Computational social science: Social data bias

Investigating social bias in CSS from both a social and computational science perspective

Project status


Research areas


Computational social science (CSS) provides a powerful approach to analyse and predict the social world, yet a widely known concern with this approach to understanding social phenomena is possible bias across all stages of the process. One underrepresented area is the application of social science theory and research to help to address social bias when building CSS tools and techniques. To provide a synthesis of the current state of knowledge across the fields of social and computational science on social bias we will conduct a scoping review and a series of workshops. We aim to publish the findings to help to raise awareness of some of the concerns and possible solutions to the issue of social bias in CSS.

Beyond the current work we hope to continue the project and to build a community of interested parties enabling further problems to be investigated and for bias to be investigated in more detail.

Explaining the science

Computational social science refers to the attempt to use computational approaches to analyse and predict social phenomena.

We consider ‘social bias’ as knowingly or unknowingly making judgments on certain groups of individuals, which typically relates to race/ethnicity, gender, or sexuality.

Social science theories refer to those developed in disciplines such as psychology, sociology, or politics, to explain actions or behaviours. 

Project aims

The work will review and summarise what is known and still unknown in the area of social bias in CSS; offering critical evaluation of the current situation, identifying ways to address the highlighted issues, and outlining priorities for future research. In addition, the workshop events will bring together individuals in the Social Sciences and CSS communities and begin to build a network for this field. Ultimately, this work will help to drive forward thinking and encourage greater input from social science theory within the development of CSS techniques and tools. In doing so, the findings will identify and outline opportunities to further draw on social science to build fair and equitable AI systems—something that is crucial for society given that this bias can have severely negative consequences for groups of people, especially minority and underrepresented groups.


The work has applications across government and industry where CSS is implemented—the findings will have implications for the design, development, and use of CSS. There are also possible benefits for academia; a comprehensive review of the current issues of social bias in CSS will indicate areas for improvement and highlight priorities for future research.