Abstract
Bystander intervention, or upstanding, is an effective antidote to cyberbullying, but entails many challenges (e.g. self-efficacy, not knowing what to do or how to upstand). Through two studies, this paper investigates collaborative upstanding, examining how a conversational partner (human or AI) can guide bystanders through these challenges in-situ. In a paired role-play study (n=24), we found that bystanders faced significant challenges in how to intervene. Even after deciding to act, how-to challenges often reignited doubts about their self-efficacy and responsibility. Using these insights, we designed ConCUR, a chatbot that (1) encourages bystanders to co-author an upstanding message, leading them to confront how-to challenges sooner, and (2) addresses how-to challenges simultaneously with other challenges that are introduced through a flexible process. Our second study (n=20) suggests such a chatbot is effective in promoting upstanding behavior in the lab setting. We discuss the implications of in-situ collaborative upstanding to upstanding education research, framing upstanding as an iterative and flexible process rather than sequential.
Introduction
When cyberbullying occurs on social media, having bystanders who speak up against offenders to defend the victims, known as upstanding, is crucial. Upstanding behaviors play a significant role in resolving a cyberbullying situation and supporting the victim. Public and active upstanding, such as public comments that call out the bully’s behavior, are particularly effective as they could encourage other bystanders to respond, as well as help mitigate the negative impact of cyberbullying on the victim’s mental well-being.
Previous work identifies five stages a bystander takes before performing an intervention: (1) noticing the event, (2) interpreting the situation, (3) assuming responsibility, (4) deciding how to intervene, and (5) taking action. However, each of these stages come with barriers that deter bystanders from taking actions. HCI researchers have created systems designed to encourage upstanding, focusing on single stages such as increasing the bystanders’ sense of responsibility. To this end, educators, HCI researchers, and designers of social computing platforms have explored methods to encourage upstanding behavior. Educational programs and frameworks (e.g. STAC) have been deployed to train and inform learners of upstanding strategies and tactics.
More recent work has suggested the potential of human-AI collaboration in teaching upstanding behavior, such as by using AI as role-play agents, or chatbots that enable students to rehearse upstanding behavior. However, to our knowledge, no previous work has yet to put this into actual use through developing and evaluating chatbots. Building on previous work, we propose a collaborative upstanding approach: a joint process of upstanding where a bystander collaborates with another entity to make intervention decisions. In this study, we explore a model where a ‘co-pilot’ guides and trains a bystander to perform upstanding behaviors through conversation.
In this paper, we explore the question of how to design chatbots that address upstanding barriers in situ.
To this end, we conducted two studies:
(1) to identify design opportunities and conversational strategies, and (2) to build and evaluate the effectiveness of a collaborative upstanding chatbot.
In the first study, we conducted a paired role-play study (N=24) to explore how collaborative upstanding can foster public, active upstanding behavior. We discovered that bystander participants were often limited in deciding to perform an intervention by a lack of how-to knowledge rather than what-to knowledge. Facing such how-to barriers also re-triggered various concerns from previous stages, such as assuming responsibility and situation interpretation. We also identified the promise of conversational interactions in flexibly addressing such challenges as they occur, as well as salient conversational strategies utilized by co-pilot participants.
Based on these insights, we designed ConCUR (CONversational Collaborative Upstanding in Real-time), an educational system based on a human-AI collaborative upstanding approach. ConCUR shifts the workflow to first address the how-to concerns, compared to a typical motivation-first approach in the bystander intervention model. To achieve this, ConCUR asks bystanders to craft an upstanding message with the aid of the AI co-pilot at the beginning of the intervention. We implemented a prototype of ConCUR and conducted a user study (N=20) to evaluate the effectiveness of this approach. Participants who used ConCUR during the study became more motivated to try public upstanding, as the how-to-first approach allowed them to address the various how-to and motivational barriers simultaneously and effectively.
Based on our findings, we discuss the design implications of using conversational systems, as in collaborative upstanding, to address upstanding barriers. We consider the possibility of understanding and designing for the upstanding process as a flexible, non-linear process. The barriers bystanders face do not necessarily follow the five-stage model in a linear and sequential way, and systems should be built to address this characteristic. We encourage future research to pay special attention to situational and social factors, considering the prevalence of cognitive-situational barriers (e.g. self-efficacy and concerns about negative social evaluation).
Conclusion
In this work, we present two design probe studies based on a collaborative upstanding approach, where conversational AI systems are used as a “guide on the side” to promote and train upstanding behavior. Our findings emphasize the role of how-to barriers in the upstanding process, as well as identify effective conversational strategies and methods to overcome these barriers. In addition, we propose the potential of a how-to-first approach to collaborative upstanding, which we implemented through ConCUR. Effective human-AI collaborative upstanding requires AI co-pilots to flexibly adapt to bystanders’ individual reservations, going through an iterative, gradual process of persuasion. This emphasizes the importance of in-situ training, as well as using motivational and behavioral nudges to promote upstanding behavior. We provide design insights that can inform future design of collaborative upstanding systems, such as utilizing iterative and flexible conversational flows, exploratory co-writing processes, and how-to-first approaches that motivate users to overcome barriers to upstanding. Finally, we present the theoretical implications of this work, expanding upon the Bystander Intervention Model and identifying intervention action as a multi-event process that includes post-action maintenance and evaluation. We hope that future researchers and designers can build upon these findings to develop improved collaborative upstanding systems to train upstanding behaviors.
