Abstract
Opinion leaders play an important role in influencing people's beliefs, actions and behaviors. Although a number of methods have been proposed for identifying influentials using secondary sources of information, the use of primary sources, such as surveys, is still favored in many domains. In this work we present a new surveying method which combines secondary data with partial knowledge from primary sources to guide the information gathering process. We apply our proposed active surveying method to the problem of identifying key opinion leaders in the medical field, and show how we are able to accurately identify the opinion leaders while minimizing the amount of primary data required, which results in significant cost reduction in data acquisition without sacrificing its integrity.
The Challenge of Identifying Influentials
In fields ranging from marketing to medicine, identifying "Key Opinion Leaders" (KOLs)—individuals whose opinions shape the behavior of their peers—is vital. Traditionally, researchers have relied on two distinct approaches:
- Primary Methods: Direct surveys where peers nominate their leaders. This is the "gold standard" for accuracy but is prohibitively expensive and time-consuming.
- Secondary Methods: Using "proxy" data like citation counts or social media follower numbers. While cheap, these often correlate poorly with actual peer influence; a highly-cited researcher isn't necessarily the person local doctors turn to for clinical advice.
One of the most striking findings of this research is that standard network metrics like PageRank, Degree Centrality, and Betweenness Centrality performed no better than random selection for identifying KOLs in real-world medical datasets. This highlights why secondary data alone is often insufficient for high-stakes decision-making.
The Active Surveying Framework
The authors propose Active Surveying, a greedy probabilistic approach that uses active learning to intelligently pick which individuals to survey next. The framework functions as a loop that integrates primary and secondary data sources.
How it Works:
- Feature Engineering: The system augments a primary nomination network with secondary data. For pairs of individuals, it examines "edge features" (e.g., co-authorship, shared hospital affiliation) and "node features" (e.g., h-index, academic rank).
- Probabilistic Classification: A logistic regression classifier is trained on currently observed survey results to predict the likelihood of nominations between unsurveyed individuals.
- The Scoring Function: The model targets "candidates"—people who have received some nominations but haven't yet reached the threshold ($\alpha$) to be confirmed as a leader. It prioritizes surveying respondents likely to push these candidates over the threshold.
- Exploration vs. Exploitation: To prevent bias, the system uses Weighted Uncertainty Sampling (WUS). It occasionally chooses to survey individuals the model is least certain about (exploration) to improve the classifier's overall accuracy, rather than just picking the most likely yield (exploitation).
Performance and Cost Savings
The study applied this method to a healthcare dataset of 2,004 physicians. The results demonstrate that the "Active" component of the algorithm provides significant efficiency gains over passive models.
Key results include:
- 30% Cost Reduction: The Active Surveying method required 30% fewer respondents to identify the same number of opinion leaders compared to the best-performing baseline.
- Real-World Impact: In a practical scenario where a survey costs $500 per person, using this method to identify 50% of the KOLs in a network would save approximately **$52,500** compared to traditional methods.
Conclusion and Future Work
Active Surveying offers a robust way to maintain the integrity of primary data while leveraging the speed of secondary proxies. While the current model assumes nominations are independent events, future iterations could explore "full joint approaches" to account for social clustering. Additionally, incorporating varying acquisition costs—recognizing that a KOL's time is more expensive than a junior associate's—could further refine the utility function for real-world application.
Summary updated