The previous chapter viewed the individual as a society of competing interests and posited an internal politics of the self. The study of human history reveals that polities may be characterized by periods of relative harmony, with effective social action to achieve common goods, interspersed with periods of bitter conflict, sometimes descending into civil war in which public goods are squandered. During periods of social peace, how are different factions able to work together despite divergent ideas about the just society? In the outbreak of civil strife, why does compromise collapse into internecine hostilities? When we come to consider the society of the self, the Hobbesian war of all against all is not easily resolved by submission to a Leviathan with a monopoly on force. We are a republic with dispersed centers of power rather than an absolute monarchy or dictatorship. How then do we achieve the benefits of cooperation and avoid the dangers of civil war? Can we come to internal compromises to which all parties accede? How do we resolve internal conflicts? Can we agree to disagree within ourselves? This chapter explores the possibility of strategic negotiation between contending parties in an internally complex self and addresses the question whether there can be intrapersonal reciprocity (Trivers 2011).
Introspection suggests that I often attempt to modify my own behavior by an internally voiced mixture of exhortation, bribes, and threats. Such intrapersonal cajolement has limited effectiveness, in part, because threats and promises to myself lack credibility. If I renege on a contract with myself, who will enforce sanctions? If I offer myself a reward now—for later good behavior—why am I bound to fulfill the second half of the bargain? Is it credible that I would knowingly do harm to myself to punish a past transgression? These attempts at intrapersonal persuasion pose a philosophical conundrum. I know what I know and I know what I want. So, why do I need to persuade myself? One might argue that internal bargaining is simply a reuse, or misapplication, of tools that are effective in the control of others’ behavior. I modify my behavior in response to the threats and bribes of others and, in turn, use threats and bribes to modify others’ behavior. Why not use these same methods on myself? Or, one might argue that intrapersonal bargaining resembles interpersonal bargaining because the self contains multiple agents with different, sometimes conflicting, agendas (Ainslie 2001).
Evolutionary conflicts arise within organisms when different replicators have different rules of transmission. Could these conflicts underlie problems of self-control? The potentially competing agendas within the self, both genetic and memetic, are complex. Therefore, to focus discussion it will be useful to consider a simple exemplar of an internal genetic conflict. For this purpose I will consider the conflict discussed in the previous chapter between the genes an individual (call him Bob, in honor of Robert Trivers) inherits from his mother and father.
Suppose that Bob’s parents divorce and remarry, and that Bob’s mother and father each have a child with their new partner. Bob thus has a maternal half-sibling (Maddy) and a paternal half-sibling (Paddy). What cost (C) would Bob’s genes be prepared to accept in exchange for a benefit (B) to Maddy? Bob’s maternal genes each have one chance in two of having being transmitted to Maddy. Therefore, his maternal genes would favor Maddy receiving the benefit, as long as the value to Maddy exceeded twice the cost to Bob (B > 2C). However, Bob’s paternal genes are absent from Maddy because Bob and Maddy have different fathers. For Bob’s paternal genes, no benefit to Maddy (no matter how great) could justify any cost to Bob (no matter how small). Bob’s maternal and paternal genes are in conflict whenever B > 2C > 0. A similar conflict exists over benefits that Bob could confer on Paddy, but in this case the roles of maternal and paternal genes are reversed.
How could Bob’s internal conflict be resolved? The traditional answer invokes a “veil of ignorance” (Rawls 1971). If genes lack information about their parental origin, they are constrained to behave in the same manner when inherited from either parent. If one of Bob’s genes is faced with a choice whether to help Maddy, the gene is equally likely to be paternal, and necessarily absent from Maddy, or maternal, with one chance in two of being present in Maddy. Therefore, an uninformed gene has an expectation of one chance in four of its copies being present in Maddy and would therefore favor transfer of the benefit if B > 4C.
What theoretical predictions can be made about how the conflict would be resolved if the relevant genes are imprinted and thus possess information about their parental origin? Genes of both parental origins would favor withholding the benefit from Maddy if B < 2C and both sets of genes would favor conferring the benefit if this directly benefited Bob (C < 0), but specific predictions of outcomes within the zone of conflict defined by B > 2C > 0 require assumptions about the mode of gene action and the relative power of the contending parties to influence the decision.
One possible resolution of a conflict is that one party has the power to dictate the outcome. If Bob’s paternal genome had the dictatorial role, the benefit would be conferred only if it was without cost for Bob (C < 0), whereas if Bob’s maternal genome had the dictatorial role, the benefit would be conferred whenever the benefit to Maddy was more than twice the cost to Bob (B > 2C). In the first case, Bob’s paternal genome could be considered to have a veto as to whether the benefit is conferred, whereas, in the second case, Bob’s maternal genome could be considered to present his paternal genome with a fait accompli (Haig 1992b). Whether a particular dictatorial decision is interpreted as a veto or fait accompli may simply depend on how the question is framed. Models of gene expression at an imprinted locus result in a dictatorial outcome when maternal and paternal alleles have conflicting interests over the level of gene expression. At evolutionary equilibrium, whichever allele favors the higher amount produces that amount (the “loudest-voice-prevails” principle).
Stalemates could result if different genes had dictatorial power over different aspects of a decision-making process. Wilkins and Haig (2001) provide an example of this process in a two-locus model of the coevolution of a paternally expressed demand enhancer and a maternally expressed demand inhibitor. Increased paternal production of the demand enhancer favors increased maternal production of the demand inhibitor, which favors increased paternal production of the demand enhancer, and so on, until the evolutionary spiral comes to a halt because any further escalation would be prohibitively expensive for both parties. At evolutionary equilibrium in this model, the loudest voice prevails at both loci and there are substantial costs of conflict.
A possibility that has received little attention is that maternal and paternal genes could reach a compromise between their competing interests and avoid the full costs of evolutionary escalation. Suppose that separate opportunities arise for Bob to confer benefits on Maddy and Paddy. If the two opportunities were considered separately, the inclusive fitness of Bob’s paternal genes would suffer if the benefit were conferred on Maddy, whereas Bob’s paternal genes would suffer if the benefit were conferred on Paddy. Therefore, if Bob’s maternal and paternal genes each had an effective veto on decisions whether to confer a costly benefit on half-siblings, neither benefit would be conferred. However, if the two decisions were considered jointly, both sets of genes might be able to increase their fitness if they could strike a deal. In effect, Bob’s maternal genes could offer his paternal genes a bargain: “We will let you help Maddy if you let us help Paddy.” The value of the deal for both parties would be B/2 – 2C. That is, each of Bob’s genes experiences the costs of conferring the benefit on both Maddy and Paddy, in return for an expected half-share in the benefit to Maddy, but no share in the benefit to Paddy (or vice versa). If B > 4C, both sets of genes would receive a net inclusive fitness benefit relative to what they would achieve if they exercised their respective vetoes. If the deal can be trusted, it should be accepted.
In the above example, genes that know their parental origin settle on the same decision rule (B > 4C) as genes that are uninformed about their parental origin. This is a consequence of artificial symmetries in the model and the limitation of bargaining to a single take-it-or-leave-it deal. Suppose instead that Maddy and Bob live together with their mother in Melbourne but Paddy lives with Bob’s father in Dublin and never interacts with Bob. Bob’s only opportunities to benefit a half-sibling involve Maddy. Bob’s maternal genes have no opportunities to exercise their veto on benefits to Paddy and, therefore, have nothing to offer in exchange for Bob’s paternal genes withholding their veto on benefits to Maddy. In this scenario, uninformed genes would use the decision rule B > 4C, but those of Bob’s paternal genes that are informed of their parental origin have dictatorial power and should exercise their veto whenever C > 0.
Now consider an intermediate case in which there are two opportunities to benefit Maddy (combined benefit to Maddy = 2B; combined cost to Bob = 2C) for each opportunity to benefit Paddy (benefit to Paddy = B; cost to Bob = C). In one possible deal (“two-for-one”), Bob’s paternal alleles would withhold their veto on both benefits to Maddy in exchange for Bob’s maternal alleles withholding their veto on the single benefit to Paddy. The value of this deal to Bob’s maternal alleles is B – 3C whereas its value to Bob’s paternal alleles is B/2 – 3C. Bob’s maternal alleles would accept the deal if B > 3C whereas his paternal alleles would accept the deal if B > 6C, but reject the deal if B < 6C. However, a deal in which one benefit to Maddy is exchanged for the benefit to Paddy (“one-for-one”) is also possible. As shown above, this deal would benefit both parties (relative to no deal) if B > 4C. If B < 4C; neither deal is acceptable to Bob’s paternal alleles who should exercise their veto. If 6C > B > 4C, the “one-for-one” deal is acceptable to both parties, but the “two-for-one” deal is unacceptable to Bob’s paternal alleles. “One-for-one” should be accepted by both parties. If B > 6C, either deal is superior to no deal for both parties, but Bob’s paternal alleles would prefer “one-for-one” whereas his maternal alleles would prefer “two-for-one.” Theory provides no simple prediction as to which deal, if any, will be chosen. The complications of multiple possible contracts are exacerbated if costs and benefits are continuous rather than discrete.
Genes of course are not subject to binding contracts. In the absence of enforceable contracts, the proposed deal—trading benefits to Maddy and Paddy—has the nature of a prisoner’s dilemma. If the effects of genes are fixed and not conditional on the behavior of the other genes with which they are temporarily associated, the deal is a one-shot prisoner’s dilemma and is evolutionarily unenforceable. Although mutual defection is the evolutionary stable strategy (ESS) if a prisoner’s dilemma is played once with a particular partner (“I stab your back, you stab mine”), cooperation can sometimes succeed if the game is played repeatedly with the same partner and there is uncertainty about when the relationship will end (Axelrod and Hamilton 1981). If this criterion—long-term association of uncertain duration—were the only consideration, strategic cooperation among genes within an organism should be possible.
Most successful cooperative strategies in iterated prisoner’s dilemmas (IPDs) require memory of past interactions to allow a player’s actions to be conditional on the past behavior of its partner. If genes were to implement a simple strategy like tit for tat (“Do what the other player did on the previous move”), they would need to perceive, and remember, a partner’s previous move and use this information to guide current behavior. This requires a degree of sophistication that some may be wary of ascribing to genes, even though genes are known to possess forms of memory and to exhibit complex conditional expression. One might ask, however, why cooperation should be more difficult to evolve for interactions between genes within organisms than for interactions between genes in different organisms.
Analytical investigations and computer simulations of the IPD have revealed a wealth of possibilities for complex behavior. In general, no strategy is an ESS and multiple strategies may coexist in a population. Polymorphism is possible because two strategies may be indistinguishable when played against each other, so that their relative performance will be determined by their interactions with third strategies, including the degenerate versions of themselves produced by mutation (Bendor and Swistak 1997). In the context of intrapersonal reciprocity, genetic polymorphism would raise the intriguing possibility that different combinations of strategies in the internal IPD could contribute to personality differences: sometimes one interest gets its way; sometimes the different interests come to a relatively harmonious compromise; sometimes the different interests contest every point of contention.
It is tempting to simply transfer results obtained from theoretical investigations of the IPD to discussions of reciprocity within the genome. I resist the temptation because most simulations of the IPD have competed strategies in a stereotypical tournament (Bendor and Swistak 1997). In each round of the tournament, pairs of strategies play multiple rounds of the prisoner’s dilemma, with “fitness” determined by the sum of payoffs. In the next round, a strategy increases or decreases in frequency determined by its performance relative to the average fitness of the population. Significantly, a strategy is inherited as a whole and does not undergo recombination. This kind of tournament models asexual or single-locus inheritance of strategies.
Consider, however, an alternative tournament in which teams play against each other, and in which the members of a team have different tasks in the collective implementation of their team’s strategy. In this tournament, a team’s relative performance in one round determines the probability that its members will participate in the next round, but the teams themselves are ephemeral. At the end of each round, members of successful teams (or their clonal progeny) are reorganized into new teams which reenter the tournament to play against another reorganized team. This kind of tournament models sexual or multilocus inheritance of strategies, with the units of inheritance determining components of strategies rather than entire coherent strategies. A successful component must do well, on average, across all the combinations in which it is tested. The sexual tournament would appear to select for a team of champions rather than a champion team. It is possible that the simpler asexual tournaments provide a good guide to equilibrium behavior in the more complex sexual tournaments, but this is something to be demonstrated rather than simply asserted.
Speculation about intrapersonal reciprocity is fun, but prompts the question how one would ever know that it occurs. How does one distinguish strategic cooperation from veil-of-ignorance cooperation or from escalated noncooperative stalemates? The problem of testability has bedeviled progress in the scientific study of internal conflicts, because there may be little that we can observe in external behavior that allows us to test hypotheses about internal conflicts. Progress in this field is likely to come from understanding the details of genetic mechanisms, just as an appreciation of internal conflicts between paternal and maternal genes has in large part depended on increased knowledge about the molecular functions of imprinted genes (Wilkins and Haig 2003). One would need to demonstrate conditionality of gene expression, not just expression that is conditional on parental origin, but also expression that is conditional on the behavior of other genes. This is a challenge, but surely not an insurmountable one.
For the flesh lusteth against the Spirit, and the Spirit against the flesh: and these are contrary the one to the other: so that ye cannot do the things that ye would.
—Galatians 5:17, King James Version
We often have the subjective experience of struggling with ourselves, of a conflict between powerful internal voices in which neither side yields nor gives up the fight, in which neither side plays fair. Elsewhere I have argued that the multiple “factions” within the genome tend to align into two opposed “parties” and that the hardliners of each party often drive the debate on substantive issues (Haig 2006). Many diagnoses of the etiology of our internal conflicts invoke two opposing forces, variously described as a struggle between impulse and control, passion and reason, temptation and conscience, sin and virtue. There is a family resemblance in the axes of dispute in these diagnoses, perhaps analogous to the first principal component in a multidimensional space of conflicting forces. And the conflict of these forces is sometimes governed by an executive controller or judge. Examples include Socrates’s simile of a tripartite soul, a charioteer struggling with a dark and white horse, or Freud’s Ich mediating the conflict between Es and Über-Ich (egregiously translated as ego, id, and superego).
Why do some minds function harmoniously whereas other are consumed in recrimination and internecine strife? The standard biological approach to mental illness is to search for a broken mechanism; but could some discontents of the self be conceptualized as an expression of a dysfunctional community rather than a malfunctioning machine? Could such a shift in perspective help to heal troubled souls? I have been timid about speculating on the implications of internal conflicts for human psychopathology because it would be easy to be misunderstood as advocating crude genetic determinism and to be accused of being insensitive to nuance and of failing to understand the “insuperable” divide between mechanisms and subjective experience.
One issue raised in this chapter is whether we can credibly threaten ourselves to coerce our own behavior. Could a “faction” within the self follow through on a threat in order to make future threats credible? “Self-destructive” behaviors immediately come to mind. Could these behaviors sometimes result from a breakdown of reciprocity among the factions of self, a fraying of an internal social contract, with the infliction of actual harm in a desperate attempt to change future behavior? For those of us who have known individuals who have engaged in self-mutilation of body or mind, the behaviors can be hard to fathom. From my perspective as a completely unqualified observer, the outward signs seem to express a degree of anger of a faction within the collective self toward the self, combined with a desire to prove to the self the strength of the faction’s resolve. But this may be a failure of empathy on my part, with an exaggerated projection onto others of my own inward anger when the “meddling” of one internal faction “frustrates” the plans of another.