Transparency data

Expert Panel for Growing up in the Online World: summary meeting minutes

Published 15 June 2026

Panel Meeting 1 – 5 May 2026

Policy context and key questions

The panel discussed the broader policy context and priorities for advice.

• The consultation aims to understand both risks and opportunities of digital technologies for children, including harm reduction and positive outcomes.
• Priority questions include:
  • How interventions align with developmental stages
  • The effectiveness of age-, feature-, and time-based restrictions
  • Impacts on vulnerable groups and equality of opportunity

The panel also discussed desired indicators of success:

• Outcomes of interest include mental health, sleep, physical activity, and educational attainment, alongside broader wellbeing.
• There is a need to assess both benefits and unintended consequences of interventions.

It was noted that:

• The Online Safety Act provides the regulatory foundation, with new policies expected to complement existing measures.
• Policy development will need to proceed relatively quickly to align with statutory timelines.

Structuring the evidence base

The panel discussed how to organise and interpret the evidence on online harms.

• There is a need for robust taxonomies of harms and logic models or driver diagrams linking digital activity to outcomes.
• Existing frameworks (for example, from regulators and international organisations) can provide a starting point but may need adaptation.

Key challenges identified included:

• Complexity in defining age and developmental stages, with limited academic consensus but a practical need for policy clarity.
• The importance of recognising differences between children, families, and contexts, rather than assuming uniform effects.
• Significant evidence gaps, particularly around unintended consequences and impacts on different groups.

The panel supported combining multiple approaches (taxonomy, logic models, theory of change) to improve policy relevance.

Evidence on “addiction” and problematic use

The panel briefly discussed the concept of addiction in relation to digital technologies.

• There are ongoing concerns about the use of terms such as “addictiveness”, with uncertainty about how well these map onto evidence.
• The evidence on technology addiction and compulsion remains complex and contested, with no settled view.

This led to agreement to explore these issues further, including through dedicated subgroup work.

International evidence and case studies

The panel reviewed emerging evidence from international policy interventions.

Australian social media ban

• Early evaluation includes multiple waves of data collection and mixed methods, but faces methodological limitations (for example, short timeframes, limited comparison data).
• Initial findings suggest a reduction in reported account ownership among younger users, though attribution remains uncertain.
• There are ongoing concerns about platform compliance and enforcement effectiveness.

The panel emphasised caution:

• Evidence is early and difficult to interpret, and there is risk of drawing premature conclusions.

Broader international approaches

• Discussion referenced frameworks in Europe and the United States, including regulatory and rights-based approaches.
• Several cross-cutting issues were highlighted:
  • The importance of effective age assurance
  • Questions about who bears responsibility for compliance
  • The balance between protection and access to benefits

The panel also noted:

• The need to ensure children and young people’s perspectives are reflected in policy development.
• A view that platforms should bear responsibility for demonstrating safety and effectiveness, drawing comparisons with product safety regimes.

Key cross-cutting themes

Across the discussion, several common themes emerged:

• The importance of balancing harms and benefits of digital technology.
• The need for stronger evidence frameworks to support policy decisions.
• Recognition of heterogeneity across children and contexts, limiting one-size-fits-all approaches.
• Challenges in evaluating interventions, particularly given limited data and complex causal pathways.

Panel Meeting 2 – 27 May 2026

Features and functionalities influencing engagement

The panel discussed the evidence on which design features drive engagement and how they work.

• There is limited direct causal evidence linking specific features to observed behaviour, partly due to restricted data access.
• Some experimental evidence suggests that introducing friction (for example, pauses before app use or requiring users to pre-commit to time spent) can reduce usage, though impacts on wellbeing are unclear.

Several broader dynamics were highlighted:

• Many commonly cited features (for example, streaks, gamification, short-form content) are widely adopted across platforms, suggesting effectiveness in driving engagement.
• Platform business models are strongly oriented towards maximising attention and time spent, and features evolve in line with that objective.
• Features initially perceived as harmful may also have beneficial applications (for example, in education or motivation), complicating regulatory approaches.

The panel also noted:

• A shift in some digital environments from intrinsic motivation (enjoyment) to extrinsic motivators (for example, rewards, streaks), which may have different wellbeing implications.
• Certain features (for example, streak-based interactions) may create pressure to maintain engagement, contributing to anxiety or dependency for some users.
• The design and replication of features across platforms may reflect tested effectiveness, though this cannot be independently verified due to lack of data access.

Emerging technologies were also discussed:

• AI chatbot features may raise risks where users prefer agreeable or reinforcing interactions, or where systems are perceived as authoritative.
• At the same time, AI may offer significant benefits, including educational support.

Relationship between use and exposure to harm

The panel explored how different levels and patterns of use relate to outcomes.

• There is evidence that very high levels of use are associated with negative outcomes, often through displacement of other activities such as sleep, education, or in-person interaction.
• For moderate use, the relationship with harm is less clear and may depend on context, content, and individual characteristics.

Key themes included:

• The importance of understanding what activities are displaced, rather than focusing only on time spent.
• Ongoing uncertainty about causation versus correlation, with many findings difficult to interpret definitively.
• Limited evidence of a clear time-based threshold beyond which harm occurs.

The panel also explored wider dynamics:

• Changing social norms, including expectations around constant availability and rapid responses, may contribute to stress and disrupted sleep.
• High usage may reflect or interact with underlying vulnerabilities, rather than acting as a sole causal factor.
• Some evidence links higher levels of use in adolescence with poorer outcomes later, though causal pathways remain unclear.

Several conceptual interpretations were discussed:

• Moving beyond simple causation or correlation towards understanding pathways to harm, including how technology may facilitate or amplify existing risks.
• Recognition that online environments can increase the scale and speed of harmful content exposure, compared to offline contexts.

The panel also noted that:

• Interventions such as bans or curfews could have unintended consequences, including limiting access to support or encouraging workarounds (for example, VPN use).

Features most associated with harm

The panel considered which specific features or mechanisms may present higher risks.

• There is limited conclusive evidence that specific features are inherently addictive; instead, engagement may be driven by content relevance and enjoyment, supported by low-friction design.
• Personalisation systems may increase engagement by delivering increasingly tailored or extreme content, particularly where this aligns with peer group norms.

Evidence and hypotheses discussed included:

• Non-personalised feeds may reduce usage more effectively than surface-level design changes (for example, removing notifications).
• Features such as location sharing may introduce risks (for example, exploitation by bad actors), even when intended as safety tools.
• Some features may shape social norms, such as normalising surveillance or constant connectivity in relationships.

The panel also emphasised:

• The importance of considering perpetrators, not only victims, including how certain features may enable harmful behaviour (for example, exploitation of ephemeral messaging or anonymity).
• The need to draw on law enforcement and safeguarding evidence to understand how platforms are used in practice for harm.

A consistent theme was:

• A significant lack of platform data limiting the ability to robustly assess how specific features impact behaviour and outcomes.

Cross-cutting insights

Across the discussion, several common themes emerged:

• Platform design and business models play a central role in shaping engagement patterns.
• The relationship between use and harm is complex, context-dependent, and not reducible to simple metrics such as time spent.
• Both features and content are important, and are often interdependent.
• There is a need to focus on mechanisms and pathways to harm, rather than relying on binary or simplified models.
• Evidence gaps remain substantial, particularly due to limited access to platform data.

Subgroup 1 – Addiction – 3 June 2026

Conceptualising “addiction” and “addictiveness”

The panel discussed the scientific and clinical validity of applying terms such as “addiction” and “addictiveness” to social media and gaming.

• The panel noted that clinical addiction is a narrow concept, involving significant functional impairment (for example, disruption to education, relationships, or daily life), and should not be conflated with high levels of engagement.
• Common measures and self-reports may capture anxiety about technology use or perceived loss of control, rather than clinically meaningful behaviour.
• The use of “addiction” in everyday language may reflect broader societal concern about technology, rather than a precise clinical phenomenon.

The panel also discussed adolescents’ self-perceptions:

• Many young people report feeling “addicted”, but this may be better understood as perceived overuse, reduced agency, or difficulty disengaging, rather than clinical addiction.
• These experiences are heterogeneous, reflecting different behaviours and risks.
• Language used by adults may shape how young people interpret and describe their own behaviour.

Platform design and engagement

The panel explored how platform design influences user engagement and potential overuse.

• Features such as infinite scroll, autoplay, and personalised recommendations reduce friction and remove stopping cues, enabling extended use.
• These low-friction environments may disproportionately affect users in more vulnerable or low-control states.
• The discussion emphasised that many services are designed to maximise time spent and user attention, reflecting underlying business models.

The panel considered how best to describe these dynamics:

• There was scepticism about the phrase “addictive by design”, with preference for terms such as “persuasive” or “habit-forming” design, which are more consistent with current evidence.
• Evidence suggests concern around problematic engagement but does not support strong claims of intentionally engineered clinical addiction.
• Some design features may also have beneficial uses (for example, in education or health contexts), highlighting the need to avoid overly broad conclusions.
• A key emerging perspective was that the central policy issue may be reduced user autonomy, rather than addiction per se.

Defining and identifying problematic use

The panel discussed challenges in defining and identifying problematic digital behaviour.

• There is no single agreed definition of problematic use, and it is difficult to distinguish harm from behaviour typical of adolescence.
• More credible indicators include functional impacts, such as effects on sleep, education, or relationships.
• However, some of these patterns may reflect general adolescent behaviour, not effects unique to technology.

Additional points raised included:

• Technology use patterns are often relatively stable, with users substituting between platforms rather than escalating in a way typical of addiction.
• Increased engagement may raise exposure to harmful content, which may be a more actionable policy concern.
• There is limited consensus on diagnostic thresholds and significant variation in prevalence estimates.

The panel discussed whether policy should focus on severe cases or adopt a broader public health approach but did not reach a firm conclusion.

Interventions and policy options

The panel reviewed potential interventions, noting that the evidence base remains limited and mixed.

• Existing usage-reduction tools often show limited long-term effectiveness, with possible rebound effects.
• Transparency around recommendation systems was highlighted as a priority to enable scrutiny and accountability.
• Some design-focused interventions (for example, restoring friction or requiring user action) may help reduce excessive engagement.

The panel also noted:

• A clearer focus may be needed on specific high-risk features or exposures, rather than broad concepts such as “addiction”.
• Educational and media literacy approaches have weak evidence and may benefit more advantaged groups disproportionately.
• A major constraint is lack of access to platform data, limiting understanding of impacts and effective interventions.

Language and framing

Throughout the discussion, the panel emphasised the importance of careful language.

• Overuse of “addiction” risks pathologising normal behaviour, weakening the meaning of clinical addiction, and potentially undermining agency.
• However, the term is strategically powerful in legal and regulatory contexts, as it implies harm and corporate responsibility.

Alternative framings were explored:

• “Persuasive design” and “habit-forming design” were seen as more accurate and flexible, allowing both beneficial and harmful uses to be captured.
• A habit-based framework may better support policy focused on shaping specific features, rather than limiting technology overall.

Overall takeaway

The panel did not reach full consensus on terminology, but agreed on several core points:

• The concept of clinical addiction is not well suited to most everyday digital behaviours.
• Problematic engagement and reduced autonomy are more relevant policy concerns.
• Platform design and business models are central drivers of engagement.
• The evidence base for interventions remains limited, particularly due to lack of data access.
• Policy should prioritise clear, precise, and evidence-based language to avoid overstating harms while still enabling effective intervention.

Subgroup 2 – Age and Development – 5 June 2026

Differential impacts by age and development

The panel discussed how impacts vary across ages and developmental stages:

• Longitudinal research suggests age-specific windows of vulnerability, with increased social media use linked to lower life satisfaction at later points, particularly for certain age groups. However, the panel emphasised the need for caution in translating these findings directly into policy.
• Younger and older adolescents experience different risks and benefits, with no consistent pattern across all outcomes.
• Early adolescence is characterised by heightened sensitivity to social dynamics, including peer approval, embarrassment and social comparison.
• Some evidence links higher use to poorer outcomes for a subset of younger adolescents, but effects are not universal.
• Older adolescents may face greater exposure to risks overall, while also gaining more from digital participation.
• Children are divided on restrictions, with concerns about reduced access to social and learning opportunities.
• Digital behaviours are not platform-specific; children move across services, suggesting a need to focus on shared features and pathways.

Interventions across developmental stages

The panel explored how interventions may vary in effectiveness by age:

• Evidence is limited on how specific features affect different age groups, particularly younger users.
• Evidence suggests that screen use after bedtime increases with age, particularly among mid-to-older adolescents.
• Ages 14 to 16 may be a critical period, where curfews or time-based restrictions could be more impactful.
• Younger children may be more receptive to restrictions.
• Older adolescents are more likely to resist, conceal, or circumvent controls.
• Imperfectly implemented restrictions may lead users to migrate to other platforms or workarounds.

Overall, the panel noted that interventions should be developmentally tailored.

Implications of age-based restrictions (“cliff edge”)

The panel discussed the potential impact of restricting access below a threshold age:

• There is very limited direct evidence on “cliff edge” effects.
• Concerns were raised that a sudden transition could lead to intensive uptake or increased risks, though this remains speculative.

Alternative approaches were considered:

• Gradual or phased access may better support development of self-regulation.
• Co-use with parents and guided exposure could build capability, though evidence is limited.
• Digital literacy and parental support are likely to be important alongside any restriction model.

Cross-cutting insights

• Impacts are highly dependent on age, development, and individual context.
• There is no clear evidence for a single optimal threshold or intervention.
• Developmental factors are central to both risk and benefit.
• Interventions must account for behavioural responses and unintended consequences.
• Evidence gaps remain significant, particularly due to limited platform data.

Panel Meeting 3 – 10 June 2026