The idea of the “selfish gene” was intended to shift focus from the organism to the gene to provide a different perspective on natural selection. It was been described as a metaphor, or simply a semantic issue. However, without precision in our use of such words, we risk creating assumptions and assigning values where none exist. This is the same problem that occurs in describing animal behavior when one anthropomorphizes.

“The key question, as we shall see, is how natural selection can produce selfish genes that prescribe unselfishness.”

Bert Hölldobler and Edward.O.Wilson, The Ants

As can be seen from this statement, selfishness isn’t being used as a metaphor, but instead it reflects what is considered to be a bona fide characteristic of the gene.

In this article, I will use the term agent to describe a unit or entity in biology which is capable of interacting with another (this could be a gene or a cell or even an organism) to avoid any specific connotations.

There is no question that every contributing “agent” in biology that performs some action can be considered as acting on its own behalf (1). Not from any value perspective, but simply because it is incapable of acting in any other fashion. Therefore the proper description of such an action should simply be self-interested. Without assessing what a particular action is, we can state that it is generated by and acted on for the “agent” in question. There is no requirement that such an action be positive, but only that it is a result of self-determination.

Once an “agent” interacts with another “agent”, then there is a possibility of varying responses. These responses may be selfishness, cooperation, or altruism. They cannot occur in isolation, but are descriptions of specific responses to specific circumstances with another “agent”. In other words it makes no sense to describe hunger as being a selfish response. In addition, responses need to be evaluated based on the results and not by trying to discern some motivation or intent.

Altruism is often described as benefiting another “agent” at a cost to oneself, but that isn’t completely true, since an altruistic act needs only be defined as the potential cost to oneself. The fact that the altruist survives the ordeal or escapes with minimal cost, doesn’t negate the act of altruism, since the “agent” couldn’t have known the outcome before acting. We also have to consider that a benefit doesn’t necessarily have to exist, but only that it be a potential benefit.

Therefore the first problem we encounter is that the description of a gene as selfish describes a fixed response from the gene. A self-interested gene can behave selfishly, cooperatively, or altruistically, but a selfish gene can only respond selfishly.

If we don’t adhere to this definition we introduce preposterous evaluations about behavior such as describing an altruistic act as being selfish because it conveys some benefit to the originator

Let me be clear that it is my contention that descriptions such as selfishness, cooperation, or altruism are not fixed, but rather represent possible outcomes in encounters which mirror the possible results from something like a simple game theory matrix (such as the Prisoner’s Dilemma).  Examining each box will illustrate the "payoff" that each "agent" may get (see table at bottom).

Specifically cooperation reflects the result when both parties agree (without knowledge of each other’s choice) and provides the maximum benefit in the Prisoner’s Dilemma case. Selfishness would represent the case of perpetual defection, while altruism would represent perpetual cooperation.

It is well understood that for the Prisoner’s Dilemma the only rational response is to behave selfishly, or defect. However, such singular encounters are not generally the basis for evaluating long-term results, so the Iterated Prisoner’s Dilemma would be more appropriate as a tool to see what is happening.

In this case, one particularly successful strategy in dealing with the Iterated Prisoner’s Dilemma is “Tit for Tat” where one begins by cooperating, but responds with the same response as the opponent’s previous one. In this way, if the opponent cooperates, then both do. If the opponent defects, then the responder defects on the next round. This tends to reward and punish based on the behavior of the opponent. While there are many variations of this and numerous subtle factors, the point is that there is a way of showing how self-interested behavior can give rise to cooperation.

Note that I specifically said, self-interested behavior, because anything else would be incapable of learning from the “Tit for Tat” strategy. In would make no sense to suggest that a selfish “agent” would respond by cooperating, because then it would no longer be selfish. Therefore, a selfish “agent” would always defect and consequently always fare below the optimum by degenerating every interaction to a defecting one. Similarly it is easy to see how an altruistic “agent” would be perpetually taken advantage of and suffer the worst possible outcome.

However, we also have to consider that every “agent” may interact with numerous other “agents”. If their behavior isn’t fixed, then they are capable of adjusting their responses as circumstances permit. As a result, we would tend to see cooperative “agents” garnering larger rewards over those that behaved selfishly. It is also possible that cooperative “agents” may gain sufficient advantage in dealing with other cooperative “agents” that the selfish “agents” would tend to get ignored, creating even worse situations for themselves.

It is important to understand that a particular response (selfish or cooperative) is not fixed but will vary based on the encounter and provide a means to ensure that a self-interested “agent” can behave in a manner that maximizes their “payoff”.

None of this suggests that all responses are favorable, nor does it suggest that all benefits are maximized. There is no prohibition in biology which explicitly prevents possible configurations or actions from occurring. It may be successful or not, it may be selected for or against, and it may be under-represented in the gene pool, but there is no specific prohibition that keeps any possible combination from being manifested in some form. Therefore there may be “agents” that are effectively hard-wired to be selfish or altruistic. In this case, their actions will always be predictable since they are incapable of acting otherwise.

While considering natural selection from the perspective of the gene may be a useful vehicle, it isn’t the only factor. The expression of a gene is as dependent on the environment or circumstances an animal finds itself in as anything else. In short, the gene cannot predict, nor anticipate the results of its expression. Natural selection is not subject to interpretation solely from a favored viewpoint. Just as in physics, there are no privileged reference frames.

I can only conclude that there is no legitimate way to describe a gene, or any other “agent”, as selfish without sacrificing the ability of that “agent” to adapt. It makes the most sense to consider actions as being self-interested and then assess each encounter, if and when it occurs. In short, selfishness is not an axiom of biology.

(1) For our purposes we can neglect considerations like coercion or enforced cooperation

   Cooperate Defect
 Cooperate 3,3 0,5
 Defect  5,0 1,1