Improving Health Risk Assessment Through the Wisdom of Crowds

Jingwen Zhang is an assistant professor of communication. Her project, which looks at collective intelligence and its impact on decision-making, was awarded a 2016-17 ISS Individual Research Grant. She provided this update in May 2018.

How did this project come about? What inspired it?

The Wisdom of Crowds argues that a group’s average judgement is more accurate than any one individual’s judgement because individual errors cancel out in aggregate. This phenomenon has been shown to offer improved accuracy in geopolitical forecasting, financial decisions, and even sports betting. Although the aggregated average opinions can be meaningful to simple prediction or estimation tasks (e.g., the average estimate of a pig's weight is closest to its true weight than any one individual's estimate), the emergence of collective intelligence and its potential influence on individual decision-making regarding health remain largely unknown. 

In 2014, my colleague and I published a review article "Collective Intelligence: The wisdom and foolishness of deliberating groups" that discusses the conditions for the emergence of collective intelligence and its impact on decision-making. We identified a few directions for applying the concept of collective intelligence in substantial domains. The idea of leveraging the Wisdom of Crowds to improve health risk assessment is one of them. More specifically, by incorporating online networks, I propose to study how changes in the structure of online information networks (e.g., decentralized versus centralized networks) can influence the overall accuracy of personal health risk assessment, including assessment of risks for cancer, sexually transmitted diseases, and chronic diseases.

How has it progressed since you received an ISS Individual Research Grant?

Over the course of the past year, my PhD students and I worked with a web developer to build an online system to experimentally test collective intelligence in various network structures. We completed the system building and experimental materials. In addition, we completed the initial testing of the system and a pilot experimental study. We are currently conducting the full experimental test involving more than 1000 participants recruited from the Internet.

What notable or surprising findings can you share at this point?

We found preliminary results indicating that decentralized networks that facilitate equal communication among all network peers was the most effective in eliciting the collective intelligence effect and improving the accuracy of individuals' risk assessment. Although centralized networks could occasionally improve the collective's judgement, their success relied on the central node's accuracy level. For instance, if a centralized network happens to have a domain expert (e.g., a public health professor) occupying the central node, everyone else in the network will greatly increase their accuracy. This finding corresponds to the idea of "opinion leader," who has the power to influence others. However, we argue that popular opinion leaders do not always hold accurate opinions (e.g., pop stars may hold misperceptions about vaccination), thus relying on centralized networks is more likely to lead the collective astray.

What is the next step?

We are currently conducting the full experimental test involving more than 1000 participants recruited from the Internet. The next step is to complete the data analyses and calibrate our preliminary findings. Importantly, we want to show that individual's improved risk assessments can also lead to changes in their attitudes regarding different preventative health behaviors, such as getting the flu vaccines and exercising regularly.

We plan to present the results at several conferences (e.g., International Communication Association, International Conference on Computational Social Science, American Public Health Association) and publish the results in a peer-reviewed academic journal.

Learn more about Jingwen Zhang.