Afsaneh Razi’s research while affiliated with Drexel University and other places

... Our results showed that teens are still being exposed to risky content (i.e., explicit and/or self-harm content) even though there are filtering and/or reporting features in place on social media. This suggests that the safety features do not reflect contextualized teens' risk experience, largely because while risk is highly subjective, safety features are designed based on the risk perceptions of the third person [99]. Hence, we need to look again at those safety features through the lens of teens' perspective. ...

Reference:

Teen Talk: The Good, the Bad, and the Neutral of Adolescent Social Media Use

... This integration of online resources into youth's everyday lives has opened new avenues for seeking support, serving as an alternative option to offline support-seeking, which is often hindered by barriers like stigma and a preference for self-reliance [106,108]. Within the SIGCHI community, researchers have consistently highlighted how youth leverage the internet to seek support in different contexts [8,59,111,147]. For instance, Pretorius et al. [106] explored young people's online help-seeking practices and found that they used online resources to independently search for credible mental health information, seek empathetic and personalized support that validated their experiences, or look for immediate help through real-time interactions such as chat features or hotlines during crises. More recently, and with the advances in technology and artificial intelligence (AI), researchers have also begun exploring how youth perceive new types of online social support, like those available from AI chatbots [25,80,91]. ...

Reference:

How the Internet Facilitates Adverse Childhood Experiences for Youth Who Self-Identify as in Need of Services

... Recent studies evaluating the efficacy of LLMs in offering mental health intervention have shown promising results. When evaluated against human responses, responses by LLMs such as GPT-4 were found to demonstrate more empathy (Luo et al., 2024), active listening, and helpfulness when responding to relationship or general health-related questions (Vowels, 2024;Young et al., 2024). Another study indicated that GPT-4, when prompted to act as a therapist, exhibits competency, empathy, and therapeutic capacity when delivering singlesession therapy to individuals seeking assistance with relationship challenges . ...

Reference:

An AI-Assisted Multi-Agent Dual Dialogue System to Support Mental Health Care Providers

... Our results indicate that algorithmic approaches to identify online risks on social media for youth could be more teen-centered and effective if they take into account the difference between the platforms and the types of risks teens discuss encountering the most on those platforms. In contrast, prior works on machine learning (ML) algorithms for detecting social media risks for youth have mainly centered on training models using available datasets [11,73,115] without considering teens' shared experiences on these platforms. Therefore, instead of focusing our efforts on collecting benchmark datasets from various social media platforms, we recommend prioritizing addressing platform-specific challenges that teens discussed when they disclosed their negative experiences on these platforms. ...

Reference:

Teen Talk: The Good, the Bad, and the Neutral of Adolescent Social Media Use

... Previous studies have examined the risk perceptions and parental controls across various technologies and applications, including Virtual Reality [3], IoT devices [4], social media platforms [5], and gaming [6]. These studies identified both common and unique risks in various contexts, including explicit content, addiction, cyberbullying, and harassment [7]- [10]. ...

Reference:

Exploring Parent-Child Perceptions on Safety in Generative AI: Concerns, Mitigation Strategies, and Design Implications

... Hence, we advocate for the development of tailored trauma-informed approaches to address their specific needs. Trauma-informed approaches have been explored in the design of various digital systems and social platforms in HCI [46,109,112,121]. This approach adheres to six core principles: 1) Safety, 2) Trustworthiness and Transparency, 3) Peer Support, 4) Collaboration and Mutuality, 5) Empowerment, Voice and Choice, 6) Cultural, Historical, and Gender Issues. ...

Reference:

How the Internet Facilitates Adverse Childhood Experiences for Youth Who Self-Identify as in Need of Services

... Wang et al. (2013) found that younger individuals are more adept at adopting new technologies due to their digital upbringing, which fosters curiosity and reliance on online platforms for various needs. Similarly, Thai et al. (2023) reported that children and youths hold positive views of AI, expressing interest in AI research and advocating for shared decision-making with AI. McDonald et al. (2023) demonstrated that Generation Z, having grown up with AI, is more likely to integrate AI tools into their studies and research. This study found that education level influences engagement with digital platforms in graduate education. ...

Reference:

Assessing the Determinants of Generative AI Integration: A Study on Google's Gemini Adoption Among Thai University Students

... Meanwhile, the internet serves as a valuable resource for CHINS and other youth, offering access to support services they may not otherwise have access to [48,63]. Peer-to-peer support platforms and forums play a crucial role in this regard, allowing youth to freely share their lived experiences anonymously if desired, connect with peers who have similar experiences, provide mutual support, and become part of a larger supportive online community, mirroring the support they might lack offline [34,56,59,66,69]. At the same time, the internet could also facilitate risks associated with at-risk youths' adverse experiences [65]. ...

Reference:

How the Internet Facilitates Adverse Childhood Experiences for Youth Who Self-Identify as in Need of Services

... Pitfalls of social media-based online peer support have also been highlighted, such as personal distress due to others' experiences [44], unhelpful interactions with others [52], social exclusion [44], and feeling of vulnerability when talking to strangers online [24]. Emerging evidence also documented other adolescent online safety issues including online sexual risks [12,58,112], exposure to explicit media content [31,100,119], and problematic media use with mental health issues [89,101,102]. ...

Reference:

Teen Talk: The Good, the Bad, and the Neutral of Adolescent Social Media Use

... Drawing from the Razi et al. case study [113], where researchers successfully gathered and analyzed similar sensitive data from social media platforms, it is evident that with appropriate methodologies and ethical considerations, collecting targeted data from Instagram to detect teens' body shaming instances is indeed feasible. By doing so, the actual risks faced by teenagers in their online lives would be acknowledged, which would result in allowing the risk detection models to accurately identify these risks on specific platforms (e.g., [10,100,112]). As such, the models would learn patterns, contextual cues, and platform-specific dynamics that would be more representative of the risks faced by them. ...

Reference:

Teen Talk: The Good, the Bad, and the Neutral of Adolescent Social Media Use