How to Prevent A Cylon Attack Using Game Theory

In my forthcoming book on game theory and social media marketing (See how I slipped a mention of my book in there? Bet you hardly noticed. That’s called contextual marketing. It works because it’s subtle.), I set out to prove that game theory has something to teach us about how the testy relationship between marketers and consumers has evolved online over the last decade. In fact, game theory worked so well in analyzing this conflict that I now use it to analyze all conflicts, including why my neighbor insists on heaving his dead Christmas tree into the alley behind my house every year.

And since all marketing and no fun makes Eric a dull blogger (it’s one of the factors, anyway), I thought I’d set out to prove just how fun game theory can be by demonstrating its relevance to Battlestar Galactica, Season 3, Episode 6: A Measure of Salvation. Yes, it is actually possible to out-geek sci-fi.

I know, I know, the show ended like three years ago. I fully admit that I am late to the cause of BSG’s greatness, but I am filled with the evangelistic fervor of the recently converted. In the episode in question, the humans, who have been relentlessly pursued by the Cylons throughout the series, stumble upon biological weapons that have the potential to destroy the Cylons for good. These weapons are, in fact, diseased Cylons captured by the humans. When Cylons are killed, they are automatically “resurrected” among their own kind, so executing the captured Cylons would cause them to transmit the disease to other Cylons, setting off a chain reaction that could wipe out their species.

I’ll spare you a recap of the many brooding conversations that precede the humans’ decision to use the biological weapons; the upshot is that they use the weapons, or attempt to, and it is this decision that stirred me from the couch and prompted me to write this post, convinced as I am that game theory would have led them down a different decision path entirely.

This actually isn’t much of a stretch. Game theory was originally developed to analyze geopolitical conflict in the Cold War, especially in regard to the use of potentially world-ending nuclear weapons. As you may have heard, we didn’t end up launching any nuclear weapons in the Cold War, but they were very useful anyway in the practice of deterrence, which was basically a matter of convincing the other side that you had the weapons and were willing to use them, so that you wouldn’t have to use them. If you think deterrence worked out better for us than the grim alternative, then you have the original game theorists (partly) to thank for it.

Deterrence, I’m suggesting, is the other option available to the humans on BSG, though it’s never discussed – it’s all a lot of hand-wringing and should-we-or-shouldn’t-we. But it’s a real option. In game theory, the most basic conflicts are analyzed in terms of two options: cooperation (pursuing self-interest that aligns with your opponent’s self interest) and defection (pursuing self-interest that is contrary to your opponent’s self-interest). Self-interest is always a given; we are not altruists when it comes to our survival.

Deterrence is an interesting gambit: it is a form of cooperation with a strong threat of defection. You are demonstrating to your opponent that you’re more than willing to defect in the next round if they don’t immediately cooperate, and you have to be willing to follow through on that threat in order for future deterrence to carry any weight. In iterative conflicts –the Cylon-human conflict has been repeated for five seasons – deterrence often emerges because the cost of defection is very high, as both sides engage in a downward spiral of bloodletting.

By Season Three of BSG, both sides are sick of the fight, but neither trusts the other enough to cooperate; each believes the other will always end up defecting. The decision to use the biological weapons threatens to worsen the downward spiral. Since it is implausible that the weapons would succeed in wiping out all of the Cylons, the humans would then face a decimated foe that believes the humans are capable of anything and must be destroyed at all costs.

Using the weapons would have a positive outcome for the humans in the short run, but in the long run the outcome would be irreversibly bad for all concerned. This is a classic game theory dilemma: weighing the cost of an action that produces a short-term win against the risk of a long-term loss.

The humans and the Cylons find themselves in this dilemma because they both violated, from the other’s point of view, a cardinal rule of game theory in iterative conflicts: Never be the first to defect. The Cylons’ preemptive strike against the humans, which destroyed most of civilization, convinced the humans that the Cylons want nothing more than their total annihilation. And the Cylons claimed that they attacked in the first place because the humans’ willingness to destroy their own creation – the Cylons – proved they could never be trusted. When both opening moves are defection, cooperation is nearly impossible to achieve.

But deterrence offers a possible way out: it is a way of signaling that you are capable of inflicting a devastating act of defection, but you choose to cooperate instead, and so it engenders a degree of trust. On BSG, deterrence could involve making the Cylons aware that the humans possess the biological weapons and have the means to keep them safe until they can be used (it’s been established that the diseased Cylons can be kept alive, though not cured, with medication). The humans would only use the weapons if the Cylons defected, thus preserving their potency as a deterrent. If the humans could continue to wield the deterrence with some authority, a fragile truce could take hold.

Easier said than done, both in life and on BSG. Game theory analysis assumes perfectly rational action, and the chief proponents of using the biological weapons, Admiral Adama and his son, both have major daddy issues that cloud their judgment. Remind you of any recent office-holders? The final decision-maker, President Roslin, is locked into a story arc that includes an authoritarian streak running through her mother-hen persona. And how much fun would a show about deterrence be?

The point is, well, the point is I love me some BSG. But the other point is that deterrence is too often the option we overlook, because we’re drawn to the unhealthy extremes of either doing everything or doing nothing. The middle path is rocky. But the fate of humanity lies in the balance. So say we all.


The Zero-Sum Game of Online Advertising

The following is an excerpt from Chapter Three of my forthcoming book: Social Media Marketing: A Game Theory Perspective, to be published by Springer in June 2010. For more excerpts, updates, an opinions on game theory and social media marketing, follow me on Twitter @unsettler, or email me at

It all began innocently enough, with a fuzzy rectangular graphic perched atop a page on October 25, 1994. The world’s first banner ad read, “Have you ever clicked your mouse right here? YOU WILL.”With stunning prescience, AT&T had extended to the Web its popular “You Will” campaign, which predicted future consumer technology, into a prediction that users would blindly click on a banner ad that offered nothing specific in return (D’Angelo). Remarkably, users did click, and that first click set Web marketing down a zero-sum path from which it is only now recovering. For nearly a decade, the click was all that mattered. It was a measurable action that brought the user in direct contact with the offer. In other words, it most closely resembled the zero-sum game of direct mail, with even better measurability. And because banner ads could be switched out easily, the ability to improve the minimax point through randomization was vastly simplified, if often overlooked.

The obvious problem is that banner ads are only partly like direct mail. For the most part, direct mail’s practical purpose is simply to get consumers to respond. If the consumer throws the envelope unopened in the trash, it accomplishes nothing. But banners could do more. As with print and broadcast advertising, the banner appears alongside free or subsidized consumer content and helps to offset its cost. As in these other media, consumers can absorb a “brand impression” while they focus on other content.

And marketers generally agree, though they may lack the game theory framework to describe it, that a brand impression sits outside of the zero-sum game. Branding is not directly transactional; it demands no immediate action by the consumer, allowing instead for the cumulative impact of repeat exposure. In its purest form, branding is a form of cooperation, inviting the consumer to participate emotionally in defining the product’s meaning. The brand marketer seeks a long-term relationship that depends on consumer goodwill in a way that direct response marketing does not.

There’ll be more on where branding fits in to game theory later. The point here is that banner advertising stood at those divergent paths from the start, and it took the path more travelled, consigning itself, perhaps forever, to the realm of direct response. The allure was irresistible: here was a medium that offered immediate, highly measurable feedback on its effectiveness, allowing the marketer to track the actual value of a given ad and media placement.

If marketers had known how that value would fluctuate, they might have chosen a different path for the medium from the start. Recall the previous axiom that any single direct market technique over a long enough span of time will produce an inexorable shift in the equilibrium point toward the consumer. It’s also axiomatic that marketers will chase their losses with more aggressive direct response tactics, producing short term gains but ultimately making a bad situation worse.

And that is, in essence, what happened to banner advertising. Fearful of missing out on the next big thing, advertisers threw money at the Web. Publishers, trying to gain dominance quickly in the race to monetize content online, obligingly raised rates. In 1998, advertisers could expect to pay an average of $37 for every 1,000 impressions (Morgan Stanley Dean Witter), which was made digestible only by the 1-2% response rates that the ads still commanded.

But from 1998 onward, that response rate slid. To sate advertisers’ appetite for impressions, publishers began saturating their content with ads. When Microsoft’s car-shopping portal, Carpoint, debuted in 1997, there were no ads on its home page. By 2001, there were at least eight, not including sponsored links and pop-ups. As a matter of simple mathematics – even the most willing user can only click on one ad at a time – click-through rates declined accordingly.

But there were other factors that hastened the decline. The most obvious is the axiomatic one: consumers in a zero-sum game become inured to marketer’s tactics over time. Tactics that produced incremental gains quickly become overused dogma, whereupon they become ineffective. Because advertisers now had to compete for eyeballs in much bigger arenas, their methods became increasingly intrusive and deceptive: strobing ads, fake interfaces, and ads camouflaged as real content.

The most notorious example, still spoken of ruefully among Web marketers, is’s “PUNCH THE MONKEY AND WIN 20 BUCKS” ad, which invited the user to brandish a virtual boxing glove to punch a virtual monkey. Millions of users were duped into clicking, only to discover that they’d won 20 “banana bucks” that could be parlayed into real money only by playing even more games. The ad was so often decried by the industry’s doomsayers that some still hold it accountable for the near-death of the medium.

The truly tragic aspect of the direction that Web advertising went is that marketers saw the writing on the wall very quickly. From its debut in 1999, the Web marketing forum Clickz began fretting about the industry’s over-emphasis on direct response, believing it would lead to a crash. Topics covered the first year included “Escaping the Cult of the Click-Throughs” (Graham 1999), “Tracking Non-Click Conversions,” and “Between a Rock and a Hard Place,” which contained the quaint observation that click-through rates were “at an all-time low” (Hespos 1999). (The average response has since declined another 500%.)

It’s easy to be smug about the inevitable consequences of the new medium’s direct-response myopia, but in truth individual marketers were simply powerless to invert the widely accepted perception that banner advertising’s primary function was as a direct response medium. The industry produced study after study showing how exposure to banner ads increased brand awareness by some measurable delta. The Internet Advertising Bureau was formed mainly to advance that agenda, by standardizing ad sizes around more brand-friendly specifications and running studies on the impact of rich media. Certainly the evidence was persuasive, but it didn’t matter, because of another axiom: given the choice between hard and soft data, marketers will always choose hard. So unless the entire industry simultaneously stopped measuring click-throughs, it remained the only metric universally accepted as an indicator of campaign performance.

Then the crash came. Advertisers were more or less content to throw bad money after good in banner advertising as long as the Internet economy was strong. But when dot-coms started to bomb with greater intensity in late 2000, dragging the rest of the economy with them, online ad money dried up overnight. Start-up online media companies canceled IPOs, and public ones like rivals Avenue A and Doubleclick watched their value vanish. The mainstream media wasted no time in declaring the era of online advertising well over, and the Web’s ad volume shrank for the first time since its inception. It remained in decline for nearly two years.

In retrospect, it seems unfair that Web marketing was sent into the desert like a scapegoat, carrying marketers’ sins on its back. To this day Web marketers still complain, and quite justifiably, that the level of accountability between online and offline advertising is badly misaligned. We still argue about brand impact and still tout statistics to persuade advertisers to accept other metrics. But none of that really matters when we look at this story through the coolly objective eyes of the game theorist. Web advertising went the zero-sum route, and zero-sum is what it got. Its zero-sum mathematics went the only direction such mathematics can: the minimax point shifted toward the consumer. But it’s also true in game theory that that which does not kill us helps us find equilibrium, and that’s what happened here.

Interestingly, at least one business journalist observed the relationship between game theory and banner advertising’s race to the bottom early on. In a piece for Business World entitled, “The Unbearable Lightness of Ad Revenue,” Frank Yu declared, “Ad budgets are a zero-sum game and so are users’ attention spans.” He predicted that as “jaded, cynical consumers” learned to tune ads out, only the top content providers could afford to stay in the game, and severe “clustering” of content and media revenue would occur. He further predicted that new platforms like PDAs would challenge the Web and force new content monetization models (Yu).

Yu was at least partly prescient, if too cynical. Web traffic did indeed cluster around top content providers, but smaller players were able to stay in the game as a result of the Web’s transparency. Media planning tools like Nielsen Online (formerly Nielsen NetRatings) were able to ascertain the dimensions of the audience on more niche sites and allow advertisers to trade volume for relevance. The predicted changes brought on by new platforms are only now beginning to occur, with marketers taking notice of the growth of mobile applications as a small but rising threat to the now-traditional online advertising model. But the fundamental problem Wu raises – that of consumers tuning out – remains the industry’s greatest challenge.

What truly saved Web advertising was the equilibrium that occurred between response rates and media costs. While the minimax point shifted inexorably toward the once-bitten-twice-shy consumer during this period, the industry survived because the cost model shifted too. The cost has stabilized around a proportional rate of return that direct-response marketers can live with; in other words, the cost of impressions dropped alongside the rate of response. This has, in turn, eradicated most of the least tolerable tactics. Pop-under ads are largely a thing of the past, and fake interactions are mostly passé.

The limitations of this outcome are the same as they are for Sierra Trading Post: a more stable zero-sum game is still a zero-sum game. It leaves marketers with the basic problem of trying to eke out performance gains from a medium that is shifting inexorably away from direct consumer engagement. The stark reality of this marketer-consumer relationship was made plain by a 2007 study that sent shock waves through the digital marketing community. A joint study by media research company Comscore and media agency Starcom showed that a stunning 50% of all clicks on banner ads came from one small slice of the Web population: Web users aged 25-44 with a household income of less than $40,000 per year. Dubbed “Natural Born Clickers,” these users spend four times more time online than average users but purchase products at significantly lower frequency. Such users tend to favor gambling, employment, and auction sites – a much narrower pattern of surfing behavior than the Web population as a whole. A 2009 update to the study showed that the minimax point was continuing to slide. The percentage of monthly clickers fell from 32 percent in July 2007 to 16 percent in March 2009, with only 8% of Web users accounting for 85% of clicks (Comscore 2009).

From a game theory perspective, the implication of the “Natural Born Clickers” phenomenon is that it undermines the precarious equilibrium in click-based banner advertising. That equilibrium is based on the idea that the cost of finding and prompting action from the right targets compensates for banner advertising’s low response rate. If, however, that low rate of response also falls short of finding the right targets, the advertiser is no longer in equilibrium. Advertisers are then paying too much for the wrong kind of results.

Obviously the industry is in need of a game-changer – a shift in the use of the medium that moves it outside of the stark give-and-take of zero-sum. Fortunately for the banner ad medium, that game-changer has come in the form of more advanced metrics that account for the effects of advertising beyond direct response. Any of us can recall an instance of having seen an ad or a series of ads and having some later decision, e.g., which cars to research, informed by those previous impressions. This is, in fact, the way that advertising has always been understood to work: as one of many factors that add up to a purchase decision. Banner advertising, by contrast, had been operating under the fallacy that only a direct and immediate action, irrespective of whatever else the user might be doing, is the only way to account for the ad’s impact. Such an outrageous supposition easily leads to the Natural Born Clickers phenomenon, as clicking on an ad bears the lowest cost for a user who is at their leisure and has no intention of purchasing.

But the advent of advanced metrics disposes of this fallacy. Advertisers can now account for “view-throughs” of an ad, i.e., the perfectly natural phenomenon of a user seeing an ad and responding later. In rich media advertising, one can now account for interaction with the ad – certainly important in making a brand impression – as well as the brand impact of the ad. And banner advertising can be evaluated for its contribution to sales rather than to the fallacious clicks metric.

The digital marketer might rightfully protest that no other advertising medium is required to justify its existence in this way; it is the equivalent of demanding that billboard advertising account for consumers that spotted the sign and then later went to the store and purchased the advertised item. But again, game theory provides a ready explanation: once the payoffs in a game have been established, no single player can unilaterally change the rules. No bottom-line focused marketer wishes to give up hard metrics in favor of more logically persuasive but softer arguments concerning brand impact.

This is precisely why the advent of social media marketing is so important to the health of digital marketing as a whole: it provides the game-changer that demands different metrics, none of them easily obtainable, for how online conversations with consumers impact brand relationships. When viewed in the context of (as opposed to in conflict with) now-traditional tactics like banner advertising, social media marketing becomes a way of continuing a conversation that may be initiated in traditional ways.

How precisely social media marketing works in symbiosis with other forms of advertising is a topic for a later chapter. The main point of recounting banner advertising’s tumultuous journey is that its evolution away from direct response and toward a more nuanced role has led the way for more radical evolutionary stages represented by social media. And that evolution is reflected in the numbers: while marketers’ investment in banner advertising dipped, then stabilized, at a fraction of its former value, their total investment in the Web has grown year over year. This has occurred because interactive media has begun, albeit slowly and with no shortage of false starts, to offer a way out of the zero-sum game of direct-response marketing. The chapters that follow will demonstrate how zero-sum has evolved into more complex gaming scenarios that involve varying degrees of cooperation. These games offer an alternative to the uneasy truce of mutually assured destruction and pave the way toward a very different future for both players.