Is it possible to accurately predict how social media will react to an event?


8 Jul 2020

Dr Jisun An is a scientist at the Qatar Computing Research Institute at Hamad Bin Khalifa University. Image: HBKU

Dr Jisun An of HBKU in Qatar is trying to use the power of machine learning and analytics to better understand human behaviour on social media.

After completing her bachelor’s degree and master’s degree in computer science from the Korea Advanced Institute of Science and Technology in 2007 and 2009, respectively, Dr Jisun An went on to complete a PhD from the University of Cambridge in 2015.

Now based at the Qatar Computing Research Institute at Hamad Bin Khalifa University, An’s research is devoted to studying news sharing on social media. She is an honourable recipient of a Google European Scholarship in Social Computing.

‘My research helps us to gain knowledge of the society we live in and to create better institutions and systems that affect people’s lives every day’
– DR JISUN AN

What inspired you to become a researcher?

I may not have ‘the moment’ where I decided to become a researcher, but I recall being inspired by social media analysis. In July 2009, I created a Twitter account and connected with my colleagues and other well-known scholars.

I enjoyed the diversity that social network brought to my information diet and I was fascinated by how quickly information spreads and how an idea can be amplified by the public voice. It was then that I realised the impact of the new era of communications.

During this time, I also read Six Degrees: The Science of a Connected Age by Duncan J Watts and Linked: The New Science of Networks by Albert-László Barabási, which further sparked my interests in social media networks.

Can you tell us about the research you’re currently working on?

Our mobile devices enable us to access the internet at all times including news, videos, information, communicating with friends and posting on social networking sites.

These interactions on online platforms are recorded as digital footprints and with the appropriate consideration of privacy, these footprints enable us to understand individual or collective human behaviour; what people like or hate, how people feel about various topics and how people behave and engage.

Thus, it has become crucial to understand human behaviour on these online platforms. My fundamental research goal is to develop new computational methods and tools for understanding, predicting and changing human behaviour on online platforms.

One of the challenges posed by online data is the diversity and complexity of the datasets. I research various types of large-scale data and compare existing tools to overcome their limitations and use them in the right way. The data is also used to develop new measurements, machine learning models and linguistic methods to understand human behaviour online (such as audience analysis and user engagement) and, furthermore, solve real-world problems such as echo chambers, political polarisation, media bias, and online harassment.

As textual data is a primary source of online data, I apply and develop neurolinguistic programming methods to capture the opinions of individuals on social media. In particular, I develop unsupervised learning models for comparative text analysis by extracting important topics, contextual differences (frames), opinions, values and beliefs. This research, for example, enables us to answer the following question: Do Donald Trump supporters consider guns safer than Bernie Sanders supporters?

I am also currently working on a project called Precision Public Health Campaigns to determine how we could use data analysis and machine learning to create the right message for the right person, and evaluate the impact on offline behaviour beyond simple impressions and click-based metrics using targeted social media advertising tools.

In your opinion, why is your research important?

To me, this question is equivalent to why we study social science. It is important because my research helps us to gain knowledge of the society we live in and to create better institutions and systems that affect people’s lives every day.

The findings of my studies help policymakers, organisations and businesses in their decision-making process.

What are some of the biggest challenges you face as a researcher in your field?

The area I’m working on is called social computing, which is a multi-disciplinary research area combining computer science and social sciences to address problems of societal relevance. Thus, I have been collaborating with various domain experts such as political scientists and health communication scholars.

Since everyone has been trained to approach a problem from different directions and we are lacking a ‘common language’, it takes time to work on the problem that both parties value. I feel finding good collaborators is one of the hardest parts of my research.

Are there any common misconceptions about this area of research?

Fundamental questions about online human behaviours are often challenged by the difficult measurement tasks of aggregating large, noisy, non-representative data and inferring the demographic attributes of online users.

Since online users do not tend to be a representative sample of the population, using raw online data may incur sampling bias in the analyses. To avoid such bias, one should accurately know who those online users are. This is challenging as online user attributes, including demographics, are not readily available.

In my research, I have been tackling challenges in measurement problems by examining the feasibility of various technologies for demographic inference to avoid sampling bias in online and social media data. I’m also investigating how different sampling strategies (gender, age, geography, number of followers etc) affect the performance of the now-casting of two common offline indices: flu activity and unemployment.

What are some of the areas of research you’d like to see tackled in the years ahead?

While there are extensive studies on measuring public opinion, we are still unable to predict the opinions of online users on a future event. I wonder if, based on events that have occurred, we would be able to predict how the public will possibly react to similar events in the future and why they would react in that way?

This research aims to assist those who need to make a decision based on social media reactions, such as businesses, politicians, celebrities and policymakers.

Are you a researcher with an interesting project to share? Let us know by emailing editorial@siliconrepublic.com with the subject line ‘Science Uncovered’.