Auflistung nach Autor:in "Wulff, Dirk U."

Gerade angezeigt 1 - 3 von 3

Affect in science communication: a data-driven analysis of TED Talks on YouTube
(Springer, 2024) Fischer, Olivia; Jeitziner, Loris Tizian; Wulff, Dirk U.
Science communication is evolving: Increasingly, it is directed at the public rather than academic peers. Understanding the circumstances under which the public engages with scientific content is therefore crucial to improving science communication. In this article, we investigate the role of affect on audience engagement with a modern form of science communication: TED Talks on the social media platform YouTube. We examined how two aspects of affect, valence and density are associated with public engagement with the talk in terms of popularity (reflecting views and likes) and polarity (reflecting dislikes and comments). We found that the valence of TED Talks was associated with both popularity and polarity: Positive valence was linked to higher talk popularity and lower talk polarity. Density, on the other hand, was only associated with popularity: Higher affective density was linked to higher popularity—even more so than valence—but not polarity. Moreover, the association between affect and engagement was moderated by talk topic, but not by whether the talk included scientific content. Our results establish affect as an important covariate of audience engagement with scientific content on social media, which science communicators may be able to leverage to steer engagement and increase reach.
01A - Beitrag in wissenschaftlicher Zeitschrift
Exploring linguistic indicators of social collaborative group engagement
(International Society of the Learning Sciences, 2023) Jeitziner, Loris Tizian; Paneth, Lisa; Rack, Oliver; Zahn, Carmen; Wulff, Dirk U.; Damşa, Crina; Borge, Marcela; Koh, Elizabeth; Worsley, Marcelo
This study takes a NLP approach to measuring social engagement in CSCL-learning groups. Specifically, we develop linguistic markers to capture aspects of social engagement, namely sentiment, responsiveness and uniformity of participation and compare them to human ratings of social engagement. We observed small to moderate links between NLP-markers and human ratings that varied in size and direction across the different groups. We discuss measurement and prediction of social collaborative group engagement using natural language processing.
04B - Beitrag Konferenzschrift
Insights from a multi-method assessment of collaborative engagement in student group
(27.09.2024) Jeitziner, Loris Tizian; Paneth, Lisa; Rack, Oliver; Wulff, Dirk U.; Zahn, Carmen
We investigate the Quality of Collaborative Group Engagement (QCGE) in Computer Supported Collaborative Learning (CSCL) employing a multimethod approach. Analyzing 38 triad groups the study combines advanced methods such as video analysis (verbal and nonverbal behavior self assessment, trained observer ratings, and natural language processing (NLP )). The results produced key insights into QCGE . First, t he observer rating s and self assessments exhibited limited variance and considerable skewness in most QCGE dimensions , significantly limiting their usefulness. Second, no nverbal behavior s and linguistic markers extracted using NLP showed small to moderate correlations with QCGE ratings, suggesting opportunities for measuring QCGE in an automatized fashion . Our study emphasizes the importance of multimethod approaches for understanding QCGE and highlights a potential to refine these methodologies using artificial intelligence to increase the accuracy and reliability of QCGE assessment.
06 - Präsentation