NEWS9 November 2023

Be aware of biases in social media content, hears MRS event

News UK

UK – Researchers should carefully evaluate the biases inherent in each social media platform before analysing data gathered in social listening exercises, speakers at a recent Market Research Society event have said.


Speaking at a Data Analytics Council event in conjunction with The Social Intelligence Lab on the use of social media data in insight, Patrick Collins, client director at Strive Insight, said relevance and context were vital elements of research online.

“Sometimes people are responding to specific pieces of content,” he said. “You need to think about context to make sure we are interpreting what people are saying correctly.”

He also noted that each platform had specific online cultures and biases, with different people attracted to each. For example, TikTok was more focused on humour and satire, while platforms like Quora were more about showcasing a person’s intelligence.

“People do tend to say very different things based on their platform,” Collins explained.

“For example, on Reddit, we find these users are more self-assured and expert, maybe because of the types of people who like to use these platforms, and maybe the way subreddits work, where the deeper and darker you go into Reddit, the more specialised the content becomes.

“YouTube is quite different as it is a content-first platform, so people are more inquisitive and open-minded in nature. We need to curate the data to overcome some of these challenges.”

Yukari Takehisa, senior insight analyst at Convosphere, said that researchers should accept and embrace the inherent differences between social media platforms.

She said: “We should understand the characteristics of social data. Each platform has different users, topics and behaviours. Have we ever had proper data representation?

“Understand what you can get and what you cannot get, and then use other data sources to get a holistic, diverse perspective.”

Owen Hanks, chief executive and co-founder at Measure Protocol, added that researchers should seek out alternative data to counterbalance a bias on a certain platform, saying researchers should try to “understand what the behaviours of that platform are as a whole”.