Abstract:Motivated by the effects of the filter bubble and echo chamber phenomena on social media, we developed a smart home device, Spkr, that unpredictably "pushes" sociopolitical discussion topics into the home. The device utilised trending Twitter discussions, categorised by their sociopolitical alignment, to present people with a purposefully assorted range of viewpoints. We deployed Spkr in 10 homes for 28 days with a diverse range of participants and interviewed them about their experiences. Our results show tha… Show more
“…In information retrieval and recommender systems, serendipity is studied as an optimization criterion to increase the exposure to novel and diverse information [54,59]. Many systems were developed to help people encounter diverse perspectives [14,68], deliberate on controversial topics [13,30,48], be aware of one's own information bubble and better control filtering mechanisms [17,27,41]. As Garrett and Resnick [19] argued, to increase people's consumption of attitude-challenging information, the key lies in presenting highquality challenging items in the right context, and/or reducing people's cognitive dissonance.…”
Section: Related Work 21 Selective Exposure Confirmation Bias and Ech...mentioning
confidence: 99%
“…People also receive information passively or through scanning. Indeed, prior works argued that conversational search and agent systems have the potential advantage of making individuals more receptive toward proactive interactions from the system [51], and HCI researchers explored leveraging conversational systems such as Alexa to broadcast diverse views [14]. It would be interesting to explore whether conversational interactions enabled by LLMs can effectively act as active nudging for diverse views, and whether they provide benefits over conventional information systems such as news feed and deliberation platforms.…”
Section: Mitigating Selective Exposure In Conversational Searchmentioning
Large language models (LLMs) powered conversational search systems have already been used by hundreds of millions of people, and are believed to bring many benefits over conventional search. However, while decades of research and public discourse interrogated the risk of search systems in increasing selective exposure and creating echo chambers-limiting exposure to diverse opinions and leading to opinion polarization, little is known about such a risk of LLM-powered conversational search. We conduct two experiments to investigate: 1) whether and how LLM-powered conversational search increases selective exposure compared to conventional search; 2) whether and how LLMs with opinion biases that either reinforce or challenge the user's view change the effect. Overall, we found that participants engaged in more biased information querying with LLM-powered conversational search, and an opinionated LLM reinforcing their views exacerbated this bias. These results present critical implications for the development of LLMs and conversational search systems, and the policy governing these technologies.
CCS CONCEPTS• Human-centered computing → Human computer interaction (HCI); • Computing methodologies → Artificial intelligence; • Information systems → Web searching and information discovery; Search interfaces.
“…In information retrieval and recommender systems, serendipity is studied as an optimization criterion to increase the exposure to novel and diverse information [54,59]. Many systems were developed to help people encounter diverse perspectives [14,68], deliberate on controversial topics [13,30,48], be aware of one's own information bubble and better control filtering mechanisms [17,27,41]. As Garrett and Resnick [19] argued, to increase people's consumption of attitude-challenging information, the key lies in presenting highquality challenging items in the right context, and/or reducing people's cognitive dissonance.…”
Section: Related Work 21 Selective Exposure Confirmation Bias and Ech...mentioning
confidence: 99%
“…People also receive information passively or through scanning. Indeed, prior works argued that conversational search and agent systems have the potential advantage of making individuals more receptive toward proactive interactions from the system [51], and HCI researchers explored leveraging conversational systems such as Alexa to broadcast diverse views [14]. It would be interesting to explore whether conversational interactions enabled by LLMs can effectively act as active nudging for diverse views, and whether they provide benefits over conventional information systems such as news feed and deliberation platforms.…”
Section: Mitigating Selective Exposure In Conversational Searchmentioning
Large language models (LLMs) powered conversational search systems have already been used by hundreds of millions of people, and are believed to bring many benefits over conventional search. However, while decades of research and public discourse interrogated the risk of search systems in increasing selective exposure and creating echo chambers-limiting exposure to diverse opinions and leading to opinion polarization, little is known about such a risk of LLM-powered conversational search. We conduct two experiments to investigate: 1) whether and how LLM-powered conversational search increases selective exposure compared to conventional search; 2) whether and how LLMs with opinion biases that either reinforce or challenge the user's view change the effect. Overall, we found that participants engaged in more biased information querying with LLM-powered conversational search, and an opinionated LLM reinforcing their views exacerbated this bias. These results present critical implications for the development of LLMs and conversational search systems, and the policy governing these technologies.
CCS CONCEPTS• Human-centered computing → Human computer interaction (HCI); • Computing methodologies → Artificial intelligence; • Information systems → Web searching and information discovery; Search interfaces.
“…Salehi-Abari and Boutilier, 2015; Le et al, 2017;Nelimarkka et al, 2018;Semaan et al, 2014;Al-Ani et al, 2012;Park et al, 2011;Trevisan et al, 2019;Hemphill and Roback, 2014;Hemphill et al, 2013;Agarwal et al, 2020;Li et al, 2018) (Mahoney et al, 2016;Zubiaga et al, 2013;Maruyama et al, 2014;Kriplean et al, 2014;Pierson, 2015;Rho and Mazmanian, 2020;Le et al, 2017;Semaan et al, 2014) Continued on next page Media consumption "Broadening Exposure to Socio-Political Opinions via a Pushy Smart Home Device" (Feltwell et al, 2020), "Is a polarized society inevitable, where people choose to be exposed to only political news and commentary that reinforces their existing viewpoints?" Golbeck and Hansen (2011) , and "When one searches for political candidates on Google, a panel composed of recent news stories, known as Top stories, is commonly shown at the top of the search results page."…”
Human–computer interaction scholars are increasingly touching on topics relatedto politics or democracy. As these concepts are ambiguous, an examination ofconcepts’ invoked meanings aids in the self-reflection of our research efforts. Weconduct a thematic analysis of all papers with the word ‘politics’ in abstract,title or keywords (n=378) and likewise 152 papers with the word ‘democracy.’We observe that these words are increasingly being used in human-computerinteraction, both in absolute and relative terms. At the same time, we show thatresearchers invoke these words with diverse levels of analysis in mind: the earlyresearch focused on mezzo-level (i.e., small groups), but more recently the workhas begun to include macro-level analysis (i.e., society and politics as played inthe public sphere). After the increasing focus on the macro-level, we see a tran-sition towards more normative and activist research, in some areas it replacesobservational and empirical research. These differences indicate semantic differ-ences, which – in the worst case – may limit scientific progress. We bring thesedifferences visible to help in further exchanges of ideas and human–computerinteraction community to explore how it orients itself to politics and democracy.
“…Although we have some evidence that this happened (as participants reported conversations around the workbook topics), this was not as prevalent as we had expected and the majority of participants did not openly look at activities shared by others-suggesting this was too personal or voyeuristic-and rarely looked at these activities during meetings. Previous work [23] has shown the effectiveness of delivering different content based on the individual, alternative workbook content, or tailoring of content for each PGR student, could help appropriately communicate desires and manage expectations. Conversational prompts were suggested made by a small number of participants in the study and could represent an alternative solution-guided chats were successfully used to facilitate online "talk therapy" between people experiencing mental illnesses [65], facilitating deep and valuable conversations, but which raised concerns that participants did not want to share-similar to problems we experienced in this study.…”
Section: Dissonance and Supply Of Social Supportmentioning
Student mental health and wellbeing have come under increased scrutiny in recent years. Postgraduate research (PGR) students are at risk of experiencing mental health concerns and this, with the often isolated and competitive nature of their work, can impact their sense of community and social connectedness. In response to these concerns, we designed Pears, a system to connect PGR students for regular “pearings” (in-person meetings) and provides activities to promote reflection and conversation. A four-week evaluation of Pears with 15 students highlighted its potential to sometimes, but not always, facilitate peer support. Some participants would instead meet formally and according to their needs, others instead used the system to make new social connections. Additionally, some participants who faced work-related difficulties during the study found using the system contributed to their stress levels. We conclude by noting how technologies that encourage peer support can help build social relationships, providing an avenue to share similar PhD experiences and guidance for those new to the experience, while importantly raising awareness and an understanding for the need to take time for self-care. However, these technologies must be utilised carefully, and are not a replacement for other sources of student support in universities.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.