Don’t fall prey to bots

Not all social media accounts on platforms like Facebook and Twitter are controlled by humans.

RESEARCHERS in the United States estimate that 15% of Twitter accounts are actually controlled by programs that send automated tweets.

However, it is surprisingly difficult to distinguish these so-called bots from human users.

Unless they have been clumsily programmed, at first glance, most bot profiles look like normal users. Even detection programs can fail to sniff them out.

But a good dose of human common sense could help you blow their cover. Below are several ways to spot a bot:

Respectability: Check who follows the account – bots often tend towards other bots. Carefully examine the profile picture and description: A photo copied directly from the Net and a missing or illogical profile description can be very suspicious, according to a government media authority in Germany.

Content: Bot accounts are also recognisable by their subject matter, tenor and references.

Bots are often prolific, posting a lot but barely engaging in dialogue except to disrupt the discussion with insults and provocation.

Also watch out for unusual sentence construction or repeating grammatical errors.

Likes and followers: Another sign of bot activity is if the account likes a lot of posts, says the authority.

Conversely, bot posts often receive very few likes or comments.

Activity: Several dozen posts a day – can these really originate from just one person?

The number 50 is often seen as a benchmark in this context – more than that, and you are likely dealing with a bot.

“This is obviously an arbitrary definition. There are also people who can post a lot,” says computer scientist Christian Grimme from the University of Muenster in Germany. “You can’t base your judgement on post numbers alone.”

Grimme adds that it can also be helpful to It is surprisingly difficult to distinguish bots – programs that send automated tweets or automatically post on Facebook unless you know what to look out for – from human users. — dpa

look at whether the account follows a normal human day-night pattern of activity – “but even that is not enough.”

Reaction time: Bots can react furiously quick because they scour social networks round the clock for pre-programmed keywords or hashtags.

For example, one Twitter bot, Pfannkuchenpolizei (“pancake police” in German), responds to every mention of the word “berliner” with the fact that the word refers to a German pastry similar to a doughnut – an error famously made by John F. Kennedy.

Test sites: Grimme doesn’t hold bot checking

sites such as Botometer or Debot in high regard. Researchers on a project Grimme works on have developed “inconspicuous” bots and tested them on these sites.

“These sites failed to flag our bots,” says Grimme of the results.

Their recognition rate reached about 50%. “You can’t really do much with that kind of information,” says the computer scientist.

While the sites can catch basic bots relatively easily, most people could probably discern them pretty quickly themselves.

Bot armies: “From a technical standpoint, it is also important to note that bots are theoretically easy to scale up. If you have a

program that can control a bot, you can easily control an entire army of bots with it,” writes Simon Hegelich, a professor of political data science at the Technical University of Munich in a paper.

Such bot armies have already been discovered in Twitter.

According to Hegelich, premium software that can control a network of up to 10,000 Twitter accounts can be purchased for US$500 (RM2,100).

The accounts can also be bought, says Hegelich, at a rate of US$45 (RM190) for 1,000 simple counterfeit Twitter accounts or US$150 (RM630) for Facebook accounts. – dpa