Social media for intelligence: research, concepts, and results

Abstract

When sampling part of the enormous amounts of social media data it is important to consider whether the sample is representative. Any method of studying the sampled data is also prone to bias.Sampling and bias aside the data may be generated with malicious intent, such as deception. Deception is a complicated (broad, situational, vague) concept. It seems improbable that an automated computer system would be able to find deception as such. Instead, we argue that the role of a system would be to aid the human analyst by detecting indicators, or clues, of (potential) deception. Indicators could take many forms and are typically neither necessary nor sufficient for there to be an actual deception. However, by using one or combining several of them a human may reach conclusions. Indicators are not necessarily dependent and will be added to or removed from the analysis depending on the circumstances. This modularity can help in counteracting/alleviating attacks on the system by an adversary. If we become aware that an indicator is compromised we can remove it from the analysis and/or replace it with a more sophisticated method that give us a similar indication.