Behavioral and linguistic characteristics of microblog users with various suicide ideation level: an explanatory study

Abstract

Objective To compare the differences in linguistic and behavioral features of microblog users with various suicide ideation level and to identify the users with high suicide ideation for possible real-time assessment of suicide probability through social network platform. Methods Microblog users were recruited to participate an online survey and982 respondents completed the survey. The respondents were divided into either high or lowsuicide ideation group according to their scores measured with the Suicide Probability Scale. The differences between the two groups in 10 behavioral and 88 linguistic features in microblog activities were analyzed. Results The average Suicide Probability Scale score was 69. 7 ± 11. 9 for all respondents; compared to the male respondents,the female had a higher subscale score of hostility( 13. 2 ± 2. 6 vs. 12. 7 ± 2. 4,P < 0. 05) and scaled a lower subscale score of desperation( 24. 4 ± 4. 7 vs. 25. 0 ±4. 7,P < 0. 05). Suicide ideation level was negatively correlated with social activeness and the frequency of future word use( r =- 0. 082,- 0. 073,all P < 0. 05) and positively correlated with nocturnal activeness and frequent use of third person singular and negative word( r = 0. 081,0. 077,0. 066,all P < 0. 05). The respondents of high suicide ideation were less social interactive,expressed themselves less in first person plural,used'future words'less frequently,were more active at night,and used'death words'more frequently than those of lowsuicide ideation group( P < 0. 05). Conclusion Microblog users with higher suicide ideation demonstrate significant differences from those with lower suicide ideation in terms of linguistic and microblog behavioral features. The study supports further exploration on online identification of suicidal individuals.

目的通過比較不同程度自殺可能性網絡用戶在微博行為和語言上的差異,歸納高自殺可能用戶的可識別特征,探討通過網絡社交平臺實時評估個體自殺可能的前景。方法在線招募微博用戶參與問卷調查,根據個體自殺可能性量表中文版得分情況,將982名受訪者分為446名高自殺可能組和536名低自殺可能組,分析2組在微博提取的10類行為特征和88類語言特征上的差異。結果全部受訪者自殺可能性總均分為(69.7±11.9)分;女性'敵意'分量表得分為(13.2±2.6)分,高于男性的(12.7±2.4)分,'絕望'分量表得分為(24.4±4.7)分,低于男性的(25.0±4.7)分,差異均有統計學意義(P<0.05);二列相關分析表明,自殺可能性水平與'社交活躍度'和'未來詞'使用頻率呈負相關(r=-0.082、-0.073,P<0.05),與'夜間活躍度'和'第三人稱單數'、'否定詞'使用頻率呈正相關(r=0.081、0.077、0.066,P<0.05);非參數檢驗結果表明,高自殺可能組的'社交活躍度'、'集體關注度'和'未來詞'使用頻率低于低自殺可能組(P<0.05),高自殺可能組的'夜間活躍度'和'死亡詞'使用頻率大于低自殺可能組(P<0.05)。結論不同自殺可能性的用戶在微博行為和語言表達上存在差異,高自殺可能性用戶與其他用戶相比社交活躍度低,夜間更加活躍,關注別人更少,使用更多表達否定、死亡的詞語,使用更少指向未來的詞語。