R.A.M.ifications: the Failure of the Rational Actor Model

The Rational Actor Model (RAM) is one of the principal foundations of Game Theory. It assumes that an actor, be it individual or corporate, will always act in its own self-interest when faced with a decision. Self-interest is taken to be objectively interpreted: there is a “best choice” in any situation that can be mathematically proven. Suboptimal choices are considered irrational. The concept of rationality itself is the very first problem with the RAM, as I will argue that there is no such thing as irrationality, and therefore rationality is a red herring. RAM tends to place all actors in the model on the same level in terms of environment, entropy, and data. This renders the model nondescriptive, as RAM cannot apply when actors come from different environments, have different weights assigned to their experiences (entropy), and have different levels of data about a situation. Data gaps are a fact of life and “perfect” data is an anomaly. These three ideas spell the death knell of the current RAM as a descriptive assumption. The only way to save RAM is to rid ourselves of the concept that there is such a thing as “irrationality”.

Irrationality does not exist because of environment

A poster child for the classical concept of irrationality is North Korea. Of the states in the world, few are more erratic, and none so erratic are so disruptive. The problem with analysing North Korea is that it can be seen through numerous lenses, and many of those lenses are not objective, but based on values judgements. Anyone who asks a question of Korea that starts with the words “why don’t they just…” and then recommends a course of action simply hasn’t put the time in to understanding North Korea as a rational actor. If anything other than the perpetuation of the regime was North Korea’s sole interest, such speculation would be potentially instructive. As it stands, we have a nation of some 20 million people which supports a tiny oligarchy – the only people who truly have “skin in the game” – in their chessboard of internal politicking. Internally, the Kim clan must assure dominance over the other petty power brokers, and externally, North Korea must milk the international community for aid while never truly allowing one state to become their sole interlocutor. Erratic acts – such as powering down the Yongbyon reactor only to power it up again secretly, kidnapping Japanese citizens and then releasing them some decades later, and sinking the occasional South Korean patrol ship then denying such activity on the world stage – serve to draw close and then alienate states each in their turn. So long as one state can give it aid and begin to make headway in the Hermit Kingdom, North Korea can afford to push away another for its own internal ends.

North Korea acts erratically, but not irrationally. Time and time again, results have proven to North Korea that such continuous games of “he loves me, he loves me not” get it the results it desires. Like a spoilt child, rewarded for his tantrums, so is North Korea a classic enfant terrible of the global stage. If such action is calculated to produce a desired result, and if such action is based on previous experiences of success, then how can it be called irrational? The best way to understand an actor like North Korea is not to ask “will North Korea react rationally to this offer”, but “what is the environment in which this action could be considered rational?” In the case of Korea, its experiences dictate that these actions will produce a desired result. The actions taken are erratic, but calculated; to consider North Korea anything but a rational actor in such a situation likely betrays an ideological presupposition on behalf of the interpreter.

In order to understand North Korea’s actions, we require access to data regarding its previous decisions and the results thereof. Environment is key – even applied Game Theory teaches us this. For example, in a 2005 paper (“Investigation of Context Effects in Iterated Prisoner’s Dilemma Game”), Evgenia Hristova and Maurice Grinberg detailed predictive strategies for cooperation in iterated Prisoner’s Dilemma games, finding that cooperation was not simply based on the games being played with the current opponent/partner, but all previous partners. Based on reading I had done in University, I determined that I would attempt an experiment myself that came out of Post-Modernist critique of the RAM through discourse analysis. I had a class of MBA students in a Business Strategy course divide into two teams. The professor of the course explained the rules for iterated Prisoner’s Dilemma in extremely competitive terms. For example, we made certain he said “opponent”, “beat”, “compete”, and “win”.  When I explained the rules, I used extremely cooperative terminology: “partner”, “resolve”, “cooperate”, and “participate”. The results were definitive, at least to me: In one-off games, the competitors always won. In iterated games, when cooperators were paired with cooperators, they by far scored highest. Any other pairing was so affected by the existing discourse – or became so disenchanted by their “partner’s” lack of cooperation – that they were dominated by defections. The history, the environment of each player, was a factor in their strategic choices. Those strategic choices were preordained by our simple choice of words. Were those choices irrational? No. They were purely environmental.

Notes on Game Theory and human choices:

http://users.auth.gr/~kehagiat/GameTheory/05PapersAdvanced/Shubik1970.pdf

http://rss.sagepub.com/content/4/1/8.full.pdf+html

http://dornsife.usc.edu/labs/alker/documents/reason_causes_games.PDF

parts 2 and 3 here!