This is a real story of a pair of identical twins who are suspected in a crime. There is CCTV and DNA evidence that could implicate either suspect. Detailed DNA testing that could resolve the guilty twin is prohibitively expensive. So both have been arrested in the hope that one may confess or implicate the other.
Entries Tagged "game theory"
Page 2 of 2
This is the sort of thing I wrote about in my latest book.
The Prisoners Dilemma as outlined above can be seen in action in two variants within regulatory activities, and offers a clear insight into why those involved in regulation act as they do. The first relationship is that between the various people and organisations being regulated banks, nuclear power stations, council departments, police agencies, journalists, etc, and the clear lessons from history are that even for those organisations that are theoretically in competition with each other, it is beneficial to both/all sides in the long run to use mutual cooperation in order to maximise their personal benefit. Whether it was Virgin and British Airways forming an illegal cartel to fix the price of fuel surcharges (a benefit to themselves which was paid for in increased prices for passengers); football shirt retailers (and Manchester United) being fined £16m for fixing the price of replica football shirts, or Barclays (and undoubtedly other banks) working together to fix the LIBOR rate, the reason why they do it is simple and unanswerable—it is in their benefit to do so.
However, when it comes down to the relationship between the regulators and those being regulated, then a completely different strategic dynamic comes into play. The ability of the regulated organisation to maximise personal benefit is then based on the ability to predict what the other side will do in response to the two options cooperate (play nicely) or betray (screw the customer). Given that in almost all cases the regulatory body has less funds, personnel, resources and expertise than the organisation it is regulating, then it becomes clear that there is little to be gained in the long run by cooperating / playing nicely, and much to be gained by ignoring the regulator and developing a strategy that focuses purely on maximising its own personal benefit. This is not an issue of ‘right’ or ‘wrong,’ but purely, in its own terms at least (maximisation of profit, increased market share, annual bonuses, career prospects), of whether it is ‘effective’ or ‘ineffective.’
In Liars and Outliers, I use the metaphor of the Prisoner’s Dilemma to exemplify the conflict between group interest and self-interest. There are a gazillion academic papers on the Prisoner’s Dilemma from a good dozen different academic disciplines, but the weirdest dataset on real people playing the game is from a British game show called Golden Balls.
In the final round of the game, called “Split or Steal,” two contestants play a one-shot Prisoner’s Dilemma—technically, it’s a variant—choosing to either cooperate (and split a jackpot) or defect (and try to steal it). If one steals and the other splits, the stealer gets the whole jackpot. And, of course, if both contestants steal then both end up with nothing. There are lots of videos from the show on YouTube. (There are even two papers that analyze data from the game.) The videos are interesting to watch, not just to see how players cooperate and defect, but to watch their conversation beforehand and their reactions afterwards. I wrote a few paragraphs about this game for Liars and Outliers, but I ended up deleting them.
This is the weirdest, most surreal round of “Split or Steal” I have ever seen. The more I think about the psychology of it, the more interesting it is. I’ll save my comments for the comments, because I want you to watch it before I say more. Really.
For consistency’s sake in the comments, here are their names. The man on the left is Ibrahim, and the man on the right is Nick.
EDITED TO ADD (5/14): Economic analysis of the episode.
Behzad Zare Moayedi, Mohammad Abdollahi Azgomi, “A Game Theoretic Framework for Evaluation of the Impacts of Hackers Diversity on Security Measures,” Reliability Engineering & System Safety, 99 (2012): 45-54 (full article behind paywall).
Abstract: Game theoretical methods offer new insights into quantitative evaluation of dependability and security. Currently, there is a wide range of useful game theoretic approaches to model the behaviour of intelligent agents. However, it is necessary to revise these approaches if there is a community of hackers with significant diversity in their behaviours. In this paper, we introduce a novel approach to extend the basic ideas of applying game theory in stochastic modelling. The proposed method classifies the community of hackers based on two main criteria used widely in hacker classifications, which are motivation and skill. We use Markov chains to model the system and compute the transition rates between the states based on the preferences and the skill distributions of hacker classes. The resulting Markov chains can be solved to obtain the desired security measures. We also present the results of an illustrative example using the proposed approach, which examines the relation between the attributes of the community of hackers and the security measures.
New paper: Dengpan Liu, Yonghua Ji, and Vijay Mookerjee (2011), “Knowledge Sharing and Investment Decisions in Information Security,” Decision Support Systems, in press.
Abstract: We study the relationship between decisions made by two similar firms pertaining to knowledge sharing and investment in information security. The analysis shows that the nature of information assets possessed by the two firms, either complementary or substitutable, plays a crucial role in influencing these decisions. In the complementary case, we show that the firms have a natural incentive to share security knowledge and no external influence to induce sharing is needed. However, the investment levels chosen in equilibrium are lower than optimal, an aberration that can be corrected using coordination mechanisms that reward the firms for increasing their investment levels. In the substitutable case, the firms fall into a Prisoners’ Dilemma trap where they do not share security knowledge in equilibrium, despite the fact that it is beneficial for both of them to do so. Here, the beneficial role of a social planner to encourage the firms to share is indicated. However, even when the firms share in accordance to the recommendations of a social planner, the level of investment chosen by the firms is sub-optimal. The firms either enter into an “arms race” where they over-invest or reenact the under-investment behavior found in the complementary case. Once again, this sub-optimal behavior can be corrected using incentive mechanisms that penalize for over-investment and reward for increasing the investment level in regions of under-investment. The proposed coordination schemes, with some modifications, achieve the socially optimal outcome even when the firms are risk-averse. Implications for information security vendors, firms, and social planner are discussed.
Companies would be better off if they all provided meaningful privacy protections for consumers, but privacy is a collective action problem for them: many companies would love to see the ecosystem fixed, but no one wants to put themselves at a competitive disadvantage by imposing unilateral limitations on what they can do with user data.
The solution—and one endorsed by the essay—is a comprehensive privacy law. That reduces the incentive to defect.
The big news in professional bicycle racing is that Floyd Landis may be stripped of his Tour de France title because he tested positive for a banned performance-enhancing drug. Sidestepping the entire issue of whether professional athletes should be allowed to take performance-enhancing drugs, how dangerous those drugs are, and what constitutes a performance-enhancing drug in the first place, I’d like to talk about the security and economic issues surrounding the issue of doping in professional sports.
Drug testing is a security issue. Various sports federations around the world do their best to detect illegal doping, and players do their best to evade the tests. It’s a classic security arms race: improvements in detection technologies lead to improvements in drug detection evasion, which in turn spur the development of better detection capabilities. Right now, it seems that the drugs are winning; in places, these drug tests are described as “intelligence tests”: if you can’t get around them, you don’t deserve to play.
But unlike many security arms races, the detectors have the ability to look into the past. Last year, a laboratory tested Lance Armstrong’s urine and found traces of the banned substance EPO. What’s interesting is that the urine sample tested wasn’t from 2005; it was from 1999. Back then, there weren’t any good tests for EVO in urine. Today there are, and the lab took a frozen urine sample—who knew that labs save urine samples from athletes?—and tested it. He was later cleared—the lab procedures were sloppy—but I don’t think the real ramifications of the episode were ever well understood. Testing can go back in time.
This has two major effects. One, doctors who develop new performance-enhancing drugs may know exactly what sorts of tests the anti-doping laboratories are going to run, and they can test their ability to evade drug detection beforehand. But they cannot know what sorts of tests will be developed in the future, and athletes cannot assume that just because a drug is undetectable today it will remain so years later.
Two, athletes accused of doping based on years-old urine samples have no way of defending themselves. They can’t resubmit to testing; it’s too late. If I were an athlete worried about these accusations, I would deposit my urine “in escrow” on a regular basis to give me some ability to contest an accusation.
The doping arms race will continue because of the incentives. It’s a classic Prisoner’s Dilemma. Consider two competing athletes: Alice and Bob. Both Alice and Bob have to individually decide if they are going to take drugs or not.
Imagine Alice evaluating her two options:
“If Bob doesn’t take any drugs,” she thinks, “then it will be in my best interest to take them. They will give me a performance edge against Bob. I have a better chance of winning.
“Similarly, if Bob takes drugs, it’s also in my interest to agree to take them. At least that way Bob won’t have an advantage over me.
“So even though I have no control over what Bob chooses to do, taking drugs gives me the better outcome, regardless of what his action.”
Unfortunately, Bob goes through exactly the same analysis. As a result, they both take performance-enhancing drugs and neither has the advantage over the other. If they could just trust each other, they could refrain from taking the drugs and maintain the same non-advantage status—without any legal or physical danger. But competing athletes can’t trust each other, and everyone feels he has to dope—and continues to search out newer and more undetectable drugs—in order to compete. And the arms race continues.
Some sports are more vigilant about drug detection than others. European bicycle racing is particularly vigilant; so are the Olympics. American professional sports are far more lenient, often trying to give the appearance of vigilance while still allowing athletes to use performance-enhancing drugs. They know that their fans want to see beefy linebackers, powerful sluggers, and lightning-fast sprinters. So, with a wink and a nod, they only test for the easy stuff.
For example, look at baseball’s current debate on human growth hormone: HGH. They have serious tests, and penalties, for steroid use, but everyone knows that players are now taking HGH because there is no urine test for it. There’s a blood test in development, but it’s still some time away from working. The way to stop HGH use is to take blood tests now and store them for future testing, but the players’ union has refused to allow it and the baseball commissioner isn’t pushing it.
In the end, doping is all about economics. Athletes will continue to dope because the Prisoner’s Dilemma forces them to do so. Sports authorities will either improve their detection capabilities or continue to pretend to do so—depending on their fans and their revenues. And as technology continues to improve, professional athletes will become more like deliberately designed racing cars.
This essay originally appeared on Wired.com.
Interesting paper: “Passenger Profiling, Imperfect Screening, and Airport Security,” by Nicola Persico and Petra E. Todd. The authors use game theory to investigate the optimal screening policy, in a scenario when there are different social groups (separated by felons, race, religion, etc.) with different preferences for crime and/or terrorism.
Interesting research paper by Shishir Nagaraja and Ross Anderson. Implications for warfare, terrorism, and peer-to-peer file sharing:
Often an attacker tries to disconnect a network by destroying nodes or edges, while the defender counters using various resilience mechanisms. Examples include a music industry body attempting to close down a peer-to-peer file-sharing network; medics attempting to halt the spread of an infectious disease by selective vaccination; and a police agency trying to decapitate a terrorist organisation. Albert, Jeong and Barabási famously analysed the static case, and showed that vertex-order attacks are effective against scale-free networks. We extend this work to the dynamic case by developing a framework based on evolutionary game theory to explore the interaction of attack and defence strategies. We show, first, that naive defences don’t work against vertex-order attack; second, that defences based on simple redundancy don’t work much better, but that defences based on cliques work well; third, that attacks based on centrality work better against clique defences than vertex-order attacks do; and fourth, that defences based on complex strategies such as delegation plus clique resist centrality attacks better than simple clique defences. Our models thus build a bridge between network analysis and evolutionary game theory, and provide a framework for analysing defence and attack in networks where topology matters. They suggest definitions of efficiency of attack and defence, and may even explain the evolution of insurgent organisations from networks of cells to a more virtual leadership that facilitates operations rather than directing them. Finally, we draw some conclusions and present possible directions for future research.
I very am interested in this kind of research:
Network Structure, Behavioral Considerations and Risk Management in Interdependent Security Games
Interdependent security (IDS) games model situations where each player has to determine whether or not to invest in protection or security against an uncertain event knowing that there is some chance s/he will be negatively impacted by others who do not follow suit. IDS games capture a wide variety of collective risk and decision-making problems that include airline security, corporate governance, computer network security and vaccinations against diseases. This research project will investigate the marriage of IDS models with network formation models developed from social network theory and apply these models to problems in network security. Behavioral and controlled experiments will examine how human participants actually make choices under uncertainty in IDS settings. Computational aspects of IDS models will also be examined. To encourage and induce individuals to invest in cost-effective protection measures for IDS problems, we will examine several risk management strategies designed to foster cooperative behavior that include providing risk information, communication with others, economic incentives, and tipping strategies.
The proposed research is interdisciplinary in nature and should serve as an exciting focal point for researchers in computer science, decision and management sciences, economics, psychology, risk management, and policy analysis. It promises to advance our understanding of decision-making under risk and uncertainty for problems that are commonly faced by individuals, organizations, and nations. Through advances in computational methods one should be able to apply IDS models to large-scale problems. The research will also focus on weak links in an interdependent system and suggest risk management strategies for reducing individual and societal losses in the interconnected world in which we live.
Sidebar photo of Bruce Schneier by Joe MacInnis.