Cambridge study highlights how social media turns values into negative traits

Privileging hostile or negative content has been a short-term success but will damage social media platforms in the longer term, says David Stillwell, professor of computational social science at Cambridge Judge Business School (CJBS) and the author of a new study on Twitter discourse.

The study, ‘Two is Better than One: Using a Single Emotion Lexicon Can Lead to Unreliable Conclusions’, involved 5 million tweets and was co-authored by Prof Stillwell and Gabriela Czarnek, of the Centre for Social Cognitive Studies at Jagiellonian University in Kraków, Poland.

The authors used two lexicons to measure ‘sentiment analysis’, which gauges sentiment based on the quantity and emotional context of word use. They concluded that social media can distort language to the point that apparently positive terms like ‘independence’ and ‘ethics’ can take on negative sentiments.

“The challenge we’ve encountered was that with exactly the same data, we arrive at different conclusions depending on which lexicon we use,” says Gabriela, who worked on the study while an academic visitor at the Psychometrics Centre at CJBS. “The question we had was: as people age, do they become more positive in their expression on Twitter? If we used one of the methods, the answer would be ‘yes’, but if we used the other, the response would be ‘no’.”

“It’s like if you measured someone’s height using a ruler and then a tape measure and you got different answers depending on which device you used,” adds David. “So then we went searching for the cause of the discrepancy, which seems to be down to political words, and we realised that you get the same results if you remove political words from one of the lexicons – the NRC.

The two commonly used lexicons to assess sentiment analysis are the Linguistic Inquiry and Word Count (LIWC) and NRC Word-Emotion Association Lexicon (NRC). Both methods show an increase in positive affect of tweets until age 50 – and after age 50 the positivity drops sharply, according to LIWC, but increases steadily until age 65 based on NRC. The research found that this inconsistency was “mostly due to a particular class of words: those related to politics”.

To pinpoint that political terms were causing the discrepancy based on age, the researchers identified four topics relevant to politics that correlated with age: politics in general (with top words such as war, world, and police); US politics (Trump, President, wall, and GOP); UK politics (Brexit, EU, Labour); and Indian politics (India, Modi, BJP, Congress).

“Even words connected to politics that seem to represent positive values – such as justice, democracy and equality – can actually result in negative sentiment because of the way those words are often used in tweets,” says David, who is academic director of CJBS’s Psychometrics Centre, where the study was conducted. “Rather than expressing positive values, people often use these words in tweets to criticise a political situation or express concern with values such as justice and democracy.”.

“Context might skew our results and that’s why it’s best, we think, to use more than one lexicon in such analyses,” concludes Gabriela.

There’s no suggestion the discrepancy between the two lexicons for over-50s is the source of the harm that social media causes. The negative impact on wellbeing – including paranoia, depression, confusion, excess anger, shame and self-hatred – could be being generated because an a priori different mindset is being amplified on social media.

“I can certainly imagine situations where the same word means different things to different people – the example that quickly comes to mind at the moment is ‘woke’,” notes Prof Stillwell.

And Twitter amplifies the negative effects of differences of opinion?

“You are right that others’ research shows that negativity gets more engagement on social media, not just on Twitter but all social media,” he says. “Given that social media feeds tend to show stories at the top that have had the most engagement, this would mean that both negative stories get more engagement and also that people are encouraged by the algorithm to post negative views if they want more likes and retweets.

“Personally I think this is a mistake with social feeds, because although controversy is engaging in the short term, I think that in the long term people get turned off of using the platform. Lots of people I know have stopped using Facebook after getting into arguments with family or friends.

“Arguments might be engaging in the short term, but who wants to regularly get into online spats with family members that you then have to meet later at Christmas dinner?”

- ‘Two is Better than One: Using a Single Emotion Lexicon Can Lead to Unreliable Conclusions’ is published in the journal PLOS ONE (Public Library of Science One).