Vom 2. bis 4. April 2025 findet in Siegen die Pre-CHI 2025 statt. Forschende aus Deutschland und den angrenzenden Ländern haben dort die Gelegenheit, ihre akzeptierten CHI-Paper zu präsentieren und zu diskutieren. Auch Aditya Kumar Purohit, Artur Solomonik, Besjon Cifliku, Hendrik Heuer und Jasmin Baake vom CAIS-Forschungsprogramm „Design vertrauenswürdiger Künstlicher Intelligenz“ nehmen an der Veranstaltung teil und stellen ihre Arbeiten vor.
Über die Pre-CHI
Die ACM (Association of Computing Machinery) CHI-Konferenz zu Human Factors in Computing Systems ist die führende internationale Konferenz im Bereich der Human-Computer Interaction und findet vom 26. April bis 1. Mai 2025 in Yokohama, Japan statt. Die Pre-CHI dient als Plattform zur Präsentation und Diskussion der akzeptierten CHI-2025-Paper. Besonders für die jüngere HCI-Community ist die Veranstaltung eine wertvolle Gelegenheit, sich mit den Anforderungen einer internationalen Top-Konferenz wie der CHI vertraut zu machen. Die zweitägige Veranstaltung umfasst Präsentationen, Diskussionen, Workshops, Podiumsdiskussionen und Demo-Sessions. Zudem wird ausreichend Zeit für informellen Austausch beim Mittag- oder Abendessen sowie bei Laborführungen geboten. Organisiert wird die diesjährige Pre-CHI von der Universität Siegen in Zusammenarbeit mit dem Institut für Sozialinformatik Bonn (IISI) und der gemeinnützigen Gesellschaft für Digitale und Nachhaltige Zusammenarbeit (DNZ).
Beiträge der CAIS-Forschenden
Die folgenden Paper unter Beteiligung von CAIS-Forschenden werden bei der Pre-CHI präsentiert:
„Lost in Moderation: How Commercial Content Moderation APIs Over- and Under-Moderate Group-Targeted Hate Speech and Linguistic Variations“.
Autoren: David Hartmann (Weizenbaum-Institut Berlin, Technische Universität Berlin), Amin Oueslati (Hertie School Berlin), Dimitri Staufer (Technische Universität Berlin), Lena Pohlmann (Weizenbaum-Institut Berlin, Technische Universität Berlin), Simon Munzert (Hertie School Berlin), Hendrik Heuer (Center for Advanced Internet Studies CAIS, Universität Wuppertal)
Abstract: Commercial content moderation APIs are marketed as scalable solutions to combat online hate speech. However, the reliance on these APIs risks both silencing legitimate speech, called over-moderation, and failing to protect online platforms from harmful speech, known as under-moderation. To assess such risks, this paper introduces a framework for auditing black-box NLP systems. Using the framework, we systematically evaluate five widely used commercial content moderation APIs. Analyzing five million queries based on four datasets, we find that APIs frequently rely on group identity terms, such as ‘black’“, to predict hate speech. While OpenAI’s and Amazon’s services perform slightly better, all providers under-moderate implicit hate speech, which uses codified messages, especially against LGBTQIA+ individuals. Simultaneously, they over-moderate counter-speech, reclaimed slurs and content related to Black, LGBTQIA+, Jewish, and Muslim people. We recommend that API providers offer better guidance on API implementation and threshold setting and more transparency on their APIs‘ limitations.
Link zum Paper: https://arxiv.org/abs/2503.01623
„Social Media for Activists: Reimagining Safety, Content Presentation, and Workflows“
Autoren: Anna Ricarda Luther (Institut für Informationsmanagement Bremen GmbH, Universität Bremen), Hendrik Heuer (Center for Advanced Internet Studies (CAIS), Universität Wuppertal), Stephanie Geise (Zentrum für Medien-, Kommunikations- und Informationsforschung (ZeMKI), Universität Bremen), Sebastian Haunss (Forschungszentrum Ungleichheit und Sozialpolitik (SOCIUM), Universität Bremen), Andreas Breiter (Institut für Informationsmanagement Bremen GmbH, Universität Bremen)
Abstract: As people engage with the social media landscape, popular platforms rise and fall. As current research uncovers the experiences people have on various platforms, rarely do we engage with the sociotechnical migration processes when joining and leaving them. In this paper, we asked 32 visitors of a science communication festival to draw out artifacts that we call Social Media Journey Maps about the social media platforms they frequented, and why. By combining qualitative content analysis with a graph representation of Social Media Journeys, we present how social media migration processes are motivated by the interplay of environmental and platform factors. We find that peer-driven popularity, the timing of feature adoption, and personal perceptions of migration causes – such as security – shape individuals‘ reasoning for migrating between social media platforms. With this work, we aim to pave the way for future social media platforms that foster meaningful and enriching online experiences for users.
Link zum Paper: https://arxiv.org/abs/2503.12924
“Scrolling in the Deep: Analysing Contextual Influences on Intervention Effectiveness during Infinite Scrolling on Social Media”
Autoren: Luca-Maxim Meinhardt (Universität Ulm, Maryam Elhaidary (Universität Ulm), Mark Colley (Universität Ulm, UCL Interaction Centre, London), Michael Rietzler (Universität Ulm), Jan Ole Rixen (Universität Ulm, Karlsruher Institut für Technologie), Aditya Kumar Purohit (Center for Advanced Internet Studies (CAIS)), Enrico Rukzio (Universität Ulm)
Abstract: Infinite scrolling on social media platforms is designed to encourage prolonged engagement, leading users to spend more time than desired, which can provoke negative emotions. Interventions to mitigate infinite scrolling have shown initial success, yet users become desensitized due to the lack of contextual relevance. Understanding how contextual factors influence intervention effectiveness remains underexplored. We conducted a 7-day user study (N=72) investigating how these contextual factors affect users‘ reactance and responsiveness to interventions during infinite scrolling. Our study revealed an interplay, with contextual factors such as being at home, sleepiness, and valence playing significant roles in the intervention’s effectiveness. Low valence coupled with being at home slows down the responsiveness to interventions, and sleepiness lowers reactance towards interventions, increasing user acceptance of the intervention. Overall, our work contributes to a deeper understanding of user responses toward interventions and paves the way for developing more effective interventions during infinite scrolling.
Link zum Paper: https://arxiv.org/pdf/2501.11814
„Social Media Journeys – Mapping Platform Migration“ (Late Breaking Paper)
Autoren: Artur Solomonik (Center for Advanced Internet Studies (CAIS)), Hendrik Heuer (Center for Advanced Internet Studies (CAIS), Universität Wuppertal)
Abstract: As people engage with the social media landscape, popular platforms rise and fall. As current research uncovers the experiences people have on various platforms, rarely do we engage with the sociotechnical migration processes when joining and leaving them. In this paper, we asked 32 visitors of a science communication festival to draw out artifacts that we call Social Media Journey Maps about the social media platforms they frequented, and why. By combining qualitative content analysis with a graph representation of Social Media Journeys, we present how social media migration processes are motivated by the interplay of environmental and platform factors. We find that peer-driven popularity, the timing of feature adoption, and personal perceptions of migration causes – such as security – shape individuals‘ reasoning for migrating between social media platforms. With this work, we aim to pave the way for future social media platforms that foster meaningful and enriching online experiences for users.
Link zum Paper: https://arxiv.org/abs/2503.12924
„This could save us months of work“ – Use Cases of AI and Automation Support in Investigative Journalism“ (Late Breaking Paper)
Autoren: Besjon Cifliku (Center for Advanced Internet Studies (CAIS)), Hendrik Heuer (Center for Advanced Internet Studies (CAIS), University of Wuppertal)
Abstract: As the capabilities of Large Language Models (LLMs) expand, more researchers are studying their adoption in newsrooms. However, much of the research focus remains broad and does not address the specific technical needs of investigative journalists. This paper presents several applied use cases where automation and AI intersect with investigative journalism. We conducted a within-subjects user study with eight investigative journalists. In interviews, we elicited practical use cases for automation and presented a prototype that combines LLMs and Programming-by-Demonstration (PbD) to simplify data collection on numerous websites. Based on user reports, we classified the journalistic processes into data collecting and reporting. Participants indicated they utilize automation to handle repetitive tasks like content monitoring, web scraping, summarization, and preliminary data exploration. We provide guidelines on how investigative journalism can benefit from AI and automation.
Link zum Paper: https://arxiv.org/abs/2503.16011