A team from India, Netherlands, Poland, and Switzerland has looked at how to improve data analysis and to reduce the inherent bias in social network analysis. Writing in the International Journal of Applied Management Science, the researchers recognise that in quantitative surveys and social network analysis, the accuracy of data can often be skewed by biases in how respondents answer the questions. One particular form of bias, known as declarative bias, poses a significant threat to the reliability of survey results, particularly when addressing complex social issues.
Declarative bias occurs when survey participants, consciously or unconsciously, provide answers influenced by social expectations, fatigue, or external pressures rather than reflecting their true attitudes or beliefs. This type of bias is particularly problematic when the research seeks to inform public policy, as it can lead to misleading conclusions about society’s attitudes and behaviour and thus inappropriate policies.
Response time testing could offer an answer. The assumption is that a more immediate response tends to reflect a stronger, more internalized opinion, while a slower response may reflect uncertainty or a response swayed by external factors, such as social desirability or reading into the questions themselves to work out what the right answer might be. By distinguishing between these types of responses, the researchers suggest that it might be possible to segregate strong answers from the flimsy.
They tested their approach on an international survey conducted in Spain and Sweden to explore attitudes toward the COVID-19 pandemic. Their results were striking. By homing in on high-confidence, fast responses, the team could see a much greater diversity of opinion. By contrast, a conventional analysis, where declarative bias was present, showed much more homogeneous opinions.
The findings have implications for public policy and health interventions based on surveys of the public or stakeholders on a given topic. For instance, public health policies based on the assumption of uniform public opinion on issues such as the pandemic might fail to address the subtleties of diverse opinions from different groups. By reducing declarative bias in the analysis of surveys, it should be possible to form policy that takes into account diverse opinions and needs.
Fernandez, G.P., Norré, B.F., Reykowska, D., Dutta, K., Nguyen-Phuong-Mai, M., Fernandez, J. and Ohme, R. (2024) ‘Social network of confident attitudes with response time testing’, Int. J. Applied Management Science, Vol. 16, No. 5, pp.1–31.