Telecom & E-Commerce: The Multi-Agent Prisoner’s Dilemma

Telecom and E-Commerce are perfect grounds for the multi-agent Prisoner’s Dilemma game. What behavior we set as a norm for the industry is critical for the success of the industry as a whole. Imitating a defector, which is the current norm, could destroy these industries; imitating cooperative behavior by all agents could give a life line and increase the size of the pie for all.

Game Theory insights for two agent Prisoner’s Dilemma game gives a standard solution to the game, one of cooperation (for the common good) rather than defecting (selfish gains for the short term), as the pay-off matrix suggests. The repetitive game symbolizes the quandary of two players trying to get market share in a common market space and by cooperating with each other both could increase the pie-size in the long term, while defecting/rejecting to cooperate would eventually reduce the size of the pie for both.

This is rather simple and does not mirror the complexities of the market where multiple agents try to compete and the interactions can hardly be didactic given that the ecosystem evolves from the collective responses of a range of constituencies, not necessarily some of these responses could be termed from Game Theory parlance as “paired interactions”; some could be more determined by choices made under bounded rationality as information asymmetry exists and behavior under uncertainty cannot be fully judged through the logical lens we are so used to; personalities of the players in question become more important sometimes.

The fact that the game has many existing players (including some winners) and some vying to enter as entrants means that the objectives of all the players cannot be the same in the game and nor can we assume that all players have full knowledge of what the other is trying to do. Assessing the response of one player against a range of stimuli actually becomes very difficult given the time constraint and without cooperation it is only to be expected that each is trying to play the game her own way for whatever objective function she is attuned to deliver.

If we look at the E-commerce market, the size of the pie is so large that growth for all players is a given; but rarely do we see each trying to cooperate with the other on setting floor to the prices therefore or to the discounting actions each take or the frills each adds to the delivery model.

The Telecom is a more puzzling game, as multi-agents respond with a more short term tactics rather than for the long term (I am not referring to the current situation where one could be baffled by the entry of a new entrant who has declared a win before even starting the game); such are the oddities of personalities in a multi-agent game. But here also the ultimate market is so huge there is no denying the huge size of the pie for all agents to enjoy.

But rarely do we see cooperation as a more rational alternative to defecting as the pay-off matrix suggests as follows:

1. Regardless of what the other agents do, each agent receives a higher payoff for defecting than for cooperating behavior.
2. All agents receive a lower payoff if all defect than if all cooperate

When defecting is set as a norm, there is more imitation that subjugates one and all.

Imagine when a new entrant enters with a deep enough pocket, the typical short term response is to gain immediate market share at the cost of the overall size of profits shrinking in the market.

Each new entrant repeating the same behavior delivers even lower profit streams for the industry. Whereas the growth in this industry is a given, with such low penetration rates, at least in a country like India.

Here comes the issue of personalities, especially when different agents behave differently to the same stimuli. We have five types of personalities:

1. Pavlovian: In which satisfactory results from a response is re-inforced (vice versa), which means the probability of cooperation or defection changes with the environment. They are more rational than the others. If all agents are Pavlovian then there will be stability and eventual equilibrium.
2. Stubborn or Stochastically predictable, where under any circumstance the response is constant.
3. Conformist: Immediately imitates the action of the majority
4. Accountant: Who imitates others
5. Greedy: Imitates the one with the highest reward

The simulation environment is important in this game and for every action of a player the responses of the others could be plotted in n-dimensions. The curve for a cooperator and that of a defector could be plotted for the pay-off.

The pay-off functions are non-linear when we have different personalities as agents and for a multi-agent game the total pay-off maximization point for the industry may not coincide at the maximum cooperation by all players, but surely it does not coincide with the minimum cooperation as is the case for most Telecom and E-Commerce players in the market.

When a new entrant, with the express intention of disrupting the market, comes into the fray, it would be prudent to target test markets to understand the responses of other agents in the market, rather than going all over with a bang.

But that is hardly to be expected from personalities who run the show; but the suffering fortunes of the industry make a gloomy story of what could have been the outcomes had cooperation instead of defection been the norm.

Setting norms for behavior would be more important for E-Commerce and Telecom players, than the tryst with big-data and analytics that almost everyone is busy with imitating each other.