from economic theories, we proposed a Smart Crowd Donate system encourages donors to express preferences to multiple projects and reallocates funds dynamically across these preferences over time. We conducted a user study in which recruited 452 participants to simulate a small scale of crowdfunding. The findings of our user study supported the idea that the Smart Crowd Donate system has potential to efficiently distribute funds to projects and allows more projects to receive the amount of money they need.

Online learning is increasingly prevalent as an option for self-learning and as a resource for instructional design. Prerecorded video is currently the main medium of online education content delivery and instruction; this affords asynchronicity and flexibility, and enables the dissemination of lecture content in a distributed and scalable manner. However, the same properties may impede learners' engagement due to the lack of social interaction and peer support. In this paper, we propose a time-anchored commenting interface to allow online learners who watch the same video clips to exchange comments on them. Comments left by previous learners at specific time points of a video are displayed to new learners when they watch the same video and reach those time points. We investigated how the display of time-anchored comments (dynamic or static) and type of comments (content-related or social-oriented) influenced users' perceived engagement, perceived social interactivity, and learning outcomes. Our results show that dynamically displaying time-anchored comments can indeed enhance learners' perceived social interactivity. Moreover, the content of comments would further affect learners' intention of commenting. Based on our findings, we make various recommendations for the improvement of social interaction and learning experience in online education.

Audio notifications are an important means of prompting users of electronic products. Although useful in most environments, audio notifications are ineffective in certain situations, especially against particular auditory backgrounds or when the user is distracted. Several studies have used behavioral performance to evaluate audio notifications, but these studies failed to achieve consistent results due to factors including user subjectivity and environmental differences; thus, a new method and more objective indicators are necessary. In this study, we propose an approach based on electroencephalography (EEG) to evaluate audio notifications by measuring users' auditory perceptual responses (mismatch negativity) and attention shifting (P3a). We demonstrate our approach by applying it to the usability testing of audio notifications in realistic scenarios, such as users performing a major task amid ambient noises. Our results open a new perspective for evaluating the design of the audio notifications.