Twitter bot
A Twitter bot is a type of software bot that controls a Twitter account via the Twitter API.[1] The social bot software may autonomously perform actions such as tweeting, retweeting, liking, following, unfollowing, or direct messaging other accounts.[2] The automation of Twitter accounts is governed by a [3]set of automation rules that outline proper and improper uses of automation.[4] Proper usage includes broadcasting helpful information, automatically generating interesting or creative content, and automatically replying to users via direct message.[5][6][7] Improper usage includes circumventing API rate limits, violating user privacy, spamming,[8] and sockpuppeting. Twitter bots may be part of a larger botnet. They can be used to influence elections and in misinformation campaigns.
It is sometimes desirable to identify when a Twitter account is controlled by an internet bot.[33] Following a test period, Twitter rolled out labels to identify bot accounts and automated tweets in February 2022.[34][35]
Detecting non-human Twitter users has been of interest to academics.[33][36]
In a 2012 paper,[1] Chu et al. propose the following criteria that indicate that an account may be a bot (they were designing an automated system):
Emilio Ferrara at the University of Southern California used artificial intelligence to identify Twitter bots. He found that humans reply to other tweets four or five times more than bots and that bots continue to post longer tweets over time.[37] Bots also post at more regular time gaps, for example, tweeting at 30-minute or 60-minute intervals.[37]
Indiana University has developed a free service called Botometer[38] (formerly BotOrNot), which scores Twitter handles based on their likelihood of being a Twitterbot.[39][40][41]
Recent research from EPFL argued that classifying a Twitter account as bot or not may not be always possible because hackers take over human accounts and use them as bots temporarily or permanently[42] and in parallel to the owner of the account in some cases.[25]
There are many different types of Twitter bots and their purposes vary from one to another. Some examples include:
Prevalence[edit]
In 2009, based on a study by Sysomos, Twitter bots were estimated to create approximately 24% of tweets on Twitter.[62] According to the company, there were 20 million, fewer than 5%, of accounts on Twitter that were fraudulent in 2013.[63] In 2013, two Italian researchers calculated 10 percent of total accounts on Twitter were "bots" although other estimates have placed the figure even higher.[64] One significant academic study in 2017 estimated that up to 15% of Twitter users were automated bot accounts.[65][66] A 2020 estimate puts the figure at 15% of all accounts or around 48 million accounts.[67]
A 2023 MIT study found that third-party tools used to detect bots may not be as accurate as they are trained on data being collected in simplistic ways, and each tweet in these training sets then manually labeled by people as a bot or a human.[68] Already in 2019 German researchers scrutinized studies that were using Botswatch and Botometer, dismissing them as fundamentally flawed and concluded that (unlike spam accounts) there is no evidence that "social bots" even exist.[69]
Impact[edit]
The prevalence of Twitter bots coupled with the ability of some bots to give seemingly human responses has enabled these non-human accounts to garner widespread influence.[70][71][22][72] The social implications these Twitter bots potentially have on human perception are sizeable according to a study published by the ScienceDirect Journal. Looking at the Computers as Social Actors (CASA) paradigm, the journal notes, "people exhibit remarkable social reactions to computers and other media, treating them as if they were real people or real places." The study concluded that Twitter bots were viewed as credible and competent in communication and interaction making them suitable for transmitting information in the social media sphere.[73] Whether posts are perceived to be generated by humans or bots depends on partisanship, a 2023 study found.[74]