社会基础的实例 改善与健康有关的合成对话 相关的社会需求 导航
medRxiv : the preprint server for health sciences
|February 12, 2026
概括
为对话代理生成现实的合成数据对于解决与健康相关的社会需求 (HRSNs) 至关重要. 社会依据实例 (SGE) 提高数据真实性,优于传统的人口统计方法,并增强医疗保健应用中的隐私.
科学领域:
- 人工智能的人工智能
- 自然语言处理自然语言处理.
- 医疗信息学 医疗信息学
背景情况:
- 与健康相关的社会需求 (HRSNs) 显著影响患者的健康结果.
- 会话代理对可扩展的支持有希望,但面临隐私和专业培训数据的挑战.
- 目前使用人格的合成数据生成方法通常会产生通用或刻板印象的输出,因为它们依赖于粗略的人口统计学属性.
研究的目的:
- 引入社会接地实例 (SGE) 来生成更现实的和细微的合成人物来进行对话AI.
- 开发和评估使用GPT-4o生成SGE和接地合成对话的两阶段管道.
- 与传统方法相比,评估SGE在提高合成医疗保健对话的多样性和真实性的有效性.
主要方法:
- 使用GPT-4o实现了两阶段的管道,以从抽象的人格属性中生成社会基础示例 (SGE).
- 采用SGE来建立合成对话生成,探索各种提示策略,包括隐式基于名称的提示.
- 通过自动指标 (Vendi Score) 和社区行为健康专家 (CBHS) 的盲目偏好评分来评估合成数据质量.
主要成果:
- 在产生SGE方面,GPT-4o表现出高可接受度 (85%).
- 动态SGE在生成的对话中显著增强了词汇多样性 (Vendi分数:289.41对252.36).
- 将动态SGE与隐性基于名称的提示相结合的模型从CBHS获得了最高的偏好评分 (布拉德利-特里分数:0.753),超过了仅SGE和明确的人口模型.
结论:
- 基于社会的实例 (SGE) 有效地利用LLM知识来创建多样化的合成数据,克服了严格的人口学本体学的局限性.
- 隐含的暗示,比如名字,会产生比明确的人口标签更真实,更少的刻板印象的表现.
- 该框架为敏感的医疗保健应用程序创建保护隐私的对话数据集,支持代理评估和开发.
相关概念视频
Social Proof
32.4K
Social proof is a form of persuasion based on comparison and conformity. People compare their behavior and actions to what others are doing and will change to conform to do what their peers do.
32.4K
Social Traps
27.0K
Social traps are negative situations where people get caught in a direction or relationship that later proves to be unpleasant, with no easy way to back out of or avoid. The concept was orignally introduced by John Platt who applied psychology to Garrett Hardin's "Tragedy of the Commons", where in New England herd owners could let their cattle graze in the common ground. This situation seems like a good idea, but an individual could have an advantage. If they owned...
27.0K
The Sense of Self: Reflected Self-Appraisal and Social Comparison
56.2K
According to Charles Cooley, we base our image on what we think other people see (Cooley 1902). We imagine how we must appear to others, then react to this speculation. We don certain clothes, prepare our hair in a particular manner, wear makeup, use cologne, and the like—all with the notion that our presentation of ourselves is going to affect how others perceive us. We expect a certain reaction, and, if lucky, we get the one we desire and feel good about it. But more than that, Cooley...
56.2K
Social Scripts
10.3K
People tend to know what behavior is expected of them in specific, familiar settings. A script is a person’s knowledge about the sequence of events expected in a specific setting (Schank & Abelson, 1977). Essentially, scripts are a particular kind of schema, one containing default values for the features within an event. In the restaurant example, the script's features include the props (e.g., tables, menu, food, and money), the roles to be played (e.g., customer and waiter),...
10.3K
Social Facilitation
36.6K
Not all intergroup interactions lead to negative outcomes. Sometimes, being in a group situation can improve performance. Social facilitation occurs when an individual performs better when an audience is watching than when the individual performs the behavior alone. This typically occurs when people are performing a task for which they are skilled.
36.6K
Social Loafing
39.7K
Another way in which a group presence can affect performance is social loafing—the exertion of less effort by a person working together with a group. Social loafing occurs when our individual performance cannot be evaluated separately from the group. Thus, group performance declines on easy tasks (Karau & Williams, 1993). Essentially individual group members loaf and let other group members pick up the slack. Because each individual’s efforts cannot be evaluated,...
39.7K


