A new study revealed that data used for several human behavior studies gathered from either Twitter or Facebook was not reliable. The findings imply that the social media is full of bias that can result in misleading information.
Computer scientists at Carnegie Mellon University and McGill University revealed that thousands of research projects related to human behavior relied on Facebook and Twitter per year. The authors of the mentioned studies were aware of the biases but continued using these data sources hoping that the large datasets would cancel the partialities.
"Most people doing real social science are aware of these issues," said Juergen Pfeffer of Carnegie Mellon in a press release.
Social media sites full of bias as researchers fail to consider that each site caters to different population and demographics.
Seventy-one percent of online adults use Facebook, 19 percent use Twitter, 17 percent use Instagram, 21 percent has Pinterest, and 22 percent use LinkedIn. The demographics for each site vary as well, which shows an obvious bias. For instance, 66 percent of Facebook users are men while only 17 percent has Twitter, according to the 2013 survey of Pew Research Center.
Another issue cited by the computer scientists is the fake accounts. For instance, search for Angelina Jolie's account in Facebook and Twitter and you are likely to stumble in more than five fan-made accounts. Twitter has launched a tool to delete these fake accounts but the system is far from hitting zero.
"Not everything that can be labeled as 'Big Data' is automatically great, but the old adage of behavioral research still applies: Know Your Data," said Pfeffer.
Most of the studies have tapped social media data to get information for various studies that aims to understand how people relate to e-cigarette perception, diabetes anxieties, political protests and more. The researchers called for those did these studies to correct the biases or admit the errors of the data.
Pfeffer recommended to other researchers to use real data, and analyze each data using statistics, experiments, simulations, and other traditional techniques, instead of social media for a more accurate finding.
This study was published in the Nov. 28 issue of Science.