Sophisticated Tit for Tat

Today we are going to tackle the question that if you support an ethic of "Do No Harm", how do you deal with others who do not support such an ethic or equivalent, whatever they claim? That is they do in fact cause harm through their actions or inattention, often and mostly unnecessarily, regardless of how they justify this or not.

There are two levels of defence here: the primary one being the one we will consider now and the secondary one being that of the legal institutions: the law, the police, the judiciary and the institutions of sanctions and punishment. We are not going to cover these secondary and fall-back institutions here but for now, just assume that they do reflect the principles of Do No Harm, and exist generally to deal with the more extreme cases, the ones that are usually beyond the scope and ability for us as individuals to deal with, such as murder and its ilk. (This is a temporary assumption for our purposes here and will be an ongoing topic of investigation to find out the reality in future posts.)

The strategy we are going to use is based on the innovative, elegant and surprisingly simple solution to a challenging question within the field of game theory, namely the "tit for tat" solution to the iterated prisoner's dilemma. One preliminary point here is that this is not argument for an evolutionary basis for morality, nor for altruism, let a lone discussion of gene centric (selfish gene) versus multiple levels of selection arguments in evolutionary biology. (You might notice that some of the wikipedia entries linked here mostly lack proper citations, a surprising issue given the wealth of data and discussion but since this is not our concern here, I merely note this and will proceed). These are all irrelevant points to the following which is a pragmatic argument based on empirical discoveries and rational analysis within game theory per se. For the merit of brevity, I will provide only the briefest outline to make the argument here and if one is unfamiliar with these concepts one can pursue the links provided above.

The dilemma for the prisoners - disregarding the questionable issue of them both having committed a crime, one could also imagine they have been framed but cannot do anything about it - is due to how one is found guilty of this crime and facing the challenge of minimising the penalty for it. This is affected by whether they confess or not to the crime and so implicate the other prisoner and they are prevented from conferring with each other. If neither confesses ("cooperate" in the jargon of game theory) they will both suffer a light sentence; if both confess ("defect") they will both suffer medium sentences; if one defects (confesses) and the other does not (cooperates) then the defector gets the shortest sentence and the cooperator suffers the longest sentence.

Now the dilemma is that it is in both their interests to cooperate and share in both suffering only a light sentence. However since the one does not know what the other might do, both are more likely to defect leading to either the lightest sentence (if the other does not defect) or more likely both suffering medium sentences. The danger of cooperating under these circumstances is that this more likely to lead to the cooperator suffering the longest sentence (because the other has defected). This is the dilemma. This might look like a highly artificial and contrived situation but many real life situations can reflect this dilemma - of which this can be considered the skeletal outline -and this applies even more so the iterated prisoners dilemma.

The iterated prisoners dilemma is one where there are an unknown number of rounds. The question becomes what strategy can one do to maximise the benefit or minimise the suffering (if still using the concept of sentences). And one does not know what strategy will be implemented by the opponent. What strategy is robust against all types of other strategies?

Through various computer tournaments the result found is both surprising and interesting. There were many strategies but the one the gain the most benefit was virtually the simplest Tit for Tat. Start "nicely" (again game theoretic jargon here but meaningful nonetheless) by cooperating and then reciprocate whatever move the opponent made in the prior round. Regardless of the opponent strategy (more or less) this approach earned the most points! There are four important insights here:

1. Tit for Tat always starts with the benefit of the doubt - it is "nice" not "nasty" on the first move. This only really works for games with unknown amount of rounds.
2. Tit for Tat never wins a game! It either draws or loses every game it plays yet it still wins overall in terms of points accumulated. It loses the battle but wins the war!
3. This is because when it plays against a nice or mostly cooperative strategy they together gain far more points and when it plays a nastier and more defector oriented strategy it frustrates or thwarts it and so both earn far less points than they otherwise could.
4. It is "forgiving"(again game theoretic jargon here but meaningful nonetheless) and, within the constraints of this format, its actions are a means of signalling to opponent strategies a means to gain more points together, by cooperating not defecting.

OK having attempted to provide the minimal description required to see how this works, please read up the links above for more information if you need to, how does this apply to the real world? Let use the example and analysis of the recent mini-series on dealing with the bigoted question "How can one be moral without god?". Indeed it was selected because it requires a relatively simple level of analysis compared to many other real world problems. It is an exemplar, if you will, of the Sophisticated Tit for Tat I will shortly adumbrate.

First of all, just applying this considering two alternatives, cooperate or defect, is what I call the Naive Tit for Tat and typically unworkable in reality. If one operates something like the Do No Harm ethic and one guiding derived rules is to do not act in a bigoted fashion, when one deals with others who operate in the same fashion both can obtain other benefits. And certainly both start in a "nice" fashion and this just continues.When one deals with a bigot, it would appear that the naive solution would be to respond back with bigotry of something like the form "how can one be moral with god?". This is the naive analysis and the one I have been arguing against.

The intent of the Tit for Tat, that can be translated to the real world, is that the response should be to frustrate or thwart this move and to signal the two possibilities: (1) that if you repeat this I will continue to thwart you and (2) if you stop I will stop too (and we can move on to more mutually beneficial dialogues and actions). This is the Sophisticated Tit for Tat.

Now hopefully one can see the reasoning behind my suggested responses in the aforementioned mini-series. These were all designed to thwart the theist's bigotry, whilst also indicating one would continue to thwart this if they pursued it, but also signalling them alternatives that they could take, there or in the future and the choice is up to them.

This is as much as I wish to present to today to get the basic idea of Sophisticated Tit for Tat. There is much to explore and expand here which I can now develop in future posts.