The use of social media in fighting corruption
A guest blog from a student group from the Copenhagen Business School, outlining the results of their study on social media and anti-corrutpion. The study explores how social media analytics could help us better understand what people think and do or do not do about corruption.
Since the creation of social media platforms, every day tremendous amounts of data are generated on Facebook, Twitter, Google+, LinkedIn and various blogs and forums. Continuously updated posts, statuses and tweets convey opinions on topics ranging from stock markets to politics, forming an alternative information source for the world to systematically detect, track, and fight corruption. Brazil’s Ministry of Justice is one successful example of using social media data and other forms of data to identify corruption and other organized crimes like drug trafficking and money laundering. Existing technology and tools allow the Ministry to rapidly uncover hidden connections, networks and create insights, substantially increasing efficiency and reducing the timeframe of investigations. This is the reason why Transparency International (TI) asked students of the Copenhagen Business School to identify the possibility of using social media data to further understand and potentially reduce corruption.
Our research confirmed the potential of social media analytics, which seem to give a reasonable picture of what people think about corruption in their countries and could even serve the purpose of identifying the nature and causes of corruption. With the advance of text mining technology, other possibilities lie in shedding light on different emotions: “What are the chances that people will protest? Do they dislike or hate corruption?” Other possibilities include the analysis of volume trends among different periods of time or the automatic categorization of the data into different themes.
The statistics derived from the data created results that are similar to TI’s Global Corruption Barometers , verifying our assumption that social media can serve as a good proxy for people’s perception and opinion on corruption. Sentiment analysis categorizes a large share of the posts as negative (92% for UK and 91% for US), consistent with our expectations but giving limited additional insights on how much corruption is occurring in these countries. Seasonality analysis, tracking the occurrence of related mentions over time, offers the possibility to set up a real-time monitoring mechanism that could help TI to better understand the general corruption trends and grasp the latest corruption events.
Our findings suggest that a tailor-made analytical tool could be needed for TI to analyze social media data and present the results to the public. An online web portal that captures digital content and provides a real-time account of corruption could be a promising option for this purpose. To refine such a tool, a precise list of keywords complemented with jargon and vernaculars should first be defined to retrieve the related data. Volume across different countries could then be benchmarked to depict trends over time. Topic clusters could also be analyzed. Promising options would be entities like “Political Parties in UK”, case studies like “Election in Brazil” and subjects like “white collar crime”. Finally, even though the current quality of sentiment analysis is unsatisfactory, the option to train a machine using corruption related text could potentially improve the quality of analysis.