This is Craig Pirrong's Typepad Profile.
Join Typepad and start following Craig Pirrong's activity
Join Now!
Already a member? Sign In
Craig Pirrong
Recent Activity
Bayes' Theorem implies that none of these intrusive surveillance techniques is likely to prove any real ability to predict who will commit mass murder.
Toggle Commented Dec 28, 2012 on Dumb And Dumber at JustOneMinute
Absolutely, Catherine. Voting systems are extremely vulnerable to manipulation for some very basic reasons. This is especially true for social networking sites due to the nature of network effects. It's well known that there are positive feedback mechanisms in networks that lead to tipping to a dominant platform. This is seen in financial markets-it is cheaper to trade where there are more trading partners, so trading activity tends to gravitate to a single exchange. It is seen in software, where "winner take (nearly) all" is the rule in both operating systems and many applications. It is seen in social networking sites-the value of the network to a single user depends on how many others participate. It is seen in telecoms, and on and on. One implication of this is that minor-trivial, even-advantages in one networked system/platform can lead to its utter dominance. Thus, the ability to manipulate even slightly the way in which information is presented can lead to a cascade. For instance, increasing the odds that a particular piece of content is viewed, and then prioritizing subsequent display of that content based on how often it is viewed embed a positive feedback that quickly leads to the dominance of the initially privileged content and the oblivion of the initially disfavored, even if the initial advantage is very minor. All of the code geek gaming tricks that you describe are classic ways of manipulating network effects. This is in the DNA of those who work in the social network world. Learning to harness and manipulate these effects is the key to creating a successful network-and as you note, it has very insidious effects when applied to voting mechanisms, or social choice mechanisms generally. Moreover, there are well known issues with making social choice through voting that make control of the rules tantamount to control of the outcome. Voting is inherently unstable. There are cycles due to non-transitivity. This further means that whoever controls the order in which issues are considered can control the outcome and increase the odds that a proposal that they prefer is adopted, even if it is not the most popular. The discontinuity in payoffs in elections (50 percent plus 1 vote take all) also leads to instability that is controlled in real-world political systems by various checks and balances-but the detailed characteristics of those checks and balances can have decisive effects on the nature-and justness-of the outcome. Voting in a social network setting therefore combines two things that are highly subject to manipulation, and where small changes to rules can have profound effects on outcomes. This means that it is highly dangerous to delegate control over such a combination to a self-appointed and unaccountable clerisy. Code then indeed becomes law, and that is a dangerous thing indeed. We basically become tools of the coders. Since (a) it is wildly inefficient to have everyone become sufficiently proficient in code to keep an eye on these jokers and (b) due to free riding effects inherent in delegated monitoring no individual has a strong incentive to keeping an eye on the jokers because that individual incurs all the cost but reaps only a fraction of the benefit of her monitoring, the ability to control the coders is minimal. The old Lenin question resonates with particular force here: Who-Whom? Which means that you are quite right. This is nothing to kid about. This is serious business. Social choice in social networks uniquely vulnerable to totalitarian outcomes. Be afraid. Be very afraid.
Craig Pirrong is now following The Typepad Team
Nov 25, 2012