#2 From Filters to Body Positivity: Opposing Social Media Messages and Adolescent Body Image – Comprehensive Study Notes

Ubiquity of social media among adolescents
- \approx 95\% of Flemish adolescents use smartphones daily (Vanwynsberghe et al., 2022)
- Established link between social-media exposure and body-image concerns (Fardouly & Vartanian, 2016; Saiphoo & Vahedi, 2019)
- Adolescents (ages 12\text{–}18) are developmentally susceptible
- Pubertal body changes ↔ identity formation (Erickson, 1968; Williams & Currie, 2000)
- Appearance salient to self-concept (Kling et al., 2018)
Dual narrative online
- Idealized images: thin, tall, young, muscular, etc.
- Counter-ideal / body-positive content: diversity in body size/ethnicity, hashtags such as #plussizefashion, “love your body”
- Need to study simultaneous exposure to opposing messages rather than single-message effects

Body dissatisfaction: negative feelings about one’s body when actual ≠ ideal (Higgins, 1987)
Idealized content: depictions that glorify narrow beauty standards
Counter-idealized (body-positive) content: imagery/text challenging those standards
Message Interpretation Process (MIP) Model (Austin & Meili, 1994)
- Media processing = active, meaning-making negotiation
Media Practice (MP) Model (Steele & Brown, 1995)
- Media selection, interaction, application embedded in social context
Resonance: fit between media message and personal/social lens (Cultivation theory extensions; Shrum, 2017)
Peer appearance culture: frequency/intensity of appearance-focused conversations among friends

H_1 Idealized > counter-idealized exposure \Rightarrow higher body dissatisfaction
H_2 Counter-idealized > idealized exposure \Rightarrow lower body dissatisfaction
H_3 Balanced exposure (ideal ≈ counter-ideal) \Rightarrow no association (or lowest dissatisfaction)
H_4 Social resonance moderates above; higher peer appearance culture intensifies effects, especially when idealized content dominates

Idealized exposure (11 items)
- Likert 1=\text{never}\;\rightarrow\;6=\text{constantly}
- Internal consistency \alpha = .954
- Example: “How often do you see people who are thin, tall, young?”
Counter-idealized exposure
- Visual (11 items) \alpha = .911; one-factor (EV =5.638; 56.4\% variance)
- Example: “How often do you see people who are fat, short, not muscled?”
- Textual (9 items) \alpha = .930; one-factor (EV =5.664; 62.9\% variance)
- Examples: messages on subjective beauty, photo editing reveals, #loveyourbody
Body Dissatisfaction
- Body-Esteem Scale—Appearance subscale (10 items, 1=\text{never}\rightarrow5=\text{always})
- \alpha = .925; reverse-coded positives
Peer Appearance Culture (5 items; \alpha = .882)
- “My friends and I talk about what we would like our bodies to look like.”
Covariates: BMI (self-reported), age, biological sex

Centering: grand-mean for idealized (I) and counter-idealized (C) variables to curb multicollinearity
Polynomial regression + Response Surface Analysis (RSA) (Edwards & Parry, 1993)
- Regression terms: I, C, I^2, C^2, I \times C
- Separate models for visual vs. textual counter-ideal content
Moderated polynomial regression for social resonance W
- Stepwise: controls → I-C, I+C, I\times C → W → interactions with W
Gender-specific analyses after measurement invariance testing (Mplus 8.7)

Mean idealized exposure M = 4.62 (on 6-pt scale) > counter-ideal exposure
Girls vs. boys
- Girls report higher idealized, textual & visual counter-ideal exposure, body dissatisfaction, peer appearance culture, BMI; lower age
- No gender diff in visual counter-ideal exposure
Platform correlations
- Idealized content ↔ Instagram, Snapchat, TikTok use (positive)
- Visual counter-ideal ↔ TikTok use (positive)

Linear effects
- Idealized \beta = .152 (p = .014) ↑ dissatisfaction
- Visual counter-ideal \beta = -.177 (p = .004) ↓ dissatisfaction
RSA surface
- Curvature along incongruence line I = -C: B = 0.18 (p < .001) → U-shape
- Lowest dissatisfaction when I \approx C (balanced)
- Slope negative B = -0.28 (p < .001): when I > C, dissatisfaction spikes more than when C > I

Linear idealized effect significant; textual marginal
RSA curvature B = 0.25 (p < .001); slope B = -0.26 (p < .001)
- Same pattern: balance best; idealized dominance worst

Measurement invariance achieved for most constructs (scalar invariance failed for counter-idealized exposure)
Girls
- Visual model R^2 = .176 (p = .011) → same U-shape; slope B = -0.32
- Textual model marginal R^2 = .145 (p = .087); slope significant only
Boys: polynomial models non-significant R^2 < .06

Direct effect: Peer appearance culture \beta = .330 (p < .001) ↑ dissatisfaction
No moderation: \Delta R^2 non-significant for both visual and textual models

Adolescents navigate mixed media diets; their personal balance of ideal vs. counter-ideal exposure predicts body image
Balanced exposure functions as protective: supports H3; dominance of idealized supports H1; dominance of counter-ideal partially supports H_2 (significant only for visual)
Visual body-positive images may be more impactful than textual messages (imagery more attention-grabbing / emotive)
TikTok identified as a key venue for both idealized and diverse content
Girls appear more sensitive; boys’ relationships weak or absent (sample size caveat)
Peer appearance talk heightens dissatisfaction universally but does not alter content-balance effect (contrary to H_4)

Encourage platforms/educators to promote diverse, body-positive visuals alongside unavoidable ideal imagery
Media-literacy interventions should stress curating a balanced feed
Parental/peer discussions could shift from appearance critique to supportive commentary to reduce overall resonance of harmful ideals
Policymakers can incentivize creators to include diverse bodies (e.g., advertising standards, platform algorithms)

Cross-sectional → no causal inference; longitudinal/experimental (incl. eye-tracking) needed
Self-report measures susceptible to bias; observational content-analysis of individual feeds suggested
Scalar invariance issues for counter-ideal measures between genders → refine items
Did not differentiate resonance for ideal vs. counter-ideal talk; future tools should separate
Incorporate pubertal status, trait self-esteem, and other identity variables in modeling