Ethan Zuckerman reviews Infotopia and discusess social decision-making

On My heart’s in Accra Ethan Zuckerman reviews Cass Sunstein’s book “Infotopia”, which discusses how the internet changes group decision making processes. Zuckerman writes:

Infotopia… In his new book, Infotopia, [Sunstein’s] become a cyber-enthusiast to an extent that would have been hard to imagine a few years ago. Specifically, he’s excited about the ways new online tools make it possible for groups of people to assemble information and accumulate knowledge. He’s become a devotee of Friedrich Hayek, the Austrian economist who saw markets, first and foremost, as a way to aggregate information held by a large group of people. There’s ample evidence that Hayek was right in an examination of the failure of planned economies – smart men sitting in a room do a far worse job of setting the price of copper ore or bread than the collected actions of thousands of consumers, iterated over time.

Markets aren’t the only way to aggregate information from a large group of people. Deliberative groups, where a set of people get together and share the knowledge they have on a problem or an issue, are favored by many political theorists, including Jürgen Habermas, who bases much of his political philosophy on the establishment of a public sphere where deliberation can occur. Sunstein is deeply suspicious of the optimistic claims made for deliberation, and cites a wealth of studies that demonstrate that deliberation, in many cases, leads to bad decisions and the reinforcement of extreme views.

(You can think of Infotopia as a caged deathmatch between Hayek and Habermas, streamed live on the Internet. Habermas taps out somewhere around page 200.)

Inspired in part by James Surowiecki’s “The Wisdom of Crowds”, Sunstein looks at the uncanny ability of large groups to jointly guess the number of beans in a jar. While individual guesses can be far from the mark, the mean guess is often surprisingly close to the actual number. This phenomenon is well explained by the Condorcet Jury Theorem, which suggests that a group of individuals – all of which have a better than even chance of making a correct decision – will have a greater chance of making the correct decision as the size of the group increases. In other words, if we’re pretty good at guessing the number of beans in a jar, lots of us working together are likely to be excellent at the same task.

Unfortunately, the converse is true as well. If each of us have less than a 50% chance of being right about a decision, a group of us will be worse at making a correct decision, with our probability of accuracy increasing towards zero as the size of the group increases. Ignorance can lower our chances of making an accurate decision, but so can political bias and preconception.

One would hope that deliberation could solve this problem – if a small number of people in a group are knowledgeable about a subject, perhaps they can convince the others of the accuracy of their claims and move the group to a result better than the mean of all their preconceptions. This turns out to be true for at least one set of problems: “eureka problems”, where the correctness of a solution is obvious to everyone once it’s been articulated. (Most chess problems, for instance, are “eureka” problems – they end in checkmate, which any player can see once it’s been diagrammed.)

But deliberation on other types of problems has a more complex outcome. Sunstein organized an experiment in deliberation with David Schkade (of the University of San Diego) and Reid Hastie (of the University of Chicago). The professors invited a set of Colorado citizens from two communities – liberal Boulder and conversative Colorado Springs – to come to local universities and deliberate three divisive political topics: global warming, affirmative action and civil rights. The groups – 5-7 randomly selected citizens from the same community – had a strong tendency to become more politically polarized over the course of the discussion. Liberals became more liberal, conservatives more conservative, and the range of ideological diversity in each group decreased.

Explaining the finding, Sunstein offers a number of explanations, each backed up by other research studies. In a group setting, people will often gravitate towards a strongly stated opinion, especially if their own opinions aren’t fully formed. An ideologically coherent group is likely to repeat a great deal of evidence for one side of an issue, giving more reinforcement for that viewpoint. People find it difficult to defy the will of a group, and may polarize to avoid interpersonal conflict.

Sunstein makes a great deal of this finding in the book, though the paper he and colleagues authored suggests that the constraints of the experiment need to be considered very carefully. The groups the researchers constructed were fairly ideologically homogenous, and no attempt was made to give the deliberating groups any information that might challenge their pre-established points of view. Experiments in deliberation which forced ideological mixing and gave both sides information to present their case had very different outcomes – Sunstein points to some of the work done by James Fishkin which finds some reasons for optimism in more guided deliberation. Sunstein and his co-authors argue that the geographic – and hence ideological – filtering they use in the experiment is a very realistic one in modern-day America, where individuals are increasingly choosing to live in communities where their neighbors share their ideologies. Still, it’s a disappointment that Sunstein doesn’t address some of the possible upsides of deliberation more closely.

Whether or not I agree with all of Sunstein’s conclusions, his quest for systems that aggregate knowledge across networks is an exciting way to look at the contemporary Internet. A large number of the most interesting projects taking place on the Internet use strategies to aggregate information from multiple users to create new knowledge – this is the magic behind Google’s PageRank algorithm, Digg’s headlines and Amazon’s collaborative filtering recommendations. Analyzing these systems in terms of their effectiveness in getting people to reveal hidden knowledge is, in my opinion, an excellent framework for evaluation. (I’m very interested, for instance, in thinking through how the folksonomy and taxonomy systems David Weinberger is exploring in his forthcoming “Everything Is Miscellaneous” use different mechanisms to assemble information from different actors to organize information.)

Leave a Reply

Your email address will not be published. Required fields are marked *