| 摘要: |
| 情感计算是指以人类情感为机器学习对象，具有情感识别和情感分析功能的智能科技。情感计算不以信息主体的身份识别为前提，旨在通过对输入端生物反馈信息或状态的分析处理而输出情感信息。为应对情感信息不当获取与利用所带来的情感操纵风险，需以信息隐私理论为框架展开风险应对之探讨。然而，现行信息隐私规范却存在风险应对不足的问题。情感信息因其具体类型的多样性而难以划入敏感个人信息的保护范围；生物反馈信息因其情感识别而非身份识别的目的难以作为生物识别信息而被保护；作为信息处理之合法性基础的告知同意也因情感信息的特殊性存在被架空的风险。鉴于欧盟《人工智能法案（草案）》中规制情感识别的经验，我国可尝试在算法治理中构建“生物数据”的概念并以之为媒介展开技术风险的分级分领域规制，从而构建可信赖情感智能的数字环境。 |
| 关键词: 情感计算 信息隐私 情感信息 生物反馈信息 可信赖情感智能 |
| DOI： |
| 分类号: |
| 基金项目:国家留学基金委（CSC）国家留学基金（留金选〔2021〕70号） |
|
| Legal Risks of Affective Computing to Information Privacy and Its Countermeasures |
|
Chu Jingyi
|
|
|
| Abstract: |
| Affective computing is an intelligent technology that uses human emotions as the object of machine learning, embodying emotion recognition and analysis functions. Affective computing is not intended to identify the information subjects and aims to output emotional information by analyzing and processing the bio-feedback information or state of the input. In order to deal with the risk of emotional manipulation brought about by the improper acquisition and utilization of emotional information, it is necessary to discuss the risk response within the framework of information privacy theory. However, the current information privacy norms have deficiencies in the risk response. Emotional information is difficult to be regarded as sensitive information because of the diversity of specific types. Bio-feedback information is difficult to be protected as biometric information because of emotional recognition rather than identification. Informed consent as the basis for information processing is also challenged due to the particularity of emotional information. Given the EU's experience in regulating emotion recognition in AI Act Proposal, it is suggested to construct the concept of “biometric-based data” in algorithmic governance and use it as a medium to carry out hierarchical and sub-domain regulation of technical risks, so as to enhance trustworthy EAI. |
| Key words: Affective Computing, Information Privacy, Emotional Data, Bio-Feedback Data, Trustworthy EAI |
