Network analysis of Twitter data to predict donation activity in a charity campaign

The vast volume and breadth of content on social media provides rich data for social analysis, and potentially an opportunity to predict people's behaviour – a practice now commonly used by commercial businesses or election campaigns. This article represents a research proposal to investigate whether similar methods could be used by small charities to effectively understand if social media behaviour can predict charitable giving around a specific campaign.

Writing a research proposal requires significant effort without actually getting to conduct the research! The methods and approach must be understood and defined to a level of detail that the research can be effective and the outcomes can be predicted. Furthermore, subject matter expertise is required to understand the research subject and appropriately refine the research methods to produce effective insights and results.

This research proposal identified some challenges with using Twitter social media data to predict social behaviour:

• Fundamentally, does social media behaviour reflect real-world behaviour – is Twitter a visible representation of the world or should it be regarded as a world? When basing commercial decisions on predicted behaviour, this can have significant impacts far greater than conceptual notions of whether we present a different identity on social media to the real world.
• Doing social media research without deep subject matter expertise or access to the centre of the user network could be less effective. This is partly driven by the commercialisation of social media platforms – where analysis of a specific entity's network (such as a brand or charity) is available as a paid-for feature to the specific account owner.
• Do behavioural patterns between audiences of significantly different size remain consistent?
• When using free software to conduct social media research, broader facets of the data could be uncovered by preparing the data in advance – such as extracting reaction types, number of reactions, response times, etc. Potentially these facets are provided more readily by commercial social media analysis packages.
• Ethical methods of social media research such as anonymisation must be considered carefully to reduce limitations and complications in re-collecting data multiple times during the research period.

Upon reflection, it was very interesting to have the opportunity to create a detailed research proposal – even without undertaking the research. Had it been carried out, the effectiveness and outcomes of the research would have been far greater as a result of the detailed research proposal, and the proposal activity is a valuable exercise that can be translated to many other tech and creative contexts.