OKCupid for voters: effects


In this third post on OKCupid for voters (previous ones: 1, 2), I discuss the possible effects such a VAA could have. The possible effects fall under three categories:

1. Effects on voter behaviour.

2. Effects on parties’ behaviour.

3. Effects on our views of the political system.

These possible effects obviously also function as reasons for or against my proposal.


Firstly, there is some empirical evidence showing that existing vote match tests have increased voter interest in politics, and it seems reasonable that more comprehensive tests, which include more questions, would increase the interest in politics further. The consensus position is of course that this would be a boon.

Another way in which OKCupid for voters could influence voter behaviour is by persuading them to vote for a party they would not otherwise have voted for. Now we know, though, that people often disregard the advice of statistical prediction rules, which I compared to VAA’s in the first post. However, there is reason to believe that people would take a different view on OKCupid for voters. Unlike the statistical prediction rules, OKC for voters will incorporate (virtually) all reasons a voter possibly could want to take into account when making their decision. By contrast, statistical prediction rules typically exclude lots of factors people deem relevant, which makes them less inclined to trust them. Also, we know that some voters already now do follow the simpler existing VAA’s advice, and there are reasons to believe that they would be more inclined to follow OKCupid for voters’ advice. Firstly, it would be more reliable and have a better reputation. Secondly, test-takers would invest more time in it, and hence they could be more disposed to justify this investment by following its advice. (This is of course not a very good reason for following OKCupid for voters’ advice, but it can nevertheless be effective.)

Suppose, then, that voters to at least some degree would heed OKCupid for voters’ advice. Would that influence in which sorts of reasons they would ground their vote? Yes, I think it would. To see that, note that political scientists and others have argued that voter behaviour is heavily influenced by emotional factors. For, instance, research shows that voters are much more likely to vote for competent-looking candidates. Similarly, many voters tend to weigh loyalty to their political “tribe” heavily when they decide whom to vote for (or so many have argued). For instance, some working-class British voters might vote Labour, or some white Southerners republican, because they think that they express their identity, rather than because of policy agreement.

However, even though many voters are in fact heavily influenced by such factors, few would argue that they should be heavily influenced by them. Instead, I hypothesize, most voters would, if asked, say that their decision mostly should be based on the basis of policy issues (e.g. taxes, welfare, etc). OKCupid for voters give them the chance to vote on the basis of this considered preference of theirs, rather than on the basis of their gut-reaction. Hence if voters would start following the advice of OKCupid for voters on a grand scale, that would presumably lead to tribalism, party leaders’ looks and charisma, scandals, and other emotional factors having a lesser impact, and parties’ stances on policy issues having a correspondingly greater impact, on voters’ decisions. Generally speaking, it would mean that voters would use their slow System 2 more, and their fast System 1 less.

This does not, though, mean that voters no longer will use System 1 at all. For instance, voters no doubt often form beliefs on policy issues using System 1. Many of these beliefs are biased (more on this below). However, OKCupid for voters should still go some way towards helping people voting in a more systematic and reason-based way.


Changed voter behaviour will force the parties to change. If voters start voting in a more issue-based fashion, parties will, obviously, be forced to adhere more closely to voters views’ of these policy issues. That goes particularly for issues that voters weigh heavily. Similarly, voters paying less attention to emotionally laden factors will cause parties to spend less energy on them.

There is also another mechanism by which widespread use of OKCupid for voters could make parties adjust their policy stances to fit those of the voters. Today it is not very salient to the parties (and especially not to their vocal grassroots) how changes in stances on policy issues will impact their popularity among voters. Party grassroots, which tend to be more extreme than the party leadership, often fool themselves into believing that policies which are disliked by the median voter could help them win elections. If voters started voting in the algorithmic fashion sketched here, it would become clearer how changed stances on policy issues actually do change parties’ popularities. That would penalize parties which fail to satisfy the mainstream of the voters.

For instance, suppose that a certain party has 25 % of the votes of OKCupid for voters’ members, and that this was known (it could be shown on the site). Then a party congress is held, and the grassroots force the party leadership to take a new, controversial, stance on an important question. That causes the party to drop to 20 % among OKCupid for voters’ members. This would make it perfectly clear for everyone that it was precisely this change that caused the popularity to drop.

These mechanisms would thus cause parties to align their promises with those of the median voter. The promise score, outlined in the last post, would, in turn, force politicians to keep their promises. In conjunction, this would mean that the government’s policies would mirror the views of the majority voters better than they do today.

To some people, it’s obvious that this is a plus, since it would mean that politics would become more “democratic”, in a sense. Others would argue that voters are systematically mistaken both on factual questions and on value questions, and that more voter influence is therefore not necessarily a good thing. For instance, Bryan Caplan argues in his highly interesting The Myth of the Rational Voter that most voters hold a number of false factual beliefs on economics (examples include a protectionist bias and the tendency to equate economic growth with job creation). Similarly, liberals often argue that voters hold irrationally conservative beliefs on value issues (such as gay marriage, European integration and immigration) and that to the extent that politicians have pursued more progressive policies than those of the majority of the voters, they have been justified in doing so.

People arguing along these lines claim, in effect, that at least on some questions, politicians, experts on various factual matters, and other politically influential groups, hold views which are in some sense superior to those of the average voter, and that they therefore should have a disproportionate influence. According to them, increased voter influence thus threatens to lead to worse policies.

There are two ways of dealing with this problem (if you think that it is a problem). Firstly, you could try to improve voters’ beliefs – i.e. make them more truth-tracking and/or more ethical. Here my method for identifying biases by reverse engineering people’s belief and preference structures could be used. By adding different politically controversial factual questions to OKCupid for voters, you could, using the method specified in a previous post, give voters an estimate of to what degree they are politically biased. This could potentially incline them to improve their cognitive practices. (I will cover more ways of identifying biases in future posts on reverse engineering of belief and preference structures.) You could also try to educate the voters in more conventional ways, of course.

Secondly, you could outsource more more decision-making to experts. Already today, many political decisions are taken by experts rather than by democratically elected politicians. Salient examples include central bank interest rate decisions, taken by independent civil servants, and judicial verdicts given by courts (many of which – e.g. those given by supreme courts – plausibly could be seen as “political”). Those who fear that OKCupid for voters could lead to a surge in populistic policies could campaign for more such explicit expert rule.


If OKCupid for voters would succeed in changing voter and party behaviour, it could potentially lead to a major discussion on how the political system should work. Such a discussion is sorely needed. Most Western democracies have seen little change of the political system in hundred years (when universal suffrage was introduced). Hence today’s political system, with its four-year election cycles (plus/minus a year or two), with its specific party-based representative system, etc, is a product of the knowledge and technological constraints of the 1910’s. It is thus unlikely, to say the least, that the present political system is optimal given our present technology (i.e., Internet, by which voters can express their opinions in a much faster and more efficient way than they could hundred years ago) and relevant knowledge (e.g. in political science).

Indeed, some academics and pressure groups such as the British Unlock Democracy and the American Citizens for Political Reform do argue for radical change of the political system, but by and large, their calls go unheeded.  On election day, substantive issues such as taxes and public spending trump meta-level questions regarding the nature of the political system. As many other institutions, the political system is inherently conservative, and it is unlikely that merely arguing that it should change will have any effects. That method has been tested for quite some time by now.

If OKCupid for voters started influence voters and parties in the way sketched above, that would, however, change the rules of the game. For instance, if people continuously fill in answers to political questions, they might start questioning why they have to wait several years to execture power directly. Hence, they might call for more frequent elections. A more radical conclusion is to discard parties and politicians entirely. You could let policy be decided directly by voters’ weighed preferences on the different issues (as expressed in OKCupid for voters) – to cut out the middlemen, which politicians essentially are, as it were. Under such a system, policy would be determined by voters and implemented by civil servants (just like today) but you would not have any politicians in between.

Another issue that would be debated more seriously is the aforementioned question of the relative powers of experts and ordinary voters. The political elites are not very explicit on this question today. From time to time, they give high-minded sermons about the will of the people, but in practice, they prefer some degree of expert rule both regarding factual questions and value questions. So do I, but I think that this should be explicitly stated and argued for.

Exactly where a debate such as this would lead us is of course hard to predict, but my hunch is that it is unlikely that it would lead to a worse system than that we have today, and most likely to a better one. After all, we do have some ability to predict how different political systems would work.


OKCupid for voters would also have other effects (for instance, it would create vast troves of data which would be very useful to political scientists and others) but I can’t cover them here.

As a final note, let me point out that OKCupid for voters would, of course, be part of the drive towards algorithmization that we see in so many areas of society. Sooner or later, this drive will affect politics. It may not come in this form, and it may not come any time soon, but sooner or later it will surely come.


Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s