Automatically generated social media content may hinder the research investigation.

Social media robots batchly spit out the contents of computer generated, referring to spreading error messages, affecting elections, and damaging public health. Now, some social scientists have put forward a new allegation: for some popular websites such as Twitter, Reddit, and Instagram to obtain research on human health and behavioral information, social media robots will confuse their audiovisual.

The data of this type of website can help scientists understand: how natural disasters affect mental health; why young people in the United States are ðÍÈô µç×Ó µç×Ó µç×Ó; how people gather through complex social networks. To carry out these studies, you have to distinguish the true sound from the “sound” of the network robot.

“The network robot is designed to go online like a live person. “

In 2013, SUNE Lehmann, computer scientist of Danish Technology, designed his first network robot. He said that the robots on Twitter were very simple, and their existence was mainly to add some of the fans of the Twitter account. Lehmann wants to let students know how such robots will manipulate social systems. To this end, they jointly designed the virtual fan robots of the singer Justin Bieber.

“Bieber Bots” is very simple and quickly attracted thousands of attention. However, social media robots are increasingly complex, and it is difficult to discover. After the US presidential election in 2016, these robots quickly became the focus of public concern; some people accused, a large number of network robots were deployed on social media and made Trump in the elections. “Suddenly, it has become something that everyone is interested.” Allem said.

After that, Allem shows evidence that network robots easier to generate electronic cigarettes to help quit smoking 1, the probability is twice the probability of real people – and the fact is that this argument is still very controversial. There is also a study that robots easier to promote marijuana unconfirmed health benefits 2. These studies are based on algorithms to estimate the possibility of a Twitter account is automatically run. But Allem said that although there are robotic testing tools such as Botometer and Botslayer, many social sciences and public health researchers still fail to take the necessary steps, which may be filtered out from their data from their data. Part of the reason here is that some researchers think that they do not have this expertise.

Amelia Jamison warned that this negligence will pollute the entire data set. Amelia Jamison has studied health differences in Maryland University. She has dug a post against vaccination on social media. She said: “You may treat them as a copy of the real discussion, listen to these sounds, but in fact they exaggerate some of the opinions of some communities themselves.” She pointed out that with her research topic as an example, if not excluding The impact of the network robot, she may think that the sound of the vaccine is increasing or in the way, and the actual situation is not the case.

Katrin Weller, a Social Science Research Institute, Germany, believes that a problem that must be resolved in this area is how to define network robots. Not all network robots are maliciously spreading error messages: Some robots will provide the latest weather conditions of the weather station, and the buoy collected sea level variation data, or general news reports. Weller pointed out that some researchers define the definition of Twitter robots, sending messages more than a certain number of accounts per day – this definition is relaxed, and a live trot account of some love screen will be missed.

Other definitions are more complicated. The robot detector and robot developers are always in a competition “you chase me.” The first generation of social media robots is still relatively simple procedures, which will only be fixedly forwarded to the posts of others. But today, with the continuous development of machine learning, complex robots that can be released from the original content have been published. Some robots will post some contents from time to time, and they will imitate human habits. For example, don’t send anything when you may fall asleep. Some robot developers will mix the content generated by the live-generation with the automatically generated content, enhance the robot’s camouflage ability.

“Once you have more understanding of the network robot and testing their way, their creators have also mastered the same information.” The quantitative health geographic researcher of Zurich, Switzerland said, “This is a very tricky The problem. “Like Lehmann, some social scientists are designing robots to carry out social experiments. Policial Kevin Munger, Pennsylvania University, has created some special robots with his colleagues that identify Twitter users using racist languages ??and accuse it. Researchers let a group of robots use white men as avatars, another group of black men as avatars. Munger found that Twitter users are more likely to dilute racist speech after being accused by the previous set of robots.

After the Bobber Robot was successful, Lehmann also designed more advanced robots to study how behavior propagate from a group to another group. However, since the network robot has already saidmodiously, the public protests that may appear let him consider giving up the method. “Everything around the network robot is passed.” He said, “I will think about: ‘I have to find a quiet corner, and I will study at the controversial.'”

Scientist Wars “Zombies”: Social Network Robots or Examination Surveys