IRIS Publications

Abstract

Abstract: This work investigated the prevalence of filter bubble or echo chamber-related phenomena, psychological factors rendering individuals resilient or vulnerable to them, and their associations to political views focusing on extremity and polarization. For this, a cross-cultural replication of a study by Sindermann et al. (2021) was conducted. As an extension, multiple political views variables were assessed to examine whether the application of different conceptualizations of political views explains heterogeneous findings across previous studies. Two samples were recruited: 390 (n = 135 males) US college students and a quota sample of 489 (n = 243 males) US adults. Participants completed personality scales and measures on political news consumption homogeneity versus heterogeneity and political views. Consistent with previous research, results revealed few individuals consume political news absolutely homogeneously. Openness was negatively related to the degree of political news consumption homogeneity, and the relationship between political news consumption homogeneity and political views yielded inconsistent, often statistically nonsignificant, results. These findings challenge the prevailing notion of filter bubbles and echo chambers as widespread phenomena and indicate that relationships between political news consumption homogeneity and political views are not necessarily deleterious with respect to extremization and polarization. As such, the results suggest that these phenomena might not be as significant for the general population as previously thought. Nonetheless, certain individuals might still find themselves in filter bubbles or echo chambers and suffer from accompanying consequences. In this regard, the present work replicates findings underscoring that individuals with lower Openness exhibit greater political news consumption homogeneity.

Abstract

Research on language as interactive discourse underscores the deliberate use of demographic parameters such as gender, ethnicity, and class to shape social identities. For example, by explicitly disclosing one's information and enforcing one's social identity to an online community, the reception by and interaction with the said community is impacted, e.g., strengthening one's opinions by depicting the speaker as credible through their experience in the subject. Here, we present a first thorough study of the role and effects of self-disclosures on online discourse dynamics, focusing on a pervasive type of self-disclosure: author gender. Concretely, we investigate the contexts and properties of gender self-disclosures and their impact on interaction dynamics in an online persuasive forum, ChangeMyView. Our contribution is twofold. At the level of the target phenomenon, we fill a research gap in the understanding of the impact of these self-disclosures on the discourse by bringing together features related to forum activity (votes, number of comments), linguistic/stylistic features from the literature, and discourse topics. At the level of the contributed resource, we enrich and release a comprehensive dataset that will provide a further impulse for research on the interplay between gender disclosures, community interaction, and persuasion in online discourse.

Abstract

Transdisciplinary processes deal with transdisciplinary problems that are (i) complex, (ii) societally relevant, (iii) ill-defined, and (iv) real-world problems which often show a high degree of ambiguity resulting in contested perceptions and evaluations among and between scientists and practitioners. Therefore, they are susceptible to multiple trade-offs. Transdisciplinary processes construct socially robust orientations (SoROs) particularly for sustainable transitioning. The integration of science and practice knowledge on equal footing (1) is considered the core of transdisciplinary processes. Yet other forms of knowledge integration contribute essentially to construct SoROs. Individuals may (2) use different modes of thought; (3) refer to various cultures with diverse value and belief systems; and (4) problems are perceived and prioritized based on roles and interests. Coping with transdisciplinary problems, (5) purposeful differentiation and integration and (6) an integration of evolutionary evolving codes of representing knowledge are necessary. Finally, (7) what systems to integrate requires consensus-building among participating scientists and practitioners. This paper is Part I of a two-part publication. It provides a conceptualization of the different types of knowledge integration. Part II analyzes tasks, challenges, and barriers related to different types of knowledge integration in five transdisciplinary processes which developed SoROs for sensitive subsystems of Germany affected by the irresponsible use of digital data.

Abstract

Transdisciplinary problems are complex, ill-defined, societal real-world problems with high ambiguity that are contested and require multiple trade-offs. Part I of this paper showed that transdisciplinary processes include seven types of knowledge integration: system (ontological), epistemological, cultural, cognitive, social conflict, evolutionary (levels of representation), and complexity-theory-based types of knowledge integration. The epistemological integration of the different modes of reasoning from science and practice is a unique selling point of the transdisciplinary process. Part II presents five transdisciplinary processes for the responsible use of digital data in different vulnerable/sensitive subsystems of Germany (mobility, health, agriculture, SME, and social media). Between 10 and 18 participants (equally representing science and practice in each group) synchronously constructed socially robust orientations as pillars of a white book. We elaborate that outcomes of a transdisciplinary process can be improved, and barriers diminished by reflecting on which of the seven types of knowledge integration are applied (see Part I). This is done for the six phases of a transdisciplinary process: (1) triggering, (2) initiation, (3) preparation, (4) planning, (5) core, and (6) post-processing. We particularly address researchers and practitioners who seek insights into how the production and integration of knowledge can be improved by transdisciplinary processes.

Abstract

We explored problematic smartphone use (PSU) using latent profile analysis (LPA) and relationships with anxiety variables, including severity of generalized anxiety disorder (GAD), social anxiety disorder (SAD), and Fear of Missing Out (FoMO) in a non-clinical sample. We conducted a web-based survey (during the COVID-19 pandemic from February to March 2020) with high school students (N = 1,797; 1,164 female; ages 13–19 years) in Tianjin, China, administering the Smartphone Addiction Scale-Short Version (SAS-SV) to assess PSU, Generalized Anxiety Disorder (GAD-7) Scale, Social Interaction Anxiety Scale (SIAS), and Fear of Missing Out (FoMO) Scale. Using Mplus 8.7, we conducted LPA on SAS-SV item responses to uncover latent profiles and relations with anxiety and fear measures. A three-profile PSU model fit the data according to fit indices and likelihood ratio tests. SAS-SV item responses were lowest in profile 1, moderate in profile 2, and most severe in profile 3. Individual PSU profiles modeled by LPA demonstrated significant differences in social and generalized anxiety severity and FoMO. Controlling for age and sex, adolescents with higher levels of anxiety were more likely to be classified as profiles 2 and 3 rather than profile 1. These findings will hopefully inspire future studies and treatments concerning the severity of PSU as it relates to various psychopathology constructs.

Abstract

Massive efforts are made to reduce biases in both data and algorithms to render AI applications fair. These efforts are propelled by various high-profile cases where biased algorithmic decision-making caused harm to women, people of color, minorities, etc. However, the AI fairness field still succumbs to a blind spot, namely its insensitivity to discrimination against animals. This paper is a critical comment on current fairness research in AI. It is the first to describe the `speciesist bias' and investigate it in several different AI systems by reflecting on the problem via a normative analysis and by probing, in several case studies, image recognition, word embedding, and language models with established methods for bias detection. We claim that animals matter morally and that discriminating against them is unethical. Furthermore, we provide evidence for speciesist biases in all the mentioned areas of AI. We find that speciesist biases are solidified by many mainstream AI applications, especially in the fields of computer vision as well as natural language processing. In both cases, this occurs because the models are trained on datasets in which speciesist patterns prevail. Therefore, AI technologies currently play a significant role in perpetuating and normalizing violence against animals. To change this, AI fairness frameworks must widen their scope and include mitigation measures for speciesist biases. This paper addresses the AI community in this regard and stresses the influence AI systems can have on either increasing or reducing the violence that is inflicted on animals, especially on farmed animals.

Abstract

In response to the coronavirus disease 2019 (COVID-19) pandemic, many individuals turned to synchronous online video communication technologies as a substitute for real-world face-to-face interactions. Evidence indicates that some users of such technologies show symptoms of exhaustion and fatigue during and after videoconferences (VCs) – this phenomenon is referred to as Videoconference Fatigue (VC fatigue). Research characterizing the possible vulnerability factors for VC fatigue is still scarce and considered to be in its early stage. Contributing to closing this gap in the existing literature is the motivation for the present study. Survey data was collected from 311 German-speaking participants to explore the relationships of VC fatigue with several psychological factors including autistic traits, social phobia, Fear of COVID-19, tendencies towards Internet Use Disorders (IUD tendencies), and Fear of Missing Out (FoMO, trait and state variables). Results showed that VC fatigue was significantly positively correlated with all of these psychological factors except state-FoMO, and corss-sectional mediation analyses provided further evidence for the positive association between autistic traits and VC fatigue. Specifically, the relationship between autistic traits and VC fatigue was mediated by Fear of COVID-19 and IUD tendencies rather than social phobia, with the latter being a preregistered hypothesis. This study adds to the literature by revealing several possible vulnerability factors associated with VC fatigue. In essence, the present work sheds light on the complex association between autistic traits and VC fatigue. We discuss the implications of our study as well as its limitations and potential avenues for future research.

Abstract

Recent progress in large language models has led to applications that can (at least) simulate possession of full moral agency due to their capacity to report context-sensitive moral assessments in open-domain conversations. However, automating moral decision-making faces several methodological as well as ethical challenges. They arise in the fields of bias mitigation, missing ground truth for moral ``correctness'', effects of bounded ethicality in machines, changes in moral norms over time, risks of using morally informed AI systems as actual advice, as well as societal implications an increasing importance of algorithmic moral decision-making would have. This paper comments on all these challenges and provides critical considerations for future research on full artificial moral agency. Importantly, some of the adduced challenges can be met by more careful technology design, but others necessarily require engagement with core problems of meta-ethics.

Abstract

The field of artificial intelligence (AI) alignment aims to investigate whether AI technologies align with human interests and values and function in a safe and ethical manner. AI alignment is particularly relevant for large language models (LLMs), which have the potential to exhibit unintended behavior due to their ability to learn and adapt in ways that are difficult to predict. In this paper, we discuss methodological challenges for the alignment problem specifically in the context of LLMs trained to summarize texts. In particular, we focus on methods for collecting reliable human feedback on summaries to train a reward model which in turn improves the summarization model. We conclude by suggesting specific improvements in the experimental design of alignment studies for LLMs' summarization capabilities.

Abstract

Machine learning (ML) techniques have become one of the most successful scientific tools and changed the everyday life of people around the globe (e.g., search engines). A vast amount of digital data sources on human behaviour has emerged due to the rise of the internet and opened the door for computer scientists to apply ML on social phenomena. In the social sciences, however, the adoption of ML has been less enthusiastic. To investigate the relation of traditional statistics and ML, this paper shows how ML might be used as regression analysis. For that purpose, we illustrate what a typical social science approach might look like and how using ML techniques could contribute additional insights when it comes to estimators (non-linearity) or the assessment of model fit (predictive power). In particular, we reveal how epistemological differences shape the potential usage of ML in the social sciences and discuss the methodological trade-off of applying ML compared to traditional statistics.

Abstract

Following recent technological developments in the artificial intelligence space, artificial agents are increasingly taking over organizational tasks typically reserved for humans. Studies have shown that humans respond differently to this, with some being appreciative of their advice (algorithm appreciation), others being averse toward them (algorithm aversion), and others still fully relinquishing control to artificial agents without adequate oversight (automation bias). Using systems thinking, we analyze the existing literature on these phenomena and develop a conceptual model that provides an underlying structural explanation for their emergence. In doing so, we create a powerful visual tool that can be used to ground discussions about the impact artificial agents have on organizations and humans within them.

Abstract

Instructional texts for specific target groups should ideally take into account the prior knowledge and needs of the readers in order to guide them efficiently to their desired goals. However, targeting specific groups also carries the risk of reflecting disparate social norms and subtle stereotypes. In this paper, we investigate the extent to which how-to guides from one particular platform, wikiHow, differ in practice depending on the intended audience. We conduct two case studies in which we examine qualitative features of texts written for specific audiences. In a generalization study, we investigate which differences can also be systematically demonstrated using computational methods. The results of our studies show that guides from wikiHow, like other text genres, are subject to subtle biases. We aim to raise awareness of these inequalities as a first step to addressing them in future work.

Abstract

Potential benefits and risks related to mainstream social media platforms and their revenue model are vigorously debated. However, a comprehensive framework of performance criteria to evaluate social media platforms and suggestions for transforming them are rare. Employing a transdiscipli­ nary approach, the present work aimed to close these gaps through semi-structured interviews with experts from academia and industry, coupled with exploratory thematic content/topic analysis.