Transparency data

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

Published 15 June 2026

This document provides an anonymised summary of what was discussed at the 5 meetings of the Department for Science, Innovation and Technology (DSIT) Expert Panel, held throughout May and June 2026. Each summary provides an overview of the meeting, key points discussed and areas of consensus reached by the members of the panel.

Meeting 1: 5 May 2026

Attendees

Professor Russell Viner (Chair), Professor Bernadka Dubicka, Professor Deborah Fry, Professor Pete Etchells, Dr. Akshi Kumar, Professor Amy Orben, Professor Netta Weinstein, Professor Sonia Livingstone, Professor Rosie Flewitt, supported by DSIT officials.

Summary of meeting

At its first meeting, the panel agreed its role, scope and priorities. Discussion focused on building on the existing consultation and strengthening the evidence base on children’s interactions with technology. Members were invited to identify any gaps in expertise to strengthen the panel’s composition. The panel also discussed international approaches, the importance of including children and young people’s perspectives, and the need to align with wider cross government activity.

Key points

The panel discussed the social media delay in Australia and the available evidence on compliance. It agreed not to draw premature conclusions about impact and noted the difficulty of attributing change directly to the ban.

The panel highlighted the importance of consulting children and young people directly to ensure their voices are central to policy decisions. Members also noted that responsibility for demonstrating product safety could sit more firmly with technology companies, drawing on approaches in other sectors, and emphasised the importance of considering related activity across government alongside DSIT policy.

Members also raised concerns about the use of the term ‘addictiveness’ and agreed that this concept requires further exploration through a dedicated subgroup.

Areas of consensus

There was clear consensus on the need for more robust analytical foundations to support policy development. This included developing clear taxonomies of online harms and using structured frameworks, such as logic models or driver diagrams, to better link technology use to outcomes. The panel also emphasised the importance of a clear theory of change to underpin decision-making, alongside recognition that children and families have diverse needs, requiring flexible and proportionate approaches.

The panel identified key gaps in the evidence base, particularly in relation to unintended consequences of restrictions and the need to consider both harms and benefits across different age groups and for vulnerable children.

Meeting 2: 27 May 2026

Attendees

Professor Russell Viner (Chair), Professor Bernadka Dubicka, Professor Deborah Fry, Professor Pete Etchells, Dr. Akshi Kumar, Professor Amy Orben, Professor Sonia Livingstone, Professor Rosie Flewitt, supported by DSIT officials.

Summary of meeting

In this session, the panel considered 3 key questions:

• which functionalities influence children’s engagement;
• how use of these services relates to exposure to harm; and
• which features across social media, gaming and artificial intelligence (AI) systems are most associated with risks to children.

Discussion explored both the potential benefits and harms of platform features, the role of design and content in driving engagement, and the complex pathways linking use to outcomes. The panel also examined emerging risks, including those associated with AI, and highlighted ongoing evidence gaps, particularly in understanding causal relationships between features, behaviours and harms.

Key points

The panel discussed the potential benefits of some platform features, including the use of gamification as an educational tool, while also highlighting risks associated with their removal. Members noted that removing features may lead to substitution effects, where new features replicate similar behaviours or harms. Discussion also explored the distinction between intrinsic and extrinsic motivators, with evidence suggesting that extrinsic reward structures can negatively impact wellbeing. The panel further considered the role of commercial determinants of health, and how power imbalances in the online environment may contribute to harm.

The panel examined emerging risks and opportunities linked to AI. Potential harms included a preference for engaging with AI systems that reinforce users’ views, and the risk that some children may perceive AI chatbots as authoritative. At the same time, members recognised potential benefits, particularly in relation to AI-enabled educational tools such as tutoring.

The panel also discussed the impact of specific features on social behaviours, including evidence that streaks may shift friendship maintenance towards continuous engagement and contribute to fear of missing out. Members raised questions about changing social norms, including expectations of constant availability and responsiveness, and how these may relate to sleep disruption and wider wellbeing. It was noted that while higher levels of use are associated with negative outcomes, there is limited evidence of a clear dose-response threshold.

The discussion highlighted the need to move beyond simple distinctions between causation and correlation, with a focus on understanding pathways to harm. Members noted evidence linking technology use to mental health outcomes, including self-harm, and emphasised the role of online environments in increasing the scale and spread of harmful content compared with offline contexts.

The panel considered the drivers of engagement, questioning whether features are inherently addictive or whether engagement is better explained by intrinsic motivation and user enjoyment, particularly through personalised content. Members noted that policy debates have often avoided content-related issues, despite qualitative evidence suggesting that increasingly extreme or socially reinforced content may drive engagement within peer groups.

Finally, the panel discussed risks associated with specific features such as location sharing, which may expose children to harm or normalise monitoring behaviours in ways that could affect future relationships. The importance of considering how perpetrators use technologies was emphasised, alongside the need for interventions that address both victim protection and the prevention of harm. Members noted that law enforcement insights can inform understanding of how features, such as disappearing messages, may be exploited for serious harm and crime.

Areas of consensus

There was consensus that the evidence linking platform features to engagement behaviour and wellbeing outcomes remains limited, and that a lack of transparency contributes to this gap. The panel also noted that while higher levels of use are associated with more negative outcomes, particularly where use displaces activities such as sleep, the evidence for impacts at more moderate levels of use is less clear.

Subgroup 1: Addiction - 3 June 2026

Attendees

Professor Russell Viner (Chair), Professor Pete Etchells, Dr. Akshi Kumar, Professor Amy Orben, Professor Netta Weinstein, Professor Sonia Livingstone, Professor David Ellis, Dr. Suzanne Gage, supported by DSIT officials.

Summary of meeting

In this session, the group considered 2 key questions:

• how ‘addictiveness’ should be conceptualised in the context of online engagement, and
• what factors drive it, including when engagement becomes harmful or problematic.

Discussion explored the limitations of the term, including its clinical meaning compared to everyday usage, and emphasised the need for more precise, evidence-based language to describe different levels of engagement and harm. The group also examined the drivers of sustained engagement, including platform design, content personalisation and social influences, and considered how these interact with user vulnerability.

Key points

The group emphasised that ‘addictiveness’ is a contested and often overused concept, with limited evidence that current patterns of digital engagement meet clinical definitions of addiction. Clinically, addiction requires significant impairment to daily functioning, whereas common usage often reflects habits, preferences or concerns about overuse. Members noted that young people frequently describe themselves as ‘addicted’, but this typically reflects a perceived loss of control or reduced agency rather than a clinical condition and may be shaped by wider societal narratives. Members addressed the popular narrative around “dopamine hits”, noting that while dopamine is involved in many rewarding behaviours, there is limited evidence to support its use as a clinically meaningful explanation for social media use.

Discussion highlighted that policy concerns may be better framed around problematic use and reduced autonomy, rather than addiction itself. Features such as infinite scroll, autoplay and personalised recommendations were identified as reducing friction and encouraging continued engagement, particularly for more vulnerable users. While these ‘persuasive design’ features can contribute to sustained use, there is limited evidence that they are intentionally designed to create clinical addiction. However, personalised recommendations (only) may also have positive applications in areas such as education or health.

Recognising cases where social media usage is associated with young people experiencing serious life problems, the group also considered how problematic use should be defined and measured, noting the absence of a single agreed framework and the challenges of distinguishing harmful behaviour from typical adolescent use. Greater engagement was linked to increased exposure to harmful content, suggesting that pathways to harm may be a more useful focus than attempting to establish addiction. Members discussed overlaps with existing research on gaming disorder but highlighted a lack of consensus on diagnostic thresholds and prevalence.

Finally, members stressed the importance of language in shaping public understanding and policy responses. There was concern that overuse of the term ‘addiction’ risks pathologising normal behaviour and obscuring more nuanced patterns of risk. Alternative terms such as ‘habit-forming’ or ‘persuasive design’ were seen as more accurate, as they capture both potential harms and benefits while avoiding overstating the current evidence base.

Areas of consensus

There was broad agreement that the term ‘addiction’ is often not used appropriately in relation to social media and gaming, and that its scientific and clinical validity in this context is limited. While members noted that addiction may remain a meaningful concept in a narrow clinical sense, they questioned its usefulness for describing most online engagement and emphasised the need for more precise and nuanced language in policy discussions. The group also highlighted a significant lack of access to data from technology companies, which limits the ability to assess impacts and identify effective interventions.

Although full agreement on terminology was not reached, there was consensus on the importance of using careful, scientifically defensible language that does not overstate the evidence while still reflecting legitimate concerns about problematic engagement.

Subgroup 2: Age and development – 5 June 2026

Attendees

Professor Russell Viner (Chair), Professor Bernadka Dubicka, Professor Deborah Fry, Professor Pete Etchells, Dr. Akshi Kumar, Professor Amy Orben, Professor Sonia Livingstone, Professor Rosie Flewitt, supported by DSIT officials.

Summary of meeting

In this session, the group considered 3 key questions:

• how the nature and severity of online harms vary by children’s age, physiological characteristics and stage of development;
• how appropriate different policy interventions are across age groups; and
• what the experience might be for children gaining access to social media at age 16.

Discussion explored evidence of differential impacts across developmental stages, including periods of heightened vulnerability and opportunity, alongside the varying suitability of measures such as access restrictions, curfews, time limits and feature-level controls. The group also considered the potential implications of age-based thresholds, including the risk of a ‘cliff edge’ effect, and highlighted the importance of digital literacy, parental support and a stronger evidence base to inform policy choices.

Key points

Evidence suggests that the impacts of digital technologies on children vary by age and developmental stage. Longitudinal research indicates there may be sensitive periods, where increased social media use is associated with lower life satisfaction, although there remains limited evidence on the benefits of restricting access at specific ages. Early adolescence, characterised by heightened social sensitivity and a strong drive for peer connection, may increase vulnerability to harms while also offering important opportunities for social support, particularly for marginalised groups. Other research shows that higher levels of use are associated with increased depressive symptoms for a minority of users, particularly among younger adolescents, while older teenagers tend to experience greater exposure to online risks, alongside recognising the benefits of digital engagement. The group also noted the difficulty of separating the impacts of different platforms, as children often move between gaming and social media environments.

The group discussed how the effects of interventions may differ across age groups, with some evidence suggesting that younger children may be more susceptible to persuasive design features, while older adolescents are more likely to engage in behaviours such as late-night usage that can disrupt sleep. Sleep disruption, particularly in mid-adolescence, was identified as a key pathway through which digital use may influence longer-term mental health outcomes. Members also noted behavioural differences in response to restrictions, with younger users more likely to accept controls, while older adolescents may resist or circumvent them, reflecting developmental differences in autonomy and privacy.

Finally, the panel considered how transitions into digital access, particularly at a fixed age threshold, may shape outcomes. Concerns were raised about potential ‘cliff edge’ effects if access is granted suddenly, although evidence remains limited. Members discussed the potential value of gradual exposure, supported by digital literacy and parental involvement, to help young people build the skills needed to navigate online environments safely.

Areas of consensus

There was consensus that the evidence base in this area remains limited, particularly due to restricted access to relevant data, and that improved data access would be necessary to better understand impacts and inform effective interventions. Members noted that certain measures, such as curfews, may be more effective for mid-adolescents, reflecting developmental patterns of use and risk. The group also agreed that where restrictions are imperfectly implemented, older adolescents are likely to evade them by migrating to other platforms, which could undermine intended protective effects and create additional challenges for enforcement.

Meeting 3: 10 June 2026

Attendees

Secretary of State, Professor Russell Viner (Chair), Professor Bernadka Dubicka, Professor Deborah Fry, Professor Pete Etchells, Dr. Akshi Kumar, Professor Amy Orben, Professor Netta Weinstein, Professor Sonia Livingstone, Professor Rosie Flewitt, Professor Julia Davidson, supported by DSIT officials.

Summary of meeting

In this meeting, the Secretary of State thanked members for their continued expertise and engagement and emphasised the importance of the panel’s input in shaping policy following the consultation. Secretary of State made clear that forthcoming decisions will represent the starting point for further work to ensure that interventions are evidence-based, deliverable and effective in practice, and that she would be keen to involve the group in further work/implementation.

The panel then considered 3 key questions:

• what unintended consequences might arise from implementing restrictions on children’s online use;
• how partial compliance with interventions may affect their overall effectiveness; and
• the risks of introducing restrictions or bans without sufficiently robust age assurance.

Discussion explored the interpretation of evidence around correlation and causation, nuances in understanding of ‘excessive use’, a range of unintended consequences and issues relating to restrictions more generally, and persistent concerns with data access for researchers, and the effectiveness of age assurance technology.

Key points

The panel highlighted the inherent difficulty of distinguishing correlation from causation in relation to children’s technology use, noting that outcomes are shaped by a wide range of interacting social, developmental and familial factors. Members emphasised that most existing research identifies associations rather than causal pathways. Members also noted that while some experimental studies suggest small improvements in wellbeing following reductions in use, findings are mixed and context dependent.

Focusing on evidence of causality - while causality has not been firmly established, many complementary lines of evidence are pointing to that time spent on social media has a small negative effect on population well-being in adolescents, making it more likely that this is the case. Also, there is some causal evidence of real but very small negative impacts at a population level and this is likely an average across most young people experiencing little effect, smaller groups who experience limited benefits/harms, and small numbers with higher levels of either benefit or harm, averaging out to a small negative impact.

The group also stressed the importance of distinguishing between salient individual cases and population-level trends and cautioned that lower observed harm at moderate levels of use should not be interpreted as clear evidence of benefit. It was further noted that there is no consistent definition of what constitutes ‘excessive’ use, and that similar levels of engagement can lead to very different outcomes depending on individual circumstances.

Finally, the panel discussed challenges associated with implementing age-based interventions, including the limitations of age assurance technologies, such as accuracy and privacy concerns. These challenges were seen as complicating enforcement and potentially weakening the effectiveness of policy measures. Members emphasised the importance of designing interventions that minimise unintended consequences, are supported by clear guidance, and are subject to ongoing evaluation.

Areas of consensus

There was consensus that the impact of technology use varies significantly depending on context, including the platform, type of content, and the level of offline support available to the child. Members agreed that definitive causal evidence is unlikely to be achievable, and that policy must therefore operate under conditions of uncertainty, focusing on identifying specific high-risk and protective features rather than applying broad assumptions across all platforms. The panel also emphasised that limited access to data is a major barrier to building a robust evidence base and evaluating the effects of platform design on outcomes such as wellbeing and exposure to harm.

Members further agreed that existing requirements are not consistently enforced, and that stronger expectations on companies may be needed to improve safety. There was also support for approaches that go beyond restrictive measures, including improving digital literacy and drawing on lived experience to strengthen policy design and ensure it is grounded in the realities of children’s online use.