Building A Deception Database With High Ecological Validity

Lie detection research based on cue theory has been carried out for decades,scholars have explored a variety of research examples and stimulation,and constructed a series of deception database for deception detection research.However,the existing video database is relatively small in number and sample,and the lack of yellow video.In addition,a part of the database is obtained from the experimental environment,the Cheater’s cheating is strictly controlled by the experimenter,the Cheater’s motive comes from the outside,and there is a certain difference between the cheater’s cheating phenomenon which occurs independently in reality and the Cheater’s ecological validity is low.All these problems restrict the further development of deception detection.Therefore,a deception database with large sample size and high ecological validity has important research significance and application value for deception detection and its related fields.In view of the shortcomings of the existing database,this paper establishes a deception database with large sample size and high ecological validity,and provides college student evaluator’s subjective evaluation information for each character in the video.The database has important research significance and application value for the research of deception detection and related fields.The main work and innovation of this paper include the following two aspects:First,a network video platform was used to collect 199 high-risk cheating or honest video clips,and 307 video clips were obtained by video editing software.There were 224 truth videos and 83 lies.There are 160 people in the database,covering three racial and ethnic groups,and 53 of them have both lying and honest videos.There are also videos of people telling the truth in relatively high-risk and relatively low-risk situations.The database has the characteristics of large sample size and high ecological validity,which can provide good experimental material for deception detection.Second,a total of 90 college students were recruited as assessors,divided into three groups(each group of 30 people)to conduct a subjective assessment of some of the videos.There were four assessment questions:whether the video character was lying,how sentimental the video character was,and how well she liked and trusted the character.The evaluation experiment provides a large amount of subjective evaluation data for the database,which can not only help advance the research of deception detection,but also provide some interesting information for researchers in other fields,such as criminal personality.The database and the assessment information were then analysed and the results were as follows:(1)the average length of video of female characters was consistent with the average length of video of male characters,the average length of black video and yellow video,the average length of black video and white video,as well as the average length of honest video and lying video;Yellow video average length and white video average length difference is significant.(2)the evaluation problem has good reliability.(3)the average correct rate of deception detection for 90 assessors was 64%.(4)there was a positive correlation between the rate of accuracy and the rate of honesty;there was no correlation between the rate of accuracy and the degree of sadness,the degree of liking and the degree of trust;Among the three variables,the sadness degree,the liking degree and the trust degree,the sadness degree and the liking degree had a positive correlation,and the rest had no correlation.This study constructed a deception database with large sample size,covering three skin colors and high ecological validity,which can provide rich data for deception detection.By analyzing the evaluation information,it is proved that the degree of sadness,the degree of liking and the degree of trust can not be used to predict the correct rate of deception detection.