5th Column is an AI that detects bots on social media. After analyzing 1.269M accounts on X.com (formerly known as Twitter), 5th Column AI predicts that approximately 64% of the accounts currently analyzed are potentially bots as seen in Graph 1.
How does 5th Column work?
5th Column AI scrutinizes accounts and their engagers, by analyzing the attributes of an account and then makes an assessment if the account is likely human or bot. When the AI suspects that the account might not be human, it then predicts a percentage score of the probability of that account being a bot. The rationale is simple: the higher the percentage, the higher the likelihood the account is a bot.
As illustrated in Graph 2, these assessment scores are distributed across percentiles which allows us to have a higher confidence in the AI’s assessment and understand how certain in aggregate the AI is.
Bot Breakdown: Two Graphs
Graph 1A: This pie graph shows the total number of accounts 5th Column has assessed as being human or bot as a percentage.
Graph 1B: This pie chart shows only the flagged accounts. It is the percentage of flagged accounts as they were created before or after Elon Musk acquired X.com (formally known as Twitter). Noting that creation dates for twtter are hardcoded into the account and cannot be changed by humans.
Graph 2: This graph shows the percentile breakdown of all accounts 5th Column has assessed. The percentile is breakdown is divided by 10s. With this bar graph you can clearly see the distribution of accounts assessed by 5th Column and their likelihood of being bots.
Disclaimer: 5th Column is a current research project by Internet 2.0. To date (Jan 2024) we have received no funding or revenue for the production of this data which aligns with X.com policies. As it is an AI assessment on flagging accounts we cannot be 100% certain if an account is a bot. Only X.com has the capabilities and information to gain certainty about accounts as they are the only holder of the account information that would allow for a verification process with the metadata to assess with high confidence accounts operated by humans. We transparently give the assessed score based on dozens of account characteristics and behaviors the AI has made an assessment on.
Has musk now purchased the poisoned chalice?
Elon Musk opened his bid for Twitter by declaring
If our twitter bid succeeds, we will defeat the spam bots or die trying!
— Elon Musk (@elonmusk) April 21, 2022
But why inherit the bots and not remove them?
Rand Corp professor Marek Posard wrote "Here's a possible hypothesis: Twitter might not want to look too closely at this problem because then they would have to remove accounts, reducing the number of reported “active users” on the platform.
Marek P.
What is 5th Column
The 5th Column platform is an is an artificial intelligence that detects bots or inauthentic accounts on social media. It is designed to track advertising fraud where up to 32% of digital spend is lost. It was originally developed in 2021 when the Pentagon was attacked on social media by the Russian made SANA botnet during the Myanmar coup.
The 5th Column algorithm utilizes a scoring system to determine whether a user is a human, a bot, or falls into the ambiguous gray area. A score of 0 indicates a high probability of the user being a human, while a score of 1 suggests a high likelihood of the user being a bot.
If an account is regularly engaging with bots then their score may tend to be high even if they are actually a human.
