Slashdot Mirror

With a little boost from 4chan, a petition for the U.S. government to build a working Death Star has reached 30,000 signatures and counting, over on the White House's Department Of Let's See How Fast We Can Get 75,000 Signatures To Legalize Pot (or as it's officially known, "We The People"). This is the website where any of the member of the public can create a petition that other users can sign, and if the petition receives 25,000 signatures in 30 days, the White House will issue an official response. (Alan Boyle is taking suggestions on how the White House should respond to the Death Star request. How about: "4chan. You will never find a more wretched hive of scum and villainy.")

Cynics will say that the whole process was already a joke anyway. Even looking at the most popular non-Death-Star related petitions on We The People, most of them express standard left- or right-wing positions on hot-button topics in a manner that's extremely unlikely to convert anyone who doesn't already agree. Since everyone already knows that those some large segment of the population holds those positions, nobody would be surprised that any one of those petitions would be able to gather 25,000 signatures, and so there would be no pressure on the White House to change any of their official positions as a result.

On the other hand, I don't think this means that online petitioning can't work. Rather, I think a sligthly different algorithm could greatly improve the quality of the suggestions that get filtered to the top and trigger a response from the White House. At least one algorithm exists that (a) would prevent the system from being "gamed" by any large, organized group (whether 4chan or the NRA); and (b) would reward the petitions that were supported by the highest percentage of the general user population (or, if you prefer, the petitions that were supported by the highest percentage of credentialed experts in a given field).

The algorithm is the same one that I've advocated for preventing cheating on digg, or identifying the best "hidden gems" among newly released songs (and political arguments), or adjudicating Facebook abuse complaints -- have each petition voted on by a random subset of users registered on the We The People site. Based on this random sampling method, the petitions that have the highest percentage of "yes" votes, are assumed to be the ones with the broadest level of support among registered users, and the ones most deserving of a response from the White House.

Example: Suppose there are 250,000 registered users on the We The People site. A user creates a new petition, and somehow manages to pass some "threshold" that is implemented to screen out blatant time-wasters. (Perhaps you have to gain 100 signatures to pass the first threshold. I'd prefer it if you could clear the first hurdle just by paying $5 with a credit card, but this might anger purists who say that petitioning the government should always be free.) Your petition then gets emailed out to 100 randomly selected other users on the site, who vote to either Agree or Disagree. (In practice, in order to get 100 votes cast, you'd have to email more than 100 people, taking into account their response rate. So if only 50% of users respond to an email request for votes, email it to 200 randomly selected users to ensure you get about 100 votes cast.) Then petitions are sorted according to the percentage of users in their sample who voted to Agree. Petitions that got a high percentage of yes-votes, could be forwarded out to a wider audience (say, 1,000 users), to ensure that the initial high percentages of yes-votes wasn't just a fluke. Users in each random sample could also include comments about why they were voting a particular proposal up or down.

This sounds deceptively simple, but it makes it much harder for an organized online movement to hack the system. Say that 4chan manages to get 25,000 registered users in an attempt to push their favored petition to the top. This still means that, on average, their voters will comprise only about 10% of the randomly selected voters in any online poll - possibly enough to give an extra boost to a petition that already had broad support from regular users, but not enough to achieve a coup all by themselves.

Perhaps you'd object that even if such a system could not be manipulated by organized mobs, it would still leave the approval rating in the hands of non-expert ordinary citizens (even if citizens registered on We The People are slightly more informed than average). Whether you think this is a good thing, depends on whether you think the purpose of the site is to reflect the will of the people, or to provide informed advice to the President.

But if you want to get a random sampling of expert opinions, that's pretty easy as well. For petitions on, say, economic matters, just have a subset of users consisting of economics professors from accredited universities across the country. (These credentials would have to be confirmed manually by White House staff, but it's not that hard to verify that someone owns an .edu address and that their university webpage identifies them as an econ professor.) Then any petition on an economic matter could be submitted to a random sample of economics professors to be rated by them. If a petition gets a rating from economics experts that is wildly different from the rating it gets from the general user population, that suggests something interesting is going on (either econ professors are out of touch, or the general public is misinformed). But if a petition gets high levels of support from the public and the relevant expert group, that would seem to justify a response from the White House, much more so than some of the idiotic petitions currently pulling 65,000+ votes on We The People.

Something almost like this has actually been done by the IGM Economic Experts Panel in Chicago, which surveyed a group of 41 economists that the IGM believed to be among the best in the world, representative of the political left, right, and center. The survey found a high degree of consensus on questions that the general public is divided on, such as the fact that 40 out of 41 experts agreed with the statement:

All else equal, permanently raising the federal marginal tax rate on ordinary income by 1 percentage point for those in the top (i.e., currently 35%) tax bracket would increase federal tax revenue over the next 10 years.

To people who have heard celebrity conservative economists claiming that raising marginal tax rates lowers tax revenue, it might come as a surprise that virtually all expert economists in the IGM's sample, including a representative number of self-described conservatives, agreed that it does not. But don't just soak the rich and call it a day; most economists in the IGM's sample also disagreed that:

The cumulative budget shortfalls in the US over the next 10 years can be reduced by half (or more) purely by increasing the federal marginal tax rate on ordinary income for those in the top tax bracket.

Of course those were questions of fact (what economists call positive economics), while petitions address questions of what should be done (what economists call normative economics, and which varies according to your values and goals). But even economists with diverse political leanings often advocate similar policies; NPR interviewed 5 economists spanning the spectrum from left to right, and found across-the-board consensus in favor of 6 proposals, which you can read here. And hey, one of them is legalizing pot!

If We The People implements a system for polling a random sample of economics experts, I think their first order of business should be to have them rate the ideas in that 6-point platform. The five-person panel claimed that all of these ideas have broad support from economists across the political spectrum, but it would be good to know for sure. And for any of those six points that has broad consensus support from experts, it should be incumbent on the White House to declare whether they agree, and if not, why not.

More generally, random-sample voting will always reveal more useful information -- whether about the opinion of the public, or about the opinions of experts -- than a petition site that lets passionate users self-organize into signature mobs. As I've been saying ever since my first story advocating this algorithm, the only site I'm aware of that currently implements random-sample voting correctly, is HotOrNot, which shows users a random series of pictures and lets users rate the picture's hotness on a scale of 1 to 10. Can we not make at least that much effort to design a working system, when it comes to deciding which petitions get a response from the White House?