人工智能辅助下的心理健康新型测评

近年来, 人工智能技术的飞速发展及应用催生了“智能化心理健康测评”这一领域。智能化心理健康测评能够弥补传统方法的不足, 降低漏诊率并提高诊断效率, 这对于心理健康问题的普查及预警具有重大意义。目前, 智能化心理健康测评处于初步发展阶段, 研究者基于在线行为数据、便携式设备数据等开展主要以数据驱动为导向的探索研究, 旨在实现更高的预测准确率, 但是测评结果的可解释性等指标尚不够理想。未来的智能化心理健康测评需要强调心理学领域知识和经验的深度介入, 提高测评的针对性和精细化程度, 加强信效度检验, 这对于智能化心理健康测评工具的进一步发展和应用至关重要。

A new type of mental health assessment using artificial intelligence technique

The application of artificial intelligence and big data mining technology in the field of mental health has promoted the development of intelligent mental health assessment. Intelligent mental health assessment entails the application of artificial intelligence technology in acquiring and analyzing data and modeling the relationship between behavioral features and mental health problems. Intelligent mental health assessment has broadened the forms of data and the analysis methods of traditional mental health assessment, enabling researchers to obtain multi-modal data based on more simulated situations and achieve more efficient and accurate assessments. At present, researchers mainly carry out mental health assessments based on social media data, smart device data, video game data, and wearable device data to explore various features related to mental health and build predictive models. Social media data mainly refer to the text content posted by users on social media, which is widely used in psychological assessment. Researchers have explored text features related to mental health. Foreign researchers mainly predict users’ mental health conditions based on the contents posted on platforms such as Facebook and Twitter. Domestic researchers mostly rely on Weibo and other platforms to conduct related research. Smartphones and other devices record individual daily behavioral data, including application software use, communication, location movement (based on GPS), etc. These behavioral data provided effective information for predicting the psychological characteristics of individuals. Besides, with the widespread use of smartphones and other mobile devices, collecting audio and video data has become more convenient. Researchers can extract features such as actions, voices, and expressions to achieve an immediate and automatic evaluation of participants’ mental health. Video game data refers to the log data of the player during the game. It contains a wealth of behavioral performance information of the individual in the virtual environment. Researchers can evaluate the individual’s abilities and psychological characteristics based on the data. Game-based assessment is mainly used to assess individual abilities and cognitive impairment. However, there are few studies on mental health assessment based on games, only some assessments of the positive personality. Mental health problems are often accompanied by obvious physiological reactions. Researchers use wearable devices to collect physiological indicators such as brain electricity, eye movements, heart rate, and skin temperature for mental health monitoring. Researchers use EEG data and eye movement data to identify mental health problems related to emotions and attention. Indicators of skin temperature and heart rate reflect the individual's mood and stress state and therefore have the potential to predict the level of individual mental health. The future research directions of intelligent mental health assessment mainly include five aspects. First, previous research on intelligent mental health assessment has often used data-driven methods to explore features and construct predicting models, which is hard to explain the complex relationship between behavioral indicators and latent mental health state. Therefore, further improvement of pertinence and refinement is demanded. Researchers should design tasks based on psychological theories, carry out meaningful feature extraction, and gradually refine from rough dichotomous diagnosis to continuous and typed diagnosis. Second, unsupervised data mining is difficult to ensure the validity and interpretability of assessment. To carry out effective assessment and reduce errors in the new simulated environment, the task design of intelligent mental health assessment should be designed based on the evidence center. Third, the current intelligent mental health assessment mainly uses the indicators in the computer field, and the relevant research considering the reliability and validity is very rare. Researchers should select prediction models based on specific tasks and test the generalization and stability of prediction models in different datasets and scenarios. Fourth, different data sources and features have unique advantages. Researchers could obtain multi-modal data for modeling and analysis with the application of the advanced technology of artificial intelligence. Finally, privacy protection and ethical issues are essential for intelligent mental health assessment. Subjects should be notified before data acquisition and use.