Oraclum’s Wisdom of crowds (WoC) survey utilizes the essential logic of the BASON Survey but its application is more important in understanding how people feel and how their environments feel and think.
While the BASON survey is the best method on the market to uncover uncertain outcomes (such as “who will win an election?”, “will people buy this product?”, “what is the optimal price for a product?”), the Oraclum WoC is better suited to uncover truthful opinions, sentiment, and even cognitive dissonance of its participants (e.g. if people are misrepresenting themselves we will notice a difference between their responses and what their social network reveals about them, thus guiding them to the “right” answer).
Just like the BASON, the WoC also represents a significant improvement over regular surveys in both accuracy and information gained.
Our value added for scientific research surveys: network analysis
- Social network graphs tell us how various socioeconomic, demographic, and religious characteristics of people correlate with political and social affiliations.
- They reveal a plethora of highly valuable information not accessible through the questionnaire alone
- For example, how people cluster into communities, which reflects their real social affiliations that often do not agree with their declared affiliations and cannot be quantified from standard surveys
- Also, how much are two communities interconnected and how distant or close individuals within and between communities are from each other
- These data are critical for understanding how informal information propagates through society and what constitutes barriers to their mutual understanding (e.g. strength of an „echo chamber”, where people are isolated from information opposing their personal worldview)
How does it work?
- For a regular WoC we also use a network graph, and draw patterns from the network => client gets direct insights from the network analysis that no other competitor can provide
- Unlike for the regular BASON we do not adjust for individual bias, but we only use the network in order to balance the sample (we use the network to uncover if the sample itself is too biased)
- Using balancing to achieve quasi-representativness of the sample (no need for detailed stratification)
- Example: Figure shows network of consumer preferences for a product. The clusters are formed only by age (which is normal and expected), meaning that there is no major bias in the given sample
What do you get from a social network analysis?
- Clustered communities based on socioeconomic, demographic, religious, ethnic, etc. characteristics
- Cross-correlations between communities and voter preferences
- Analysis of social affiliations versus political affiliations
- Centrality measures (Laplace centrality, k-cores, betweenness, community detection)
- Cross-country comparisons of network clusters (for multiple-unit analysis)
- Assortativity within networks – how do hubs interact with other hubs