Sundar said，When people talk about the trustworthiness of bet365 cricket odds，They are talking about the performance of AI and how it reflects reality、The ability to reveal the truth。Only developers use the correct ones、Good quality data repeatedly trains bet365 cricket odds algorithms，Only then can it gain the trust of the public。In fact，Society’s concerns about the credibility of bet365 cricket odds，More concerns about the data used to train AI algorithms。

It is not easy to clarify the quality criteria of training data to non-experts。Sundar and others believe，Improving the transparency of the data annotation process、Disclose the accuracy of data annotation，will help the public establish an objective bet365 cricket odds credibility evaluation system。

To prove the hypothesis，Researchers recruited 430 experimental participants，And let them interact with an bet365 cricket odds system online。Participants informed，This AI system can discern facial expressions in social media images。The development team collected nearly 10,000 facial images，Add joy to each picture、Sadness、Angry、Fear、Surprised、Labels for disgust or neutral emotions。Finally，The development team will use the labeled data set to train the bet365 cricket odds system。Participants were also informed，A total of more than 500 people participated in the annotation work。In addition，The researchers of this experiment deliberately modified the labels，Two different annotation situations were produced: In one case，The label accurately describes the facial emotion in the image；In another case，Half of the number of labels did not match the facial emotion shown in the image。

Next，Experimental participants were randomly assigned to three bet365 cricket odds performance testing groups: bet365 cricket odds has no performance、Prejudiced performance、Behave without bias。Participants in the latter two groups observed the entire process of bet365 cricket odds processing tasks。Among them，The second group member saw it，bet365 cricket odds system completed the emotion classification task of images of white people with 100% accuracy，The emotion recognition accuracy of black images is 0%。When they realized the AI ​​system was racially biased，The level of trust in it has also been significantly reduced。But，Sundar et al also observed，The second group members’ emotional connection to the AI ​​and willingness to continue using the system in the future has not declined。

Sundar et al. invented the term “training data credibility” to describe whether the data used to train bet365 cricket odds models is credible、Reliable。Sundar thinks，The development team has the responsibility to allow users to form their own judgments about the trustworthiness of the bet365 cricket odds system，Instead of convincing them to blindly trust the AI ​​system。The researchers also recommend that developers and designers continue to disclose to users the specific process of annotating data，Develop users’ ability to identify high-quality labels，Establish regulatory awareness of bet365 cricket odds。This requires developers to come up with more creative training data sharing models，Reduce user burden while increasing user interest in participation，And avoid misleading users。

(Liu Yuwei/Compiled)