Relevance & Research Question

Certain population sub-groups are consistently under-represented or excluded from survey samples, or they may appear in insufficient frequencies. In the UK, where a named sample frame is often unavailable, current strategies for boosting samples can be prohibitively costly or impractical, particularly in self-completion survey formats. As survey research in the UK strives to become more inclusive, it is essential to explore effective methods for incorporating and enhancing the representation of these under-represented subgroups.
Methods & Data

We employed respondent-driven sampling (RDS) to recruit young adults aged 18 to 24 from the UK, using the NatCen online probability-based panel. Participants completed a brief online questionnaire and could recruit up to five peers using unique digital codes, with incentives for participation and successful recruitment. This process spanned multiple waves, aiming for a sample of about 1,500 participants and allowing for an assessment of recruitment cooperation and network characteristics in a self-completion survey format.

Results

We examined the patterns and predictors of recruitment cooperation in RDS using probability-based seeds. The study also assessed the extent of non-response at various recruitment stages and explored how socio-demographic and network factors influence recruitment success. Intra-chain correlation revealed a strong clustering of respondents by ethnicity, while other characteristics, such as gender, had a less pronounced impact on recruitment correlation. Additionally, the study evaluated the effect of seed composition on recruitment outcomes and considered the implications of violating key RDS assumptions for the quality of the inferences made. Notably, incorporating seeds from outside the target population led to the recruitment of a significant proportion of young adults within the sample.

Added Value

This study aims to enhance the evidence base for using RDS as a survey data collection method, both in the UK and internationally, by integrating empirical findings with methodological diagnostics. It also introduces a novel approach designed to amplify the voices of young people, a demographic that remains under-represented in social surveys. This perspective has the potential to make surveys more inclusive and methodologically robust.

Relevance & Research Question Major health organizations have raised alarms about rising youth loneliness (WHO, 2025), with reports suggesting young people experience the highest loneliness rates and saw the sharpest increases between 2018 and 2022 (OECD, 2024). However, reported prevalence varies dramatically across surveys—from 24% to 50.5% for young adults in different European studies—raising a critical question: are we witnessing a genuine loneliness epidemic or methodological artifacts? While genuine increases are plausible given documented declines in face-to-face social contact (OECD, 2025), the magnitude may be partially explained by methodological shifts from traditional face-to-face to web-based data collection. This study examines how survey mode influences loneliness measurement, with particular focus systematic variation by age, potentially distorting our understanding of youth loneliness trends.

Methods & Data This study quantifies data collection mode effects by comparing loneliness reports across web-based, face-to-face, and telephone collection modes in the Survey on Income and Living Conditions. The analysis focuses on identifying age-specific patterns in how survey mode affects loneliness measurement, examining whether the magnitude of mode effects differs between younger and older respondents. To address potential limitations due to mode self-selection, we use a variaty of techniques including controlling for respondent’s health status and limiting the analysis to a subsample of countries. Results from other surveys are included for rubustness (SOEP, Germany). Our research additionaly addresses the role of stigma (national aggregates by age group from the EU Loneliness Survey) to explain observed patterns.

Results The findings reveal striking age-specific mode effects: young adults report significantly more loneliness in online surveys compared to face-to-face interviews, while older adults exhibit the opposite pattern, reporting higher loneliness levels in face-to-face settings. These divergent patterns suggest that stigma associated with loneliness and social desirability bias operate differently across the lifespan.
Added Value By demonstrating that mode effects vary systematically by age, this research helps disentangle genuine increases in youth loneliness from measurement artifacts introduced by the widespread transition to web surveys. The study contributes to research in the field of survey methodology by documenting how demographic characteristics interact with mode effects for (stigmatized) subjective measures.

Relevance & Research Question
Despite unprecedented psychological distress in conflict zones and growing treatment gaps, little is known about population readiness to adopt AI-based therapeutic interventions. Following the October 7th attack, Israel experienced severe mental health crisis with 31.5% reporting treatment needs but only 18.3% seeking help, against healthcare strain and six-month waiting periods. This study examines AI therapy acceptance predictors, addressing: How age and gender influence perceived treatment needs and actual help-seeking? What is the relationship between age and AI therapy readiness? Do socioeconomic factors predict beliefs about AI therapy accessibility?
Methods & Data
We conducted a representative online panel survey of 502 Jewish adults in Israel through Ipanel in June 2025. The survey measured demographic characteristics, perceived need for psychological treatment, actual treatment-seeking behavior, and readiness to seek AI-based therapy using a validated 6-item scale (α=.854). Respondents ranged from ages 18-87 (M=42.66, SD=16.33) with balanced gender distribution. We employed chi-squared tests for demographic-treatment relationships, Pearson correlations for age-readiness associations, independent samples t-tests for age group comparisons (18-40 vs. 41-87), and Spearman correlations for income-related beliefs..
Results
Significant correlations emerged between age and perceived treatment need (χ²=103.88, p=.002) and between gender and perceived need (χ²=7.57, p=.006), with women reporting higher needs (37.2%) than men (25.8%). Age moderated gender effects on actual help-seeking in older adults (χ²=4.00, p=.045). AI therapy readiness showed negative correlation with age (r=-.114, p=.005), with younger adults (M=16.59) demonstrating higher readiness than older adults (M=15.16; t=2.787, p=.003). Higher household income correlated with decreased belief that AI therapy serves those unable to afford treatment (r=-.135, p=.002).
Added Value
This research reveals critical demographic and socioeconomic barriers to AI mental health intervention adoption in crisis contexts. Findings challenge assumptions about universal digital health acceptance, showing those most likely to benefit—older adults and lower-income populations—demonstrate lowest readiness. The substantial gap between perceived need (31.5%) and actual help-seeking (18.3%) highlights intervention opportunities, yet age-based resistance suggests technology alone cannot bridge treatment gaps without addressing acceptance barriers. Results have implications for designing age-appropriate AI interventions, targeting implementation strategies for crisis populations, and understanding how socioeconomic factors shape perceptions of digital therapeutic alternatives during healthcare strain.