How bots manipulate social media - What in the World podcast, BBC World Service

0:00 So when we're talking about bots, how do they make money?

0:03 So there's kind of like a multi-level thing going on here.

0:06 There's the companies who run them and they

0:09 kind of sell their services to different people,

0:12 whether that be companies trying to sell their products,

0:14 whether that be to countries trying to spread propaganda.

0:18 Hey, I'm Iqra and this is What in the World from the BBC World Service.

0:22 Let's be real, the comments section can be a bit of a wild west.

0:25 You never know what you're going to get there.

0:27 Some of the most popular comments can be bots.

0:30 And let me tell you, we've definitely had a few bots in our comments too.

0:34 And maybe bots can even create a viral moment.

0:37 Recently, the singer Chappell Roan faced some backlash online after

0:40 her bodyguard apparently left an 11 year old fan in tears.

0:44 It turns out the security guard wasn't actually

0:47 working for her, and now there are reports

0:49 that the whole thing could have been driven

0:51 by bots designed to spread hate and misinformation.

0:54 So in this episode, we're going to find out how much of the content

0:57 we see online could be bots and who could be behind them too.

1:05 Okay, so let's find out a bit more.

1:06 The BBC's disinformation reporter Jacqui Wakefield is right here with me.

1:10 Hey.

1:11 Hey.

1:11 Okay, so, Jacqui, we're going to chat about Chappell Roan in a bit,

1:15 but first we've got to get into the definition of a bot and bot farms.

1:20 What are they?

1:21 So a bot is just a non-human.

1:24 So robot non-human activity.

1:27 As for bot farm, when I first learned this- kind of blew my mind.

1:32 In essence, it's just a rack of smartphones which

1:36 are programmed to post on fake social media accounts.

1:39 So what happens when someone runs a bot farm is they just have a rack of phones,

1:44 and all of those phones are posting all at once,

1:47 usually hundreds, sometimes thousands of phones.

1:50 So when you're talking about all

1:51 these different phones that are programmed together,

1:53 how is that actually looking in practice?

1:56 Is it a person that's doing that or?

1:58 Yeah.

1:58 So typically a person can create a programme that they put into all the phones,

2:02 which then allows them to post.

2:04 So this has sort of been changing a lot in recent years.

2:08 So at the start, it would have been someone, you know,

2:11 having to maybe create the content and then

2:13 send that to the phones for them to post.

2:15 Then it became more easy with AI to have

2:18 the phones come up with their own content or, you know,

2:22 the machine learning algorithm to come up with its

2:24 own content to then post on social media.

2:26 So now there's a lot less human intervention.

2:29 Yeah, you touched on it a bit there,

2:30 but can we find out who is running these bot farms and bot campaigns I guess?

2:35 So, yes, sometimes and there have been many occasions where researchers

2:40 have been able to find out who's running the bot farms.

2:44 So there's kind of like a multi-level thing going on here.

2:47 There's the companies who run them and they

2:50 kind of sell their services to different people,

2:52 whether that be companies trying to sell their products,

2:55 whether that be to countries trying to spread propaganda,

2:59 or whether that be to celebrities or PR

3:02 agencies trying to kind of change public opinion.

3:04 But that being said, it's not always that easy to detect.

3:09 Sometimes we can say that there is a bot farm,

3:11 but it's really difficult to know whether that is just a random bot,

3:15 a random group of people trying to make some money,

3:17 or if there is something a bit deeper going on.

3:20 So when we're talking about bots, how do they make money?

3:24 There's a few ways they make money.

3:26 So either you set up a bot farm and use it to just drive traffic.

3:30 So you make money off the typical social media monetisation pathways.

3:35 The second is through online scams.

3:37 So getting people to click through to products

3:40 or phishing links where you scan them.

3:43 And then the third way is people hire out their bot farms to other companies,

3:48 countries, or people or influencers to sell

3:52 their services to sway public opinion.

3:55 And getting to the situation with Chappell Roan.

3:58 What exactly happened there?

4:00 What do we know happened there?

4:02 Okay.

4:02 So there was a bit of a online scandal.

4:05 I'm not sure if you saw it.

4:06 So in essence, what happened was a Brazilian football player

4:11 posted about his stepdaughter who walked past Chappell Roan's table.

4:16 The security guard cornered her and told her not to bother her.

4:19 Chappell Roan wasn't aware.

4:21 Anyway it caused a massive firestorm over whether

4:25 or not Chappell Roan was mean to her fans,

4:28 whether she was mean to children and there have

4:31 been a few things before where Chappell Roan has,

4:33 you know, told off paparazzi, asked fans not to approach her.

4:36 So that kind of dialogue has been going on for a little while.

4:39 And Chappell Roan, in response to the incident,

4:42 said she felt sad for the child and added that she hadn't

4:45 been aware of the incident and it didn't involve her own personal security.

4:50 It later turned out that the security

4:51 guard was actually working for someone else.

4:53 What happened in this case was

4:56 a research company called Gudea-they've also previously looked

5:00 at Taylor Swift and other celebrities as well- but they looked at the posts.

5:05 About 100,000 posts generated about

5:08 this Chappell Roan football player daughter scandal.

5:12 And they did find that bots contributed significantly to the online discourse.

5:18 What they found was it was only 4% of users they thought were bots,

5:23 which were contributing to the discourse,

5:25 but they accounted for 23% of the posts.

5:29 And most of the posts that they thought were originating

5:32 from bots were the ones that were stirring up more hateful comments,

5:38 misinformation, and kind of trying to, like, spread lies and spread,

5:42 I guess, like a bit of dissatisfaction about Chappell.

5:45 But what's in it for the bots?

5:48 Like, why do we get these pile ons from bots.

5:51 It's really strange because it could be a couple reasons,

5:55 especially in the case of Chappell Roan,

5:57 who is one of maybe the more- she's out and loud about some

6:02 of her opinions and a lot of people disagree with some of those opinions.

6:05 I guess so the first one is it could be people, countries,

6:10 or organisations that disagree with some

6:12 of her more outspoken thoughts and opinions.

6:15 That's one possibility.

6:17 The other possibility is it could just be for money.

6:20 So we know that a lot of commentators, influencers talk about celebrity culture.

6:24 It's very possible that the bots are just piling

6:27 in because this is a popular topic of the moment,

6:30 and therefore they want to drive traffic

6:32 to their own profiles and make money through that way.

6:35 But also it could be people trying to sway public opinion against Chappell Roan.

6:40 It's really hard to say at this stage,

6:42 and it comes to that whole idea of attribution is really difficult.

6:46 So we've got a situation where social media is full of bots,

6:50 but people use social media to talk to actual people,

6:53 and that's not necessarily what they're getting.

6:55 So what are the social media companies doing about this?

6:58 So they are doing some things.

7:01 So everyone's probably clicked on like- 'I'm not a robot CAPTCHA box',

7:06 So that's one step that social media companies have taken.

7:09 And then there's sort of like purges

7:11 of bots that social media companies have done.

7:13 In 2014, there was like a really famous

7:16 like Instagram purge where all of a sudden

7:19 a lot of celebrities followers halved and a lot

7:23 of activity on the website went down.

7:26 Do you remember?

7:27 And it was like a very like,

7:28 gossipy moment of like, okay, who's had bots to follow them.

7:31 Who mostly had bots in their following.

7:33 A lot of influencers struggled that day.

7:36 So what happened there was just like the company came up

7:40 with a way to sort of detect like that bot activity,

7:43 like people who didn't post, people who logged on at random times.

7:47 So I've been flagged as a bot before on Depop,

7:50 which is like a vintage reselling app for clothes.

7:53 Apparently my activity was irregular and that I was using the app too much.

7:59 Sorry.

7:59 And that's kind of like a way that social media companies also monitor it.

8:03 However they have been criticised a lot that they're not doing enough.

8:08 So researchers from the University of Notre Dame analysed

8:11 the AI bot policies of the main social media platforms,

8:15 and did find that pretty much on all of them,

8:17 they could get through by programming bots

8:19 to get around these CAPTCHAs and other monitoring techniques.

8:24 Yeah, I'm just saying that I can totally relate.

8:28 I've been flagged as a bot before,

8:30 not on Depop, but when trying to get concert tickets.

8:33 And I think when it's- I've been on the website for too long,

8:37 refreshing, and it thinks I'm a bot.

8:39 It's so classic.

8:39 You keep joining the queue.

8:40 You must be a bot.

8:41 I was putting in offer after offer on this skirt,

8:44 and I think they were like, this can't be a person.

8:47 So if someone's listening and thinking, well,

8:50 how do I protect myself from these bots, how can we tell who is a bot?

8:54 What would you say to them?

8:56 AI now, the quality of these bots and what they post,

9:00 it's really good and it's really hard to distinguish it from real people.

9:04 However, there's like a few tips and tricks.

9:06 So the first is are they posting all hours of the day?

9:11 Because that probably indicates they're not sleeping,

9:13 which probably means that there's some fake activity going on there.

9:16 The other one that I like to look at is does the country

9:20 that the account originates from, does that not

9:22 quite line up with what they're posting?

9:24 So for example, if they say like they're a USA patriot,

9:27 but they're based in Sri Lanka, that's just a flag.

9:29 It's not necessarily oh, that's a bot, but it's just something to look out for.

9:34 And to be honest, just like a general rule that I use for myself,

9:38 particularly on platforms like TikTok,

9:40 where typically the video is human generated

9:44 but the comments will be flooded with bots,

9:46 often what I do is just think about what I think

9:49 about the content of the video before I open the comments,

9:53 just to make sure that you know you're not being swayed constantly

9:57 by what other people or non people are telling you to think.

10:01 And then lastly, it's just to Google it

10:03 and check if what you're seeing is misinformation or disinformation.

10:08 Jacqui, I'm going to keep those in mind next

10:09 time I'm having a look at the comments section.

10:11 Thank you so much.

10:12 Thanks so much.

10:14 Thank you for joining us today.

10:15 You know what to do for this episode.

10:17 Make sure you comment below, like and subscribe.

10:20 Speaking of your comments, I wanted to shout out Aidan,

10:22 who left a comment on our episode

10:25 about whether social media is making Christianity cool.

10:28 He said, please give us more content like this.

10:31 Aiden, I really hope you're not a bot, but for now,

10:35 I'm Iqra, this is What in the World from the BBC World Service.

10:39 I'll see you soon.