deliberative nirvana and software design myopia, Mar. 2009 edition

Ages ago, I tried to write a senior thesis about the potentials and pitfalls of bringing deliberative democracy to the internet. The thesis failed, badly. There were a lot of reasons for that failure1 but in the end the biggest reason was that I let the perfect be the enemy of the good. When, at some point during the year, I realized that the internet was (gasp) not going to create a deliberative utopia, I quit altogether- it never once crossed my mind that it might be worthwhile to examine how the internet could fall short of an ideal but still be better than the offline world. In fact, it took until last year- in the midst of the election campaign- for me to have that ‘ah-ha’ moment.

And so now in the back of my mind I keep toting up the little examples of ‘so close, so far’ that keep cropping up. There are tons of them, because to their great credit, the Obama campaign and administration seem determined to push the edges of the possible in this area2. But I do wish that more people had an idea of the issues and values involved, and how merely naively asking questions on the internet can greatly diverge from the nominally democratic values people are trying to advance.

The example that finally spurred me to blog a bit, and try to get some ideas written down, was a post on the google public policy blog titled ‘Citizen participation that scales: a call to action’. It’s a fine little post, noting that the recent Obama ‘Open For Questions‘ was driven by Google’s ‘Moderator’ tool, which (being a Google product) is built to scale virtually infinitely, or at least to happily cope with the 3M+ votes and 100K+ questions. Google pats itself on the back for this:

We think technology can be a force for greater accountability and access between citizens and their elected officials. We’re excited that the White House has chosen to use the power of cloud-based applications like Google Moderator and App Engine to scale the president’s direct dialogue with the American people.

And Google should pat itself on the back for this. This is a big step forward from the insanely skewed filters of the traditional media- it’s impressive to compare the (mostly) substantive nature of the questions being asked by this group with the ‘gotcha’/news cycle driven questions that often make up the average White House press conference.

Of course, Google’s focus on ‘scale’ makes it sound like the only problem here is an engineering problem about how many people can use the system before it bogs down:

[T]hanks to the scale that App Engine provides, this application can now support tens of thousands of people at once. This gives everyone the chance to be heard in a way that gives priority to the issues that matter most to the broader group.

Tens of thousands of people can vote, ergo, we get issues that matter most to the broader group! Technology- specifically, server scaling technology- has solved the problem. No thought given to user interfaces; no thought given to what values those interfaces are expressing.

Not surprisingly the resulting questions have some issues. Most predictably, almost half of the most popular questions (in techpresident’s accounting) were substantive… about marijuana legalization. Now, don’t get me wrong- marijuana legalization is actually a reasonable question to ask the president.3 But does anyone seriously think that the huge number of votes for marijuana-related questions (top three vote-getters in budget, for example) actually represents American public opinion in any reasonable way? In fact, the huge number of marijuana questions actually represents a transparent attempt to game the system. That the system was gamed did not come as a surprise to anyone who has thought about the problems of democracy online. Treating the problem as merely an exercise in scaling up a very simple question tool designed for well-intentioned, very homogenous users – Google engineers – was a recipe for a mess in the much more complicated real world, where anti-gaming and moderation techniques are a must have.4

Even if, miraculously, no one choose to game the system like NORML and others apparently did, there are all kinds of other potential design issues with software built for democracy-scaled online deliberation. Most notably, unlike the small, homogenous group of Google employees for which this tool was first built, American politically engaged computer users are not at all representative of America as a whole.5 For example, we are extremely, extremely unlikely to have had friends killed by the police, so one important perspective in the discussion over criminal justice reform is unlikely to ever get reasonable representation in a forum like Open For Questions, no matter how much scale the backend can provide. Biases of this sort- who has more access to technology? who is more likey to use it? who is more likely to use it effectively? who will game it and how?- are of course impossible to eliminate merely with software design, but the google post (and virtually all other coverage of the Open For Questions experiment) have been shockingly devoid of skepticism of the design of the software. They all seem to blithely assume that you can just throw up a polling tool on the web, and voila, democracy.

Again, I don’t mean to be completely negative here- my thesis was torpedoed by that. Like Carolus and the early TV innovators, Google, the Drupal team, and others are doing valuable work, and this technology improves a great deal on letters to the editor and other ancestors which were also badly gameable. We shouldn’t throw this baby out with the bathwater. At the same time, it is very easy6 to ignore the deeper, less obvious ramifications on democracy of the design of the code that we use- who participates? under what conditions? how does UI design affect those things? We should all be sensitive to these limitations and constantly demand better of the technology that (more and more) is going to significantly control how we relate to our government and to each other.

  1. Krissa moved to Africa; my advisor was not technically savvy; it was a fallback topic; etc. []
  2. see, e.g., http://www.aaronsw.com/weblog/rssstimulus []
  3. Even those who (like me) don’t smoke should be very concerned about the cost of imprisonment, drug violence, and lost potential tax revenues; the president’s dismissive answer reflects poorly on him. []
  4. Though careful they don’t get too complicated, or else they’ll scare off non-technical people and lead to accusations of non-transparency. Yes, I’m talking about you, slashdot. []
  5. Almost certainly more representative than newspaper editors or TV network owners, but still, not representative. []
  6. particularly for engineers, but also for non-engineers who don’t fully grasp the implications and limitations of technology []