Lots of bots or maybe nots: A process for detecting bots in social media research

The use of bot messaging, that being artificially created messages, has increased since 2010. While not all bots are bad, many have been used to share extreme and divisive views on a range of topics, from policy discussion to brand electronic word of mouth. The issue with bot messaging and its prevalence is that it can affect researchers’ understanding of a topic. For example, if 25% of a dataset is fabricated, decision-making may result in a loss of profit or poor policy formation. To counteract the use of bots, this research note offers a framework to alleviate the potentially destructive nature of bot data and ensure the cleaning of data is thorough and beneficial to decision-making based on social media commentary. The framework is a four-step process, which includes thematic, automated, and characteristic identification stages. We provide three case studies to demonstrate the approach and conclude by providing key practical implications