In Defense of Internet Polls

Internet | Building/Consensus
Internet polls had a promise once. They are a cheap mechanism from collecting quantifiable opinions from people who want to give it. Typically the data from Internet polling is instantly visible. This promise has been somewhat muted, due to a slow pace of innovation on the part of the community software industry, as well as some caution offered up by the polling industry.

For example, here’s what the National Council on Public Polls has to say about Internet polls:

“While different members of NCPP have different opinions about the potential validity and value of online surveys, there is a consensus that many web-based surveys are completely unreliable. Indeed, to describe them as ‘polls’ is to misuse that term.”

“Unscientific pseudo-polls are widespread and sometimes entertaining, if always quite meaningless.”

Chris Suellentrop of Slate echoed this in a 2000 piece Why Online Polls are Bunk.

Even the term “polls” is not specific enough. There’s behavioral polling, which is highly desired in trying to figure out how people are going to vote for an election, or which traffic routes people take. Then there is opinion polling, which asks respondants what they think, or how they feel. It is opinion polling which is more debatable and of thus more interest for online polling. Below I’ve given some mild proposals to address the concerns about onlin epolls.

1. A website could make the effort to ask users to register and provide some demographic information. This is done by the websites of four of the leading national papers (WSJ, NYT, LAT, WP), but many local papers, such as Boston Globe’s Boston.com, don’t bother. Registration would be a simple way to ensure that votes are counted once, and facilitate the collection of demographic collection. Using demographic information, and applying quota sampling to normalize it to population levels, is how Harris Interactive calculates its online polls

2. The modifier “scientific” for polls means that the data can be extrapolated for a larger population. Still, that doesn’t mean that “unscientific” polls are outside the realm of scienctific observation. Consider the polls on Vote.com (one of the remnants of the 2000 Election’s purported “grassroots” polling sites), run by Dick Morris. Certainly these polls do not reflect the whole population, but it can be said that they reflect the people who visit vote.com (see my analysis).

3. Other things that aren’t “scientific” are Internet forums, letters to the editor, letters to elected officials, political rallies, public debates, the United States Senate– all necessary and familiar parts of deliberative discourse. All of these go into the decision-making process. None can truly claim to accurately represent the opinion of the larger community. The presence of a scientific poll to has the adverse effect of displacing democratic discourse. There is no reason to consider the question of whether to fire the Secretary, if it’s already been “answered” by a scientific poll.

Thus Internet polls may lead opinion instead of reflecting it. To deal with this, polls should be understood to be part of the Constructive Media process and not merely a dead-end. They should pose detailed questions which demand of their respondents to think. This is the Deliberative Polling technique developed by James Fishkin at the University of Texas (and now at Stanford). Deliberative polling demands that the survey group be familiarized with the necessary background information before making their final decision.

Deliberative polling is promising, but still requires a large amount of logistics to set up. Thus it’s been excercised a limited number of times. A more lightweight format can simply be to associate polls with particular web publications, as I do here with ViewPoints. The ViewPoints are ways to quantify readers’ opinions in reaction to a media artifact.

This may help explain while the term of “polling” has escaped its strict meaning as defined by pollsters. People see opinion polling as an instrument of democracy– a way to measure opinion. They have been given the opportunity to take polling into their own hands through the Internet. They want numbers, and they want to get others to participate. Scientific polling may find itself making some space for some good ol’ populist online polling.