One of the least-developed sets of eDemocracy tools in the UK is websites which provide advice to people on how to vote.
These typically ask the user to answer questions on political issues before providing a personalised response on how their views compare to the policy stances taken by different parties or candidates.
However, some excellent research by Wall Matthew of Kieskompas.nl in Amsterdam, Krouwel André of Free University, Amsterdam and Vitiello Thomas of Sabanci University, Istanbul has raised some interesting questions about the interaction between online voting advice and political beliefs.
The Dutch context
Their paper looked at the effect of Kieskompas.nl on how its users voted in the 2010 Dutch legislative elections.
The Netherlands has more experience with such websites than anywhere else, given that they originated there, and also provided an interesting political backdrop to the study.
The research highlights a trend over recent decades to "more volatile patterns of partisan sentiment and of voting behaviour" with many Dutch voters seemingly deciding on which party to support during the course of the campaign. And the party system is fragmented enough to give voters a wide range of choices.
"In such a chaotic informational context, a clear, personalized vote recommendation may be more appealing to voters than it would be in a more stable, less fragmented system."
The research compared data from the site's log files on the recommendations given to users with responses to surveys of their views before they received the advice and again (on an opt-in basis so not representative) after the election.
It also sought to deal with the conundrum which arises from the fact that the site is designed to identify the party that is ideologically 'closest' to the user. This being the case, there is a need to distinguish between cases of the site exerting 'influence' over the actual vote cast by the user from instances of the site correctly identifying the user's voting intentions.
However, in a helpful piece of data collection the site asked its users to estimate the likelihood that they would ever vote for each of the main parties competing in the election. This could then be compared with the advice they were given and their actual votes in the election.
When users were asked to characterise what, if any, effect they thought the site had on their vote choice, 29 per cent stated that it had 'no effect' while 71 per cent said they consciously experienced some form of 'effect'.
Some 30.2 per cent indicated that the experience was one of 'preference confirmation' while 26.1 per cent stated that their visit presented them with previously unconsidered options.
Smaller numbers said the visit directly shaped their vote choice – either by helping them to choose among several parties that they were considering (9.2 per cent) or by leading them to vote for a previously unconsidered party (1.2 per cent).
However, amongst the survey respondents, 26.5 per cent followed the voting suggestion given to them by the website while 73.5 per cent did not.
This is not the full story, however.
"The crucial element for understanding the nature of the influence exerted by the kieskompas.nl site on its users is the interplay between users' pre-existing preferences and the advice that the site generates."
The report authors conclude that "people were only likely to 'follow' the advice issued to them when the recommended party was one that they were strongly considering voting for already".
But they add that "there is evidence for the existence of a causal [website] 'effect' on user vote choice".
"Ultimately, the findings reported in this article are relatively straightforward – it appears that our survey respondents were influenced by the automated advice that they received on kieskompas.nl, but that they did not follow the site's advice when the recommended party was not seen as a contender for their vote... before the advice was received.
"Being advised by a [website] to vote for a party which one was already seriously considering appears to crystallise that pre-existing preference, making it demonstrably more likely that the site user will go on to vote for the recommended party.
"Incongruent advice appears to be, for the most part, disregarded by site users when they vote."
So the lesson from this paper is that the site can have influence, but from the range of parties which the user is already inclined towards.
Transparency and credibility
The one piece of potentially interesting data that isn't covered in the study is how wide this range actually is in practice.
If voters are theoretically prepared to vote for one of a broad range of parties then the impact of the website could be larger than if voters are only leaning towards a couple of parties.
The authors also make one other sensible point which should be taken to heart by anyone running an eDemocracy-related site.
They state that "the political science community must police such sites, and total transparency of party coding and advice generating procedures must be the standard for a [voting advice] site to be considered credible".