Scientists have concluded social media data shouldn't be used as a tool to gauge human behavior or trends since they're too biased.

Computer scientists from McGill University in Montreal and Carnegie Mellon University in Pittsburgh reveal evidence this data causes flaws in studies, especially in massive social media data sets.

Like Us on Facebook

The research details several issues in using social media data sets along with strategies to solve them. For example, different social media platforms certainly attract varying users. One such platform is Pinterest, which is dominated by female users ages 25 to 34.

Data feeds available to the public rarely present an accurate representation of the platform's overall data, which includes their filtering of data. The overall design of social media platforms can also affect how users behave.

For example, Facebook doesn't provide a "Dislike" button and this makes negative responses harder to detect.

Also, there's the fact that rising numbers of spammers and bots that disguise themselves as real users on social media are included in data that supposedly predicts human behavior.

New methods in deciphering and interpreting social media data make them seem more accurate than they actually are since data are taken from easy to classify users, topics and events.

Derek Ruths from the McGill's School of Computer Science said researchers should be more aware of what exactly they're analyzing when working with social media data. Social media sites have the ability to change their proprietary algorithms, that's why their filters yield unclear data that's practically unusable.

This study was published in the journal, Science.

TagsWhy Social Media Data Analysis Can Lead to Bad Science, pinterest, Instagram, social media bad science data human behavior, Why Social Media Data Do Not Represent Real Human Behavior