AI Chatbots and Social Media Platforms Show Critical Gaps in Supporting Vulnerable Users

Communication Gaps Threaten Support for Vulnerable Users

Two groundbreaking studies presented at the 29th ACM Conference on Computer-Supported Cooperative Work and Social Computing (ACM CSCW 2026) reveal critical failures in how AI chatbots and social media platforms serve vulnerable users1

1

2

2

. The research exposes a troubling disconnect between technological promises and actual user experience, particularly affecting children on social platforms and Alzheimer's caregivers seeking AI caregiver support. Dr. Pamela Wisniewski, principal research scientist at the International Computer Science Institute, emphasized that communication gaps become critical when people turn to these systems during moments of vulnerability1

1

.

Social Media Safety Messaging Falls Short on Transparency

Researchers analyzed over 350 press releases and safety-related blog posts from YouTube, TikTok, Meta, and Snapchat spanning five years to understand how platforms communicate about youth safety features2

2

. The findings reveal that social media platforms use selective, vague language that makes accountability nearly impossible. Platforms frequently described safety features as "rolling out" or being tested in certain regions without clarifying which protections were actually available to which users1

1

. When claiming features were "improved" or "stronger," companies rarely provided baselines, evaluation methods, or measurable outcomes to verify effectiveness.

Platform Design Shifts Responsibility Away From Companies

The research uncovered a concerning pattern in how social media companies frame safety discussions. Platforms communicated extensively about what young people see or whom they interact with, but provided minimal information about data collection practices, how children gain platform access, how youth-created content circulates, or how companies profit from amplifying that content2

2

. This approach frames safety as individual behavior and parental oversight rather than addressing platform design and business practices. Wisniewski noted that platforms shift responsibility for preventing harm onto young people and families, even though many risks stem from invisible company decisions that users cannot control1

1

. The findings arrive amid mounting lawsuits over social media's negative impacts on youth.

AI Chatbots Struggle to Match Human Support for Alzheimer's Caregivers

The second study, which received a CSCW 2026 Impact Recognition award, compared how ChatGPT, CareGPT, clinicians, and peer caregivers responded to 85 real questions from Alzheimer's caregivers on ALZConnected2

2

. Using 28 measures covering linguistic style, emotional content, and clinical content, researchers found distinct response patterns. Peer caregivers provided time-anchored narratives with community grounding, while clinicians offered concise, advice-forward guidance within clinical boundaries1

1

. AI chatbots, however, produced formal, verbose, highly structured "consultation" voices that failed to capture the nuanced support vulnerable users need.

Specialized Training Fails to Humanize AI Responses

The analysis revealed minimal differences between ChatGPT and CareGPT responses, despite CareGPT receiving special training on peer-support discussions2

2

. This suggests that current AI training methods cannot bridge the gap between technological capability and genuine human connection. While chatbots scored higher than peer caregivers on politeness and emotional tone, they consistently missed the real needs behind caregiver questions. The findings matter because people increasingly rely on AI systems to navigate difficult, consequential situations where support for vulnerable users proves essential.

What This Means for Accountability and Future Development

Both studies highlight urgent needs for transparency and accountability. For social media, platforms must clearly communicate what safety features do, who can access them, and provide evidence of effectiveness. Wisniewski stressed that meaningful accountability requires platforms to move beyond vague promises1

1

. For AI development, the research suggests current chatbot architectures cannot replicate the contextual understanding and emotional intelligence that vulnerable populations require. Researchers hope these findings will inform guidelines for responsible communication about youth safety features and shape how AI systems are designed for sensitive applications. As AI chatbots become more prevalent in healthcare and support roles, understanding their limitations becomes critical for protecting those who need help most.