There is a key trade-off in polling between anonymity of respondents and accuracy of results. We’re going to look at a few examples and outline how Revolve Polling makes use of a novel approach to achieve the highest level of both.
When we’re talking about polling, it could reference anything from basic social media polls to more advanced research methodologies used by large organizations. Almost every approach suffers from having to compromise within the anonymity/accuracy trade-off.
Let’s think about what anonymity really looks like in a polling application. In order to provide users the highest level of anonymity possible, you must not be storing or logging any associated identifiers that link them to their provided response for a survey or poll. Right away there is a clear and obvious issue here, these poll results would be terribly unreliable so this extreme end of the trade-off is barely usable.
Now at the other end of the trade-off we require some form of ID verification, strong authentication and store associated data from poll responses to ensure the highest level of accuracy possible. This gives us the desired “one person - one vote” outcome but it entirely compromises any sense of anonymity or privacy. This also could drive a lack of desire to participate or to share openly.
Of course in reality there is a fair bit of compromise and balance within actual approaches taken. We’ll look at a few examples:
Social media polls
- Typically requires an account and the service providers store associated data. No anonymity unless accounts are semi-anonymously registered but then their system suffers from results being potentially exploited by large, multi account controlling actors. Decent for fun polls but unreliable for proper research.
StrawPolls
- Does not require an account to participate but tracks users through IP address and cookies to prevent trivial attempts to vote multiple times. Low anonymity and also unreliable for serious research as these barriers can still be exploited by committed actors.
LinkedIn polls
- While similar to other social media polls, LinkedIn polls also allow the poll creator to view how all participants voted. This may enhance accuracy but takes anonymity to the lowest level possible, even providing hesitancy to participate at all.
IVR/RDD
- Relies on contacting phone numbers that can be possibly linked to known identities. Aside from a reluctance to answer, potential participants may be unwilling to share their thoughts. Anonymity is very low in order to gather the most accurate results possible. This is commonly employed by professional research organizations who must be accountable to their results.
Online Panels
- Similar to IVR/RDD, online panels are used by research organizations and similar identification issues exist in pursuing accurate results. This makes anonymity very low and often panels might track a user's responses over time.
It’s important to note that this list is definitely not exhaustive and mostly used to outline how the anonymity/accuracy trade-off manifests through various polling methods. Many organizations are spearheading great innovations and solving these problems through interesting new approaches.
So with a gained understanding of the trade-off, it’s now possible to outline how Revolve Polling does not need to make a compromise and instead can achieve high anonymity and accurate results.
The high level technical explainer for our research technology has been outlined in multiple blog posts already so we can save a full repeat breakdown here. If you would like a deeper dive check out Understanding Revolve Polling and Zero-Knowledge Proofs for Everyone.
What is key to point out is that through a combination of ID verification during account registration and zero-knowledge cryptography, we can achieve verifiability of eligible users, end to end uniqueness of accounts + survey responses and also absolute anonymity when survey responses are received and processed on our end. We do not track any personal identifying information or other metadata when responses are received, instead relying entirely on the soundness of the zero-knowledge cryptography protocol we are using.
This allows us to confidently stand apart from the many current survey/polling approaches in saying that we simultaneously achieve high anonymity and high accuracy. And from a data storage perspective, we can say that even a potential catastrophic data breach would not compromise data about our users or their responses.
If you would like to read more, please visit our Blog and stay tuned as we continue building and look forward to launching soon.