TL;DR
Turn quiet afternoons into listening time
- Thousands of audiobooks, podcasts and originals
- Listen on your phone, tablet or Echo — also offline
- Cancel anytime
Two studies to be presented at CSCW 2026 found gaps between the support social media platforms and AI chatbots appear to offer and what users can verify or experience. Researchers identified vague platform disclosures about youth safety features and found chatbot replies to Alzheimer’s caregivers were more advice-focused and less grounded in personal context than peer caregivers’ responses.
Two studies being presented at the ACM Conference on Computer-Supported Cooperative Work and Social Computing examine gaps in how social media platforms explain youth protections and how AI chatbots respond to people caring for someone with Alzheimer’s disease. Researchers found that platforms’ public descriptions of safety features could be difficult to verify, while chatbot answers to caregiver questions tended to offer structured advice with less personal context than peer caregivers’ replies.
In the youth-safety study, researchers analyzed more than 350 press releases and safety-related blog posts published over five years by YouTube, TikTok, Meta and Snapchat. They found that platforms sometimes described features as rolling out, under testing or available in only certain regions, leaving readers unable to tell which users could access them. Claims that features had been improved or made stronger were not consistently accompanied by baselines, evaluation methods or measurable outcomes.
The analysis also found that platform communications gave more attention to what young people see and whom they interact with than to areas such as data collection and sharing, access to platforms, circulation of youth-created content and how companies may profit from or amplify that content. The researchers said framing safety around individual behavior and parental oversight can place responsibility on families, even when risks are shaped by platform decisions that users cannot see or control.
A separate study compared responses to 85 real questions from Alzheimer’s caregivers posted in the ALZConnected online community. Researchers assessed answers from peer caregivers, physicians, ChatGPT and CareGPT, a chatbot trained on peer-support discussions, using 28 measures of language, emotional content and clinical content. Chatbots scored higher than peer caregivers on politeness and emotional-support markers, and were comparable to clinicians on informativeness and physician-alignment measures. Their replies were more formal and advice-focused, while peer caregivers offered more narrative and community-grounded responses. The researchers found little difference between the two chatbot systems on the measures used.
How Safety Information Shapes Accountability
Families and policymakers need to know not only that a platform says it has a safety feature, but who can use it, how it works and whether there is evidence it helps. When those details are vague or hard to check, public statements give families limited basis for judging protections, and make it harder to assess company claims amid scrutiny over social media’s effects on young people.
The chatbot findings matter because caregivers may turn to AI while dealing with complex, emotionally difficult situations. A reply can be polite, informative and organized yet still miss concerns that emerge from a person’s circumstances. The study suggests that measures of tone and clinical alignment alone may not capture whether a response meets a caregiver’s broader needs. It does not establish that chatbots cause harm or measure health outcomes.
As an affiliate, we earn on qualifying purchases.
Two Studies at CSCW 2026
The work will be presented at the 29th ACM CSCW conference, taking place in Salt Lake City from October 1 to 14, 2026. The youth-safety paper, titled “Analyzing social media claims regarding youth online safety features to identify problem areas and communication gaps,” is scheduled for the Supporting Youth and Families session on October 14, from 9:18 to 9:36 a.m.
The caregiver-response study received a CSCW 2026 Impact Recognition award. Both studies examine the distance between promised technological support and what people can understand or experience. The first reviews companies’ public communications; the second compares written answers to the same caregiver questions across four types of responders.
““If we want meaningful accountability, platforms need to communicate much more transparently about what their safety features do, who has access to them and what evidence exists that they are effective.””
— Pamela Wisniewski
AI chatbot for Alzheimer's caregivers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Limits of the Available Evidence
The youth-safety study reviewed public communications; the source material does not report independent testing of the features or establish how much protection they provide in practice. It also does not specify how often each platform used vague descriptions or provide a platform-by-platform comparison of outcomes.
The caregiver study assessed responses to 85 questions using 28 measures, but the reported findings do not establish how caregivers acted on the replies, whether advice was accurate in individual cases or what effect it had on care. The available account also does not detail all measured factors or the study’s limitations. Its findings describe patterns in the responses analyzed, not every chatbot interaction or caregiver experience.
As an affiliate, we earn on qualifying purchases.
Conference Presentation and Guidelines
Researchers are scheduled to present the youth-safety study at CSCW on October 14. They hope the findings will inform guidelines for responsible communication about youth-safety features. The conference presentation will provide an opportunity to discuss the analysis, but the source material does not identify a planned policy change or a timeline for new guidelines.
For the caregiver study, the reported next step is its presentation as an award-recognized CSCW paper. Further detail about the researchers’ plans, or changes to chatbot systems based on the comparison, has not been announced in the available report.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did the youth-safety study examine?
Researchers reviewed more than 350 press releases and safety-related blog posts from YouTube, TikTok, Meta and Snapchat over five years, looking at how the companies described risks to children and their safety features.
What did researchers find about platform disclosures?
They found that availability and effectiveness could be hard to verify. Some features were described as rolling out, being tested or limited to certain regions, and claims of improvement were not consistently paired with baselines or measurable results.
How did chatbots’ caregiver responses compare with human replies?
Chatbot responses scored higher than peer caregivers on politeness and emotional-support markers and were comparable to clinicians on informativeness and physician-alignment measures. They were generally more formal, structured and advice-focused, while peer caregivers gave more narrative, contextually situated answers.
Does the study show that CareGPT was better than ChatGPT?
The analysis found little difference between CareGPT and ChatGPT on the measures used. The report does not establish which system gives better advice or produces better outcomes for caregivers.
When will the youth-safety study be presented?
It is scheduled for October 14, 2026, from 9:18 to 9:36 a.m. during the Supporting Youth and Families session at CSCW 2026 in Salt Lake City.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
