The first study arrives at a moment of intense public and legal pressure on the social media industry. A swirl of lawsuits over the negative impacts of social media on young people has pushed platform safety practices into courtrooms, legislative hearings and front pages around the world. Yet according to Pamela Wisniewski, principal research scientist and leader of the Socio-Technical Interaction Research Group at the International Computer Science Institute, an essential piece of that public conversation has been missing. The question, she argues, is not only whether platforms are harming children, but whether families and policymakers can actually understand what protections platforms claim to provide and whether those protections work as advertised. Her team’s findings indicate that safety communications from the largest platforms are frequently selective, vague and difficult to verify, a combination that makes meaningful accountability nearly impossible.

To reach that conclusion, the researchers conducted a systematic analysis of more than 350 press releases and safety-related blog posts published by YouTube, TikTok, Meta and Snapchat over a five-year period. These communications represent the primary channel through which platforms describe their youth safety efforts to parents, journalists, regulators and researchers. The team looked for specific information about the availability of safety features, the mechanisms by which those features operate, and the evidence supporting claims of effectiveness. What they found were notable gaps across all three dimensions. Platforms, the study showed, routinely communicated about safety in ways that left readers unable to determine what protections actually existed, who could access them, or whether they had ever been shown to reduce harm.

The details of those gaps are instructive. Platforms sometimes described safety features as rolling out, being tested, or available only in certain regions, language that makes it extraordinarily difficult to establish which users are actually protected at any given time. A parent reading that a parental control tool is gradually rolling out cannot know whether their child’s account is covered. Platforms also made claims that features had been improved, strengthened or made more effective without consistently providing baselines against which improvement could be measured, evaluation methods that could be examined, or measurable outcomes that could be independently checked. In the absence of such specifics, assertions of improvement function more as reassurance than as verifiable evidence, and the researchers argue that this pattern undermines the public’s ability to hold companies accountable for the safety environments they create.

Beyond vagueness, the study uncovered a striking asymmetry in what platforms chose to discuss. Communications emphasized what young people could see on the platforms and whom they could interact with, topics that lend themselves to narratives about individual behavior and parental oversight. Far less attention was devoted to how platforms collect or share young people’s data, how children gain access to platforms in the first place, how youth-created content is circulated, or how platforms may profit from or amplify that content. These latter topics touch directly on platform design choices and business practices, the structural forces that shape risk in the first place. By framing safety as a matter of personal responsibility, the researchers concluded, platforms tended to shift the burden of preventing harm away from the companies that engineer these environments and onto young people and their families, even though many of the most significant risks are shaped by decisions that users cannot see or control.

The researchers hope their findings will inform guidelines for responsibly communicating about youth safety features, giving regulators, advocates and journalists a clearer framework for evaluating platform claims. Wisniewski and her colleagues will present the study, titled Analyzing social media claims regarding youth online safety features to identify problem areas and communication gaps, during the CSCW 2026 session on Supporting Youth and Families on Wednesday, October 14. In an era when platform safety commitments are increasingly written into legislation and cited in litigation, the ability to parse what companies actually say about their own protections may prove as important as the protections themselves.

The second study turns from corporate communication to artificial intelligence, asking what happens when people begin relying on AI to navigate some of life’s most difficult and consequential situations. The research, which received a CSCW 2026 Impact Recognition award, compared how chatbots, clinicians and peer caregivers responded to real questions submitted by people caring for someone with Alzheimer’s disease. The source material consisted of 85 genuine caregiver questions posted on ALZConnected, an online community focused on dementia care. Rather than testing the systems with hypothetical prompts, the researchers grounded their analysis in the authentic, often desperate inquiries of family members grappling with a progressive and unforgiving illness.

The methodological approach was rigorous. Using 28 measures spanning linguistic style, emotional content and clinical content, the team assessed responses to the same questions from four distinct responder groups: peer caregivers drawing on lived experience, physicians offering professional expertise, ChatGPT as a general-purpose large language model, and CareGPT, a chatbot specially trained on peer-support discussions. The results revealed a consistent and telling pattern. Peer caregivers provided time-anchored narratives and community grounding, answering from within the shared experience of caregiving. Clinicians provided concise, advice-forward, clinically bounded guidance. Chatbots, by contrast, produced a formal, verbose, highly structured consultation voice, polished and comprehensive on the surface but distinct from both human modes of response. Notably, the analysis found little difference between CareGPT and ChatGPT responses, suggesting that the specialized training on peer-support discussions did not move the needle on making the chatbot’s answers feel more human.

The quantitative comparisons added nuance to that qualitative picture. Chatbots scored higher than peer caregivers on politeness and emotional support markers, and they were comparable to clinicians on informativeness and physician-alignment measures. Yet chatbot responses tended to be less narrative, less contextually situated and more narrowly focused on delivering advice, while human responders, particularly peer caregivers, covered a broader range of issues, including emotional processing, legal and financial concerns, and the practical realities of daily care. In other words, the machines excelled at the surface features of supportive communication while the humans surfaced the full, messy landscape of what caregivers actually needed to know. Wisniewski cautioned against equating a polished, supportive-sounding AI-generated response with meaningful support, noting that caregiving is emotional and deeply contextual, and that human caregivers brought lived experience and raised issues the AI often missed. As these systems become more embedded in health and caregiving, she warned, there is a real risk of designing for the appearance of support rather than the full complexity of what people need.

Wisniewski observed that both studies are ultimately about the same phenomenon: the gap between the promise of technological support and what people actually experience. With AI, that gap can be critical, because people increasingly turn to these systems in moments of vulnerability, and a response can sound helpful while still failing to recognize the real need behind the question. With social media, the gap takes the form of safety communications that reassure without informing. The stakes of closing these gaps are rising. Later in October, Wisniewski will participate in a convening of experts examining how digital technology use affects brain development, serving on the planning committee for the National Academies of Sciences, Engineering, and Medicine workshop on Digital Technologies and the Developing Brain: Insights from Neuroscience, a public event scheduled for Monday, October 26. Her second study, titled One question, four voices: How advice for Alzheimer’s caregiving differs between caregivers, clinicians, and large language models, will be presented during the CSCW 2026 session on AI-Assisted Healthcare on Tuesday, October 13. Together, the two projects deliver a clear message to the technology industry: transparency about limitations, and honesty about what support truly requires, are not obstacles to innovation but prerequisites for protecting the people these systems claim to serve.

Subject of Research: Communication gaps in social media youth safety claims and AI chatbot health advice for dementia caregivers

Article Title: New research reveals gaps in social media risk communication and AI health advice

Article References: New research reveals gaps in social media risk communication and AI health advice. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: social media, youth safety, online risk communication, AI chatbots, ChatGPT, dementia caregiving, Alzheimer's disease, platform accountability, CSCW 2026, human-AI interaction, parental controls, health communication