Implementation of OKCupid for voters


In my last post I essentially proposed that somebody should build an OKCupid for voters. In this post, I’m going to address a number of questions pertaining to the construction of such a comprehensive VAA; namely:

1) Who would do it most efficiently?

2) What would be their motives for launching the VAA?

3) What could they do to minimze the risks of launching the VAA?

4) How could voters and parties be persuaded to use it? (This question was to some extent addressed already in the last post, but here I will touch on some further issues.)


Whoever constructs this VAA should ideally have:


1) Access to expertise on politics (preferably political scientists, but renowned political journalists is another option) and on constructing the algorithms, etc, needed to get the site running. This means that they probably should have:

a) A respected name among politics experts. If you don’t have that, it’s less likely that they’re going to want to participate.

b) Substantial resources, either in monetary form or in form of expert labour willing to work for free.

2) Lots of visitors. It would be much harder for an independent site to attract test-takers, than for an established site with thousand of recurring visitors. Therefore, the latter option is strongly preferable.

3) A respected name among potential test-takers, in three ways:

a) They should not be seen as politically biased to the left or right.

b) They should be seen as having a direct interest in improving politics, and not only in increasing their profits.

c) They should not be seen as infringing upon people’s privacy. The reason this is important is that the site would obtain vast amounts of sensitive data, and if users did not trust the site’s handling of that data, they would be put off.


There are lots of different organizations that could launch a comprehensive VAA: large Internet companies like Google, dating sites like eharmony (which recently launched a matchmaking site connecting employers with employees), universities or other non-profits such as Wikipedia or Reddit, major newspapers, social networks, etc. There is no reason to exclude any alternative at this point, but two alternatives which seem particularly interesting in view of the above criteria are, firstly, public service broadcasters such as BBC, which ticks all of the boxes, and, secondly, a joint launch involving one right-wing and one left-wing newspaper (say New York Times and Wall Street Journal). This move would presumably rebut the bias objection which a VAA launched by New York Times alone would face. But there are also other options such as Nate Silver’s (Input here is much appreciated.)


Let us now turn to what possible motives these sites would have for launching the VAA. These can be divided into commercial and non-commercial motives.

1) Starting with the non-commercial motives, serious media and its journalists think that they, as the “fourth estate”, play a vital role in a well-functioning democracy. By collecting and disseminating information of interest to voters, they enhance the democratic process, their theory goes. Now one can debate whether that is true or not (I happen to think that it by and large is) but that is beside the point, which is that this self-image should motivate serious media to launch the VAA. The purpose of the VAA is, namely, precisely to collect and disseminate information in a way that helps voters to make more informed choice.

2) The main commercial motives are as follows.

a) A VAA launched on, say, the BBC, New York Times or Wall Street Journal’s website would presumably attract more visitors to these sites, and make existing visitors spend more time there (and conversely; cf the symbiosis between the dating site Guardian Soulmates and The Guardian.) This would mean more ad revenues (BBC doesn’t have ads on their website but seem very interested in increasing their number of visitors neverthless).

b) Ideally, the statistics on VAA users’ preferences on the different issues would be freely available. However, the hosting site could also choose to sell some sets of statistics, and analyses of them, to political parties and others (this has to be done in a way that respects users’ privacy, however). For instance, parties could get advice on how they should change their policies and/or the relative weights they attach to them, in order to increase their popularity.


There are of course many issues pertaining to the running of the site that one could discuss. For instance: why hasn’t a site like this already been launched? Perhaps this tells us that it’s harder to run it than I’ve made it sound. Business not being a field of expertise of mine, I will refrain from discussing most issues related to the running of the site, however. Instead, I will focus on one way of reducing the risks of launching the VAA which I think is especially important to point out.

As we saw, many news sites already have simple votematch tests today. There is, however, a big difference between them and the kind of comprehensive (and therefore more costly) VAA I envision here. A cautious approach would therefore be to start off by launching a less comprehensive VAA and then expand it step by step. You could launch the test with a moderate number of questions, say 100-200, and if that goes well, gradually expand it. (In order for the site to gain credibility, the first version has, however, to be well-crafted and reasonably comprehensive.)

An important issue to consider here is that users shouldn’t have to re-do their weightings when new versions of the test is launched. One possible method to achieve this is as follows. The first version of the test could exclusive contain quite general questions, whereas new versions would add new, lower, layers in the hierarchy of questions. Default answers to these questions would be generated from users’ answers to the more general questions (in the way specified in the previous post). As a result, users wouldn’t have to adjust their answers or weights to existing questions when new questions are added.


In the last post, I argued that having different layers of questions, a permanent profile, and a robustness score, should increase voters’ willingness to use the site. Let me now discuss some further issues concerning how to get voters and parties to use the site.

Firstly, let me note that it is of course quintessential to get the parties on the boat. Without their answers, the site adminitrators would not only have to fill in the parties’ answers to the various questions, but also fill in the relative weights they attach to them. Especially the latter threatens to be very contestable. It’s much better if you can get the parties to do this for themselves.

I think, however, that there are reasons to believe that the parties would be quite positive towards this idea. Parties and candidates running for the Swedish parliament in the upcoming elections have been willing to fill in answers to up to 49 questions to various votematch tests. Granted, they did not have to fill in the relative weights they attach to the different questions (which may be harder) but I nevertheless think that there is room for optimism. Parties and candidates are generally proud of their stances on different policy issues and happy to let people know about them.

That said, if parties were given questionnaires with thousands of questions, they might refuse to fill them in. This is thus another reason for taking a step-by-step approach, since parties will be more disposed to fill in a smaller number of questions. If the VAA becomes popular, it will be hard for parties to decline to answer additional questions.

Another mechanism that should make parties more inclined to respond to the questionnaire is party competition. If some of the parties in a multi-party system start answering the questions on a large scale, the others may be forced to follow, on pain of seeming opposed to announcing their policies publically.


Let me now turn to two issues which are important both for voters and parties. Firstly, the VAA has to be seen as objective both by voters and parties. In fact the more influential it is, the more objective it must be, or voters and parties will start complaining. In particular, the questions need to be neutrally phrased. If they are phrased in a way that favours a particular party, the others party will be inclined to jump ship. This is an important issue that requires a lot of thought and care.

Secondly, it is important that parties fill in their true answers to, and true weigthings of, the different questions. If parties start filling in false answers or false weigthings on a large scale for strategic reasons, voters’ and other parties’ trust in the VAA will plummet.

Insincere weigthing of questions are probably more likely than insincere answers to the questions, so let us focus on the former. As for an example where insincere weighting of questions pays off, consider the following scenario. Suppose that party A in fact considers question X as more important than question Y, but that they know that their policy on question Y is more popular than that of party B, whereas the converse is true for question X. Then party A has an incentive to falsely weigh question Y as more important than question X.

There are two reasons to believe that this won’t be as big problem as you’d think at first glance. Firstly, media, party grassroots and core voters will object to weightings of questions which are too far removed from what is perceived as the party’s true weighting. Party leaderships are thus severely constrained; they cannot answer these questionnaires in whatever way they think is the most strategic.

Secondly, the weightings will, in effect, function as declaration of the relative strength of your election promises. This means that if you weigh a question very heavily, it will look very bad if you don’t implement it if you become part of the government. This should make parties less inclined to weigh questions insincerely.

To make sure that the parties actually fulfill their promises, a “promise score” could be calculated on the basis of the governing parties’ question answers along with the relative weights they attach to them, on the one hand, and their actual policies when in power, on the other hand. News media of course already do publish articles on broken promises, but a promise score like this would make it much clearer to voters to what extent the government has fulfilled its promises. This would give the government a strong incentive to fulfill its promises, which, in turn, should make them avoid giving promises they could not or would not want to fulfill. Hence such a promise score should force parties’ to fill in the questionnaire sincerely.


Summing up, I have given some reasons for why (potentially efficient) hosting sites, parties, and voters, would be interested in a comprehensive VAA. These reasons are of course not conclusive. I am perfectly aware that there might be major problems that I have not thought of (and invite readers to point them out). At the same time,  the fact that there a) already exist popular dating (and now, as we saw, also employer-employee) matchmaking tests that look quite similar to the test I’m proposing and b) already exist less comprehensive VAA’s, the proposal should not look too outlandish.

I don’t even think that the general idea is very original. What’s important is to get the details right. A website can be great on nine scores out of ten but fail due to a subpar performance on the last one. Hence it’s important to do a fair amount of theorizing and testing before launching the VAA.





Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s