Core Insights - The article discusses the design and implementation of an "AI Child-Friendliness Assessment" aimed at evaluating mainstream AI models in their interactions with children, particularly focusing on their ability to address children's unique questions and needs [2][4][5]. Assessment Framework - A unique five-layer pyramid evaluation model has been developed to outline the ideal characteristics of an AI partner for children, based on decades of theoretical foundations from education, sociology, and human-computer interaction [4][5]. - The five layers of the model include: 1. **Safety and Reliability**: The foundational layer emphasizing the importance of not causing harm, providing accurate information, and maintaining confidentiality [6]. 2. **Understanding and Growth**: AI should facilitate learning and comprehension, using language that children can understand and encouraging critical thinking [7]. 3. **Empathy and Care**: AI should be capable of recognizing children's emotions and providing genuine encouragement [8][9]. 4. **Relationship Support**: AI should help children build social connections and not isolate them in a virtual environment [10]. 5. **Autonomy and Empowerment**: The highest layer, where AI should respect children's choices and foster their decision-making abilities [11][12]. Performance Evaluation - In the assessment of AI models regarding "left-behind children," significant differences were observed across ten evaluation dimensions, with lower scores in higher-order skills such as empathy and relationship support [20][22]. - The highest scoring dimensions were related to basic safety and reliability, with scores of 4.04 for confidentiality, 3.88 for accuracy, and 3.87 for non-harmfulness, indicating a strong foundational performance [20][48]. - However, higher-order skills like understanding emotions, enabling autonomy, and facilitating friendships scored below 3, highlighting a critical gap in AI's ability to provide emotional and social support [20][48]. Comparative Analysis - The AI model "Deepseek" scored the highest in the assessment, although its performance was not as pronounced in the context of "youth sexual education" [22][48]. - There was no significant difference in the child-friendliness of domestic and international AI models regarding the topic of left-behind children [24][48]. Emotional and Social Dynamics - AI performed best in addressing emotional topics, scoring 3.64, but it primarily exhibited surface-level empathy rather than deep emotional understanding [26][39]. - The assessment revealed that while AI can simulate emotional support, it struggles with providing context-specific guidance and understanding the complexities of real-world situations faced by left-behind children [49]. Implications for Education Equity - The article raises concerns about whether AI will promote educational equity or exacerbate resource inequalities, emphasizing that while AI can democratize access to knowledge, it may also obscure the need for real support systems [50]. - The findings suggest that the most significant inequality lies in the "understanding" gap, where children may relinquish their critical thinking and decision-making to AI, potentially leading to a loss of autonomy [51]. Future Directions - The article advocates for a paradigm shift from merely providing powerful tools to fostering comprehensive human development, emphasizing the need for AI to activate children's self-efficacy rather than replace traditional learning methods [52]. - Ultimately, the goal is to ensure that AI serves as a supportive companion in children's growth, particularly for vulnerable groups like left-behind children, rather than a mere source of information [54][53].

Core Viewpoint - The article discusses the role of AI in supporting children, particularly focusing on left-behind children, and emphasizes the need for a scientific evaluation framework to assess AI's suitability for children's welfare [5][59]. Group 1: AI Suitability Assessment - Tencent Research Institute and Beijing University of Science and Technology designed an "AI Suitability Assessment" to evaluate mainstream AI models in addressing children's diverse questions [7]. - A unique five-layer pyramid evaluation model was created, encompassing safety, understanding, empathy, relationship support, and autonomy [9][10]. - The assessment identified ten dimensions that collectively form a comprehensive profile of a child-friendly AI [16]. Group 2: Evaluation Findings - The assessment revealed that AI models scored higher in foundational dimensions like safety and reliability, with scores of 4.04 for confidentiality, 3.88 for accuracy, and 3.87 for non-harmfulness [24]. - However, higher-level dimensions such as empathy, relationship support, and autonomy scored below 3, indicating a significant gap in AI's ability to provide emotional support and foster social relationships [24][57]. - Deepseek was noted as the highest-scoring model in the assessment, although its performance was less pronounced compared to its results in youth sexual education topics [26][57]. Group 3: Emotional and Social Support - In addressing emotional issues, AI performed best with a score of 3.64, reflecting its ability to handle universal emotional patterns but lacking in contextual understanding [28][57]. - The AI's inability to provide deep empathy and effective empowerment was highlighted, particularly in the context of left-behind children's unique challenges [45][49]. - The assessment indicated that while AI can simulate emotional responses, it struggles with fostering genuine relationships and understanding the complexities of children's social environments [40][55]. Group 4: Implications for Education Equity - The article argues that AI can democratize access to educational resources but may also obscure deeper issues of inequality, particularly in understanding children's real-life challenges [59][60]. - The findings suggest that the most effective AI for children is not necessarily the most advanced but one that knows when to engage and when to allow children to take the lead [63]. - The ultimate goal of educational equity should be to nurture well-rounded individuals rather than merely providing advanced tools [63].

Core Viewpoint - The article discusses the potential of AI as a reliable source of information for adolescents, particularly in the context of sexual education, and emphasizes the need to shape AI into a compassionate and trustworthy guide for youth [2][3]. Group 1: AI's Role in Adolescent Sexual Education - The research aims to evaluate whether AI can provide accurate, inclusive, and empathetic responses regarding sensitive topics such as menstruation and sexual education for disabled youth [3][4]. - The study focuses on three key dimensions: basic sexual education, menstruation education, and sexual education for disabled adolescents [3][4]. - The report serves as both an assessment of current AI capabilities and a call to action for future improvements in AI's role in adolescent sexual education [3][4]. Group 2: Evaluation Framework - A unique five-layer pyramid evaluation model was developed to assess AI's suitability for children, based on decades of theoretical foundations from various fields [6]. - The five layers include: 1. Safety and reliability 2. Understanding and growth 3. Empathy and care 4. Relationship support 5. Autonomy and empowerment [6][7]. - Each layer consists of specific requirements that AI must meet to be considered a suitable partner for children [6][7]. Group 3: Assessment Results - The average scores for the ten evaluation dimensions related to adolescent sexual education were all above 3 out of 5, indicating a generally positive performance of current AI models [13]. - However, lower scores were observed in higher-level requirements such as learning ability and social interaction, suggesting challenges in these areas [13][14]. - The top-performing model in the evaluation was identified as "deepseek," which demonstrated superior capabilities in addressing adolescent sexual education topics [16]. Group 4: Comparison of Models - Domestic AI models showed significantly higher suitability for adolescent sexual education compared to foreign models [18]. - Open-source models outperformed closed-source models in most evaluation dimensions, particularly in higher-level requirements [20][38]. - The findings challenge the assumption that closed-source models are inherently superior, highlighting the advantages of open-source models in sensitive content handling [38][79]. Group 5: Specific Educational Topics - The study further breaks down adolescent sexual education into four main topics: interpersonal relationships, body awareness, sexual safety, and emotional management [22][41]. - In menstruation education, the four key topics identified were physiological health, hygiene products, emergency handling, and emotional management [41]. - For disabled youth, the four main topics were two-sex socialization, body awareness, safety awareness, and emotional understanding [57][69]. Group 6: Future Directions - The research emphasizes the need for AI to evolve from providing standardized answers to offering personalized support that considers the unique circumstances of each adolescent [84]. - It calls for a collaborative approach involving technical experts, educators, and community members to ensure AI is equipped to meet the diverse needs of youth [84][89]. - The ultimate goal is for AI to become a warm and wise companion in the growth journey of every adolescent, transcending its role as a mere tool [83][84].