SPECIAL: Do Your Social Media Posts Feed the Trolls?

We have some all-hands-on-deck agency commitments today, so no regular issue, but in its place, some new marketing research about brands and online toxicity.

There's lots to consider when you're crafting a social media post:

1. Is this a good image?

2. Am I using the right hashtags?

3. Is the voice and tone good?

Well, add one more thing to think about: Whether it will inflame the trolls and cause increased toxicity on your social channels.

Dr. Marcelo Nepomuceno is an associate professor of marketing at HEC. That's the business school at the University of Montreal. He is co-author of a paper published a couple of weeks ago in the Journal of Interactive Marketing. It's called Should We Feed the Trolls? Using Marketer-Generated Content to Explain Average Toxicity?

NOTE: This transcript was machine-generated and may contain errors.

How did you define toxicity?

Well, the way we defined toxicity is basically any post that has a very negative content connotation. And to measure it, we use the Perspective API, which is basically an API created by Google that measures toxic comments in the comment section in a social media context, for instance.

So it's looking for keywords like crappy, mad, angry, that thing?

And all the words that you're not supposed to teach your kids, yes.

Right. What did you set out to study?

Initially our study was to, I will take a step back, going to studies that I did before this one. So basically at the beginning, real beginning, I was trying to understand what drives engagement online in a social media context. What type of content should you post to maximize your engagement? Because we know from the literature that engagement will drive sales, it will drive a lot of other important KPIs. So at that moment I was looking at three types of content that we study.

So basically we're looking at contents that we call a social dimension. So those are basically contents that are trying to create a social connection with the audience, okay? We have quality dimension, which are trying to demonstrate the quality of the product or service that you're offering to your customer. And you have the selling dimension. So you're trying to sell something, either indirectly or directly, very explicitly. And then you try to, we also try to tag what they're trying to sell, okay?

So initially I was trying to study that. That was one paper that I think I'm not going to talk about it today, but that motivated me then to study this paper, which was trying to see how can you avoid toxicity. Because the mentality that I had when I was doing this study was to how can I avoid toxicity because there is all this literature saying that negative word of mouth, toxicity, being one example of it,

will have negative impacts for your brand, for your image, et cetera. So we're trying to see what content, what type of content in the post itself may generate more or less toxicity so you could avoid it. That was what I had in mind in the beginning.

And what did you learn?

So basically, we had some very interesting insight that we had, is that we have out of those three dimensions. We found that social dimension increases toxicity, okay? So whenever the marketer is trying to create a social connection with the audience. Trying to bond, trying to create some social connection with the audience in the post itself, that generates more toxicity, okay?

Can you give me an example of a post like that? That might be something I'm guessing I don't know, what are your plans this weekend? Post it here in the comments.

Yeah, it's very much the way that the post is written, but it can have like bonding, so you try to get a social bond connection. You can have like social spotlights, so this is a study that was done in the video game community. So you have a gamer who has just got this achievement in the game, he posted something, and then you have this award that you really put in that particular player in the social spotlight for the whole community. So this is the type of things that you're trying to really, you acting as if you are your friends, as if the brand itself is the friend of the user, of the consumer. It's a lot.

Oh, I see. OK. that's the tone of the social post. And that generates more negative comments.

Yes. And I think when you think about the Internet trolls, it makes a lot of sense. What do we know about Internet trolls? Internet trolls are those people who feel pleasure when you look at studies in psychology. What motivates trolling? One of the big motivations of trolling is the pleasure, is sadism basically. It's the pleasure that they feel when they are creating havoc in a community. And the social post is when the content creator is really trying to connect with the audience.

So it's a perfect setting to create this havoc because some people are going to be very attached to that brand and they come with toxic content and that creates just fight within the group. And they get pleasure out of it.

So that covers off the social posts. There were two other types of brand posts?

Yes, so the other one is quality. So basically when you're trying to demonstrate the quality of your product, so in this case, we're talking about the game industry. So we're looking at the features of the game, descriptions of what you can do in the game, immersion of the game culture, lore, and this stuff. So everything that is related to quality also generated more toxicity, okay? So this is basically the same logic. So they are going on a...

It's a way to attack the community because it's attacking the brand, it's attacking the qualities of that particular game. And then people are going to react and there you go, you start having insight. And then the third dimension is what we call selling dimension. So this is when the company is trying to sell something. So they're trying to sell the product, they have maybe a promotion, maybe they have, you're trying to convince people to buy the game somehow.

That's when trolling does not happen, when toxicity does not happen. So that's when this selling dimension happens, toxicity is reduced. Okay. Of course, I need to just highlight one important aspect of our method is that we do not tag only one out of those contents. We may have a post that have multiple contents at the same time. Okay. And I'm simplifying the story a little bit here because out of those three dimensions, we had sub dimensions. So we had 20 different sub-dimensions that we go over in the paper. But just to simplify a little bit, we grouped them into three big dimensions to make it a little bit easier to communicate findings.

Your research, as you mentioned, focused on the video game space. Do you think you'd have had different results if you'd have studied it in a different category, like online commerce?

I haven't done online commerce. I've done it in some other industries as well, such as music industry. I've done that as well in the food industry and what's the other one? I blanked now. But anyways, in the music industry, the one that I'm studying a lot now, and we have very similar results in terms of toxicity impact on final behavior. So on this study that we're talking about today, we did find that toxicity increased product usage, meaning that when you have more toxicity, people play the game more.

But in the music industry, what we had, what we found as well, this is not published yet, so I'm just giving you a bit of insight of a larger study that I'm conducting now, is that we saw that toxicity increased sales of music artists. So there is some a, when I got data of sales data, I managed to see this positive impact of toxicity on real consumer behavior.

That surprises the hell out of me. Like everything I think that we're taught in marketing school is that, other than the all news is good news comment, which I don't think is particularly accurate, but maybe it is. What, what do you think is behind that? That, that increase, an increased level of toxicity could actually benefit the brand.

It's a very good question. I'm actually doing some follow-up studies. We're going now to the lab, trying to do experiments with real consumers to try to understand what's going on. I have a few hypotheses that I can give you. I cannot really tell you what's going on, because I'm not sure yet. But one possible hypothesis is simply the algorithm that are behind those social medias. So in this study was done with Instagram, Twitter, and Facebook. So it may be that toxicity itself generates more engagement and we have some evidence for that in our data. So the more there is toxicity, the more you get comments. The more a particular post gets comments, more people are going to see that game, are going to see that post, are going to be exposed to that game. The same thing goes to music artists. More people are going to be exposed to a particular post from a particular artist.

And that greater exposure leads to more play, playtime, or to more sales of the music artist. So it may be that, maybe something behind the algorithm, but it may be something more psychological that's going on. And if it's psychological, there are two hypotheses that I'm studying now with my graduate students. So one of them is basically a self-defense mechanism by the community, like thinking on an individual level, when you have a troll of someone in the community being very aggressive, very toxic,

It is not only that game that is being attacked. You are a gamer, you may know this. Once you're really into a particular game and you are part of that social media community, attacking that game, it's not only attacking the game itself, it's attacking you as well, because you get so invested in that game, playing for so long. Same thing for an artist. You get so invested on that artist that once that artist is attacked, it's as if your own identity is attacked.

So as a way to defend your ego, to defend your identity, you end up playing more or buying more songs. It may be that that's going on. It's simply the community trying, that infighting is driving people to defend themselves and play more, to reaffirm that identity.

And then there is then a third hypothesis here, which is basically maybe there is also some compassion going on here. So consumers may be feeling some compassion towards the brand or towards the game itself, and down there that compassion is driving them to play or to buy the music.

Your data set was from 2016 to 2018. Obviously, online communities morph over time. How do you think your findings would have differed if you'd have had access to more recent data?

I think something that I'm very curious about is to see the TikTok era type of data. And I think it would be interesting to see, there is some studies already, academic studies looking at that to try to content analyze content of videos. And I think that will be a very interesting avenue of research. But looking at how much has it changed from 18 to 23.

I think the big question mark for me would be also how the algorithm may have an impact on the findings. The algorithms are really black boxes that we do not have access to, so we don't know to what degree the findings we have are based on the algorithm or based on the content themselves. Another one that's a big factor that I do not have as a control variable is whether that particular post was paid or not paid that may increase the performance of engagement so on and so forth.

I know you didn't study those, but what does your gut tell you?

I think, of course, if the post is paid, you're going to get a better performance. So I would love to have access to data of companies that actually paid, not paid. And then you can put that in the model as a control variable and see how many clicks that you're going to get from a paid post. And then you can do interact paid post with certain contents. See how those do things together may increase or decrease sales.

What surprised you the most about your findings?

I think what surprised me the most was what surprised you as well, which was basically the impact of toxicity, the positive impact of toxicity on product usage. And the analysis that we did was basically looking at how long did toxicity have a positive or negative impact on usage. And we looked at the impact up to 10 days, okay? And we saw that the impact was positive during 10 days. That means that if you post something today, there is a spike or increase.

there is a toxic post today, sorry, by a user, increases product usage during 10 days. So you can see for how long that impact takes place and we can see that impact lasts for 10 days. It's quite impressive when you think about it. So it's quite counterintuitive as well because as we were saying before, we teach our students to avoid negative word of mouth, to avoid negative content.

And basically what we see is that the results, the gut, the first impression that we have when you see those results is that we should actually try to embrace the trolls because they are helping the brands in some ways. Having said that, that's not my main recommendation for this study. I don't think that's what we should do necessarily. I think we need a little bit more data to make sense of what's going on.

So social posts, product info posts all attract more trolls. Sales posts don't. Do we just drop the social and product posts?

I think the key message here for marketers, what they should do differently would be to reduce selling as much as possible. Because selling, it has a negative impact on engagement. Okay, that means you're going to get less likes, less comments. It gets a negative impact on trolling, and it also has a negative impact on product usage.

Wait a minute, sales posts, I thought, maybe I'm misunderstanding here. I thought sales posts reduced the trolls in toxicity.

Yes, that's it. So when you have posts about selling, that reduces toxicity, but it also reduces engagement in general, such as likes and comments and things like that. And it has a negative relationship also with product usage, a direct relationship with product usage that is negative. So really, any way you look at it, it has a negative impact through toxicity and through a straight impact on product usage.

So I think, which in a way makes a lot of sense, that people are not on social media to be sold. They are there to engage with their community, with their friends, and to engage with the brand, to know what's up, to know what's new. So selling in itself has a negative impact with only one exception. When you have posts that are about sales promotion, okay? So click here to buy 50% off thing, you know? Then you have a positive impact.

But if it's just trying to push sales and trying to just convince people to buy, that doesn't help.

Dr. Marcelo Nepomuceno is an associate professor of marketing at HEC. That's the business school at the University of Montreal. He is co-author of a paper published a couple of weeks ago in the Journal of Interactive Marketing. It's called Should We Feed the Trolls? Using Marketer-Generated Content to Explain Average Toxicity?