Effect of online video infotainment on audience attention

The paper addresses how infotainment characteristics in online news videos influence audience attention within the attention economy, where human attention is scarce and monetizable. With rapid growth in social media use and the prevalence of online video (e.g., YouTube, Bilibili), media outlets increasingly rely on video content and platform algorithms to compete for attention. The study notes a gap in empirical analyses linking infotainment features to measurable attention outcomes in online video. The research objective is to investigate the influence of infotainment—across content and presentation dimensions—on multiple facets of audience attention (breadth, depth, engagement, validity) using video-level indicators from a major social video platform.

Grounded in attention economy theory, the review outlines how quantified social media metrics (views, likes, shares, comments) serve as proxies of attention and commercial value. Psychological perspectives suggest attention is shaped by object features and activity context, which media can engineer through content and presentation choices. Celebrity influence can restructure attention markets by commodifying personas and signaling popularity. Infotainment—blending information and entertainment—has evolved to include soft news, personalization, emotionalization, sensationalism, and stylistic choices that aim to attract broader audiences. Prior work documents the role of emotions, humor, and provocative headlines in boosting engagement, and the migration of news consumption to social platforms. The paper integrates disparate variables (headline features, celebrity presence, news value, presentation richness) under a unified infotainment framework spanning content (e.g., emotional polarity, storytelling, topic hardness/softness) and presentation (e.g., duration, multimodality, tags/hashtags, media source).

Design: Empirical, cross-sectional study using linear regression to assess relationships between infotainment features and four dimensions of audience attention. Dependent variables (all log-transformed after adding 1): breadth (y1 = log(playbacks+1)); depth (y2 = log(0.289×likes + 0.422×coins + 0.289×collections +1)); engagement (y3 = log(0.688×bullet screen comments + 0.312×comments +1)); validity (y4 = log(shares+1)). Weights for composite dimensions were derived via principal component analysis (maximum variance method).

Data source and sampling: Bilibili, a major Chinese video-based social media platform with bullet-screen interactions. Videos were drawn from four information partitions (hot spots, society, global, comprehensive). Following popularity rankings, the top 600 videos per sub-partition from 05/01/2022 to 08/01/2022 were selected (n=2400); after excluding unavailable live links, 2395 videos remained. Popularity on Bilibili is computed by a weighted formula: hot = 0.25×playbacks + 0.4×likes + 0.3×collections + 0.4×coins + 0.4×bullet_screen_comments + 0.4×comments + 0.6×shares.

Independent variables and coding: Emotional polarity (positive, negative) based on event valence (1=yes, 0=no); celebrities presence scored 0 (none), 1 (name in title or label), 2 (name in both); content storytelling summed presence of conflict, surprise, drama (0–3); news topic softness (soft=1, hard=0); sensational headline scored by count of features (e.g., emotive punctuation, colloquial/online language, eye-catching keywords, ellipses, colons/citations, curiosity-triggering incompleteness, exaggerated causality); duration (seconds) representing time fragmentation; diverse presentations counted presence of modalities/elements (e.g., video, photos, background music, special sound effects, concurrent vocals, subtitles, emphatic fonts, explanatory text); number of topics (hashtags/# or triangle-dot topics); number of tags (labels); media nature (authoritative media=1; mass/self-media=0). Control variables: number of followers (channel fan count), days since release (relative to 08/22/2022 crawl date), use-time peak (4–11 p.m., later dropped), workday vs. non-workday (later dropped).

Statistical analysis: Linear regressions estimated effects on four attention dimensions. Multicollinearity was low (all VIF < 2). Use-time peak and workday were removed for insignificance across models.

Descriptive insights: Most sampled videos were short (≤120s: ~72.9%), with release activity concentrated between 10:00–20:00. Authoritative media produced most videos (76.6%). About 44% contained clear emotional polarity; 57.5% showed storytelling features; 50.5% had sensational headlines (score ≥2). Tags were common; two-thirds used topics/hashtags at least once.

Regression highlights (beta, significance):

• Emotions: Positive emotion negatively affected breadth (β=-0.06**), depth (β=-0.169***), engagement (β=-0.076***), and validity (β=-0.104***). Negative emotion positively affected breadth (β=0.058**), engagement (β=0.125***), validity (β=0.16***), but negatively affected depth (β=-0.075***). Thus, H1a not supported; H1b partially supported.
• Celebrities: Positive effects on breadth (β=0.05*) and validity (β=0.064**); H2 supported.
• Content storytelling: Positive effects on breadth (β=0.1***), depth (β=0.129***), engagement (β=0.065**); H3 supported.
• News topics: Soft topics increased breadth (β=0.047*); H4a not supported, H4b supported.
• Sensational headlines: Increased engagement (β=0.044*); H5 supported.
• Duration (time fragmentation): Negative on depth (β=-0.515***); positive on engagement (β=0.317***) and validity (β=0.078**); H6 partially supported.
• Diverse presentations: Positive on breadth (β=0.114***), depth (β=0.114***), engagement (β=0.047*); H7 supported.
• Number of tags: Positive on depth (β=0.089***), engagement (β=0.047*), validity (β=0.078***); H8 supported.
• Number of topics: Negative across breadth (β=-0.094***), depth (β=-0.102***), engagement (β=-0.058**), validity (β=-0.127***); H9 not supported.
• Media nature (authoritative): Positive on breadth (β=0.065**) and validity (β=0.134***), negative on depth (β=-0.053**); H10a/H10b partially supported.
• Controls: Followers positively associated with all dimensions (breadth β=0.135***; depth β=0.051**; engagement β=0.166***; validity β=0.175***). Days since release negatively related to breadth (β=-0.065**) and validity (β=-0.063**), positively related to depth (β=0.115***); engagement ns.

Hypothesis summary: H2, H3, H5, H7, H8, H4b supported; H1a, H4a, H9 not supported; H1b, H6, H10a, H10b partially supported.

Findings indicate infotainment affects audience attention heterogeneously across dimensions. Negative emotions promote participation and sharing (engagement, validity) but reduce depth (likes/coins/collections), while positive emotions suppress attention broadly, suggesting neutral tone may better elicit positive evaluative feedback. Celebrity presence broadens reach and encourages sharing, consistent with influencer effects. Storytelling enhances breadth, depth, and engagement by making content more recognizable and discussable. Soft news expands breadth, signaling audience appetite for approachable topics alongside hard news. Sensational headlines raise engagement by leveraging emotive/curiosity cues. Shorter duration boosts engagement and sharing but can undermine depth; longer videos may support deeper appreciation. Multimodal, richly presented videos elevate breadth, depth, and engagement by intensifying sensory stimulation and clarity. More tags facilitate discovery and interactivity, increasing depth, engagement, and validity, whereas too many topics dilute focus and distract audiences, reducing performance across all dimensions. Authoritative media increase breadth and cross-platform sharing (validity) through trust and credibility, whereas self/media with more subjective color may garner more depth-type responses. Overall, infotainment should be applied moderately and purposefully, aligning tactics with desired attention dimensions while safeguarding objectivity to avoid drift toward misinformation.

The study unifies diverse determinants of online news video performance under an infotainment framework and operationalizes audience attention into four measurable dimensions: breadth, depth, engagement, and validity. Using a large sample from Bilibili and regression analysis, it shows that storytelling, celebrity presence, soft topics, sensational headlines, rich multimodality, and tagging generally enhance attention, while excessive topics and positive emotional framing can hinder outcomes; negative emotional framing has mixed effects, and shorter duration trades depth for engagement and sharing. Practically, publishers should emphasize narrative crafting, judicious sensationalism, multimodal enrichment, and effective tagging while limiting topic sprawl and calibrating emotional tone and length to target specific attention dimensions. Normatively, moderation is essential to avoid distortion and misinformation risks. Future work should broaden platforms and include non-popular videos to test generalizability and causal mechanisms.

The dataset comprises only popular videos from a single platform (Bilibili), limiting generalizability. Non-popular content and other platforms were not analyzed. Temporal dynamics were cross-sectional, and some coding relies on content judgments. Future research should include non-popular videos, multiple platforms, and designs addressing causality.