At this week’s Spring Seminar Series at the Cyber Policy Center (CPC), Professor Jeffrey Hancock, Director of the Stanford Social Media Lab and Co-Director of the CPC, presented key takeaways from the recently published 2025 Stanford Youth Safety and Digital Wellbeing Report. 

The Report is a product of a workshop, co-hosted by the Stanford Center for Digital Health and the Social Media Lab, which brought together experts from academia, public health, industry, and civil society with the goal of establishing a taxonomy of potential harms online based on existing scientific literature, determining how to best measure those harms, and providing recommendations to mitigate them. 

Core Categories of Harms:

The talk began with a discussion of the Report’s approach to categorizing potential harms, called the Integrated Harm Framework (IHF). The IHF is based on 22 discrete forms of potential harms, consistent among scientific literature. These harms can be categorized into four key conceptual categories:

1. Threats to Safety/Criminal Activity: Examples include child sexual abuse material (CSAM), extortion or sextortion, or threats violence.
2. Health and Wellbeing: These include harms that have the potential to undermine an individual’s wellness, and include content related to self-harm, eating disorders, or addiction, among others.
3. Other Content-Driven Harms: These would have immediate tangible impacts, but nevertheless still put the user’s online wellbeing at risk, such as misinformation, hate speech, or exposure to lude or violent content.
4. Other Harms: This captures harms which may not have direct or easily measurable impact, but undermine the individual’s online wellbeing, including algorithmic bias or rights infringements.

Key Themes From the Workshop:

After laying out the IHF taxonomy, Hancock presented the key themes that emerged from the workshop, as discussed in the Report. 

1. Prevalence vs. Severity: How do we balance these two considerations when assessing harm online; for example, social comparison is highly prevalent but not very severe, whereas child sexual abuse material (CSAM) is not as prevalent but much more severe.
2. Mechanisms vs. Outcomes: Some harms, like social comparison, were identified as potential explanations for the outcomes, rather than outcomes that should be measured themselves.
3. Actionability vs. Platform Alignment: If social media is one of several causes of increases in anxiety, what is the action item for a platform given their existing incentives?
4. Youth-Specific vs. General Harms: Children and adolescents have specific vulnerabilities compared to the broader public. For example, body image content is uniquely important for young people, different from general populations.
5. Societal Issues vs. Social Media-Only Issues: To what extent can we distinguish broader social problems with those that are unique to social media. For instance, social comparison is a psychological phenomena that is not specific to social media, but social media can exacerbate this problem. 

Recommendations:

Hancock concluded with recommendations for key stakeholders: platforms, policymakers, and researchers. The topline: be specific when discussing harms online. When we think about harms with specificity, we can better understand the connectedness between the wide range of potential harms, how those harms might affect a wide range of unique individuals differently, and how these take different forms across a diverse platform landscape.

On Tuesday, April 29 at 1 PM Pacific, Jonathan H. Chen MD, PhD, Assistant Professor of Medicine (Biomedical Informatics) and of Biomedical Data Science at Stanford University, gives his talk Artificial Intelligence in Medicine - Real Magic or Technological Illusions? To learn about this, and to register for future Cyber Policy Center events, visit CPC's event page.