Well, this applies to prisoner's dilemma games if you know exactly how many times you will "play" the game with a given other player, including the case where you will only play one time.

But for repeated prisoner's dilemma games, where you don't know what games is the last against a given player, a tit-for-tat strategy massively dominates an always-defect strategy.

Humans have evolved in environments where repeated prisoner's dilemma situation is common, and have evolved emotions such as empathy and consciousness for interactions with cooperative agents.

We ALSO have evolved (to a lesser or greater degree) the ability to turn such emotions off when dealing with people who attempt the always-defect strategy (such as psycopaths).

In fact, psychpathic behavior only becomes viable in populations that have been using tit-for-tat for so long that they've "forgotten" how to punish defectors (ie almost all agents switched to always-cooperate).

But once you introduce a few always-defect agents into the population, the tit-for-tat agents will soon dominate again.

At least for the brain wireing for interacting with members of our own "tribe". When dealing with people outside our "tribe", it's much more likely that we will only play them once, and in such cases, always-defect may indeed dominate.

However, there is also a tribe-level game, where the tribe can act as agents. And in tribe-vs-tribe games, depending on circumstances, both tit-for-tat and always-defect can be optimal strategies. Always-defect can, ultimately, lead to genocide, while tit-for-tat can lead to alliances, trade and intermarriage.

People who think that always-defect is always the optimal mode of behavior tend to be people who either don't consider it for their close relations, or who have very few close relations. For instance, someone who lives alone in the city and is either unemployed or tend to have purely transactional work relations.

What such people tend to forget, is that there is an artificial presence in the city, called the police, that can impose a level of safety that can remove the need for high trust, pro-socal relationships. In a society with police, all you have to do is to know what behavior lands you in prison, and otherwise you can be a selfish bastard.

As soon as you remove the police, loners like that are free game for "predators", and only those who have a "tribe" have any kind of protection. Such a "tribe" can be a clan in an Afghan mountain area or it can be a gang in a high-crime city. While a gang member may have little to no empathy or conscience when dealing with outsiders, maintaining good relations with other clan members can be critical (Depending on what specific gang it is.)

I think this is why it is very important with communities: so that your interactions with other people are mostly with the same people instead of mostly strangers.

The other important aspect is that people thinks "we have the same values, so we are the same team", as you will otherwise always select to deflect due to mistrust of the other part.

Police on its own is not enough, they cannot solve most crimes without help from citizens. A police officier without good informers is not good police ("The Wire", see IMDB). So police is only really effective in a high trust, pro-socal relationships society.

> But for repeated prisoner's dilemma games, where you don't know what games is the last against a given player, a tit-for-tat strategy massively dominates an always-defect strategy.

If we're considering repeated prisoner's dilemma games, you also have to consider the gain/loss of power between the various players. The classic prisoner's dilemma consists of two equally impotent players. Over a series of games, a player could aggregate enough power to change the nature of the game. They could aggregate enough power/resources to alter the parameters of the game, like using those resources to offer a substantial reward for cooperating with them or confessing. Inversely, they could use that power to enforce penalties on other players for betraying them.

The question then becomes whether an always-defect player can accumulate enough resources to change the rules of the game before the other players switch to tit-for-tat.

The answer in the real world appears to be "mostly yes". Uber appears to have chosen "always defect", but they got enough power through their defections that we've become effectively powerless to penalize them for defecting. Theranos is a counter-example where they were eventually punished, though they made the mistake of playing the prisoner's dilemma against already powerful players.

Without getting too political, Trump has a history of defecting on business relationships and the man was elected president.

> At least for the brain wireing for interacting with members of our own "tribe". When dealing with people outside our "tribe", it's much more likely that we will only play them once, and in such cases, always-defect may indeed dominate.

Within the tribe, the power levels tend to be flatter, though, making it easier for an always defect player to seize enough power to not have to play the game anymore. I.e. if the player can become moderately wealthy, the resources they offer their tribe would likely outweigh their history of defection. Company towns are an example of this; virtually all the residents agreed the company was awful, and yet they stayed (for a long while).

> As soon as you remove the police, loners like that are free game for "predators", and only those who have a "tribe" have any kind of protection. Such a "tribe" can be a clan in an Afghan mountain area or it can be a gang in a high-crime city. While a gang member may have little to no empathy or conscience when dealing with outsiders, maintaining good relations with other clan members can be critical (Depending on what specific gang it is.)

Sure, but that's not the prisoner's dilemma anymore, so always-defect is probably not an optimal choice.

A crux of the prisoner's dilemma is that a player can benefit by choosing to defect when the other player chooses to cooperate. If cooperating results in the best individual outcome and the best societal outcome, it's not a dilemma anymore. Cooperating is the obvious best choice.

This is part of what makes always-defect viable. Most situations do not mirror the prisoner's dilemma, and cooperating is often the best individual outcome. If you look at an always-defect player, they might be cooperative in 9/10 or 99/100 situations because only 1/10 or 1/100 were prisoner's dilemmas where they could benefit by defecting.

I don't think "always defect" is globally optimal, only for prisoner's dilemmas. Always-defect globally implies doing so even when it is clearly a sub-optimal choice (like driving your car into a brick wall just because the police said not to). The parameters of the dilemma scope it to only situations where defecting confers gain.

First I disagree that "success" is to be considered only in terms of accumulating material goods. It's also not JUST about feeling fulfilled and loved.

Second, I'm not denying that people are sometimes involved in "games" where always defect is indeed optimal. My objection is to the claim that always defect is universally optimal, even if you narrow "success" down to just maximizing wealth.

And, as I stated above, even prisoner's dilemma games tend to have tit-for-tat as the dominant strategy, since most games are repeated an unknown number of times.

That being said, there is an important element of competitiveness in capitalism. When you negotiate your salary (or that of your employees), you and the counterparty have conflicting interests. If you're naîve when playing the capitalism "game", you're likely to be taken advantage of.

I don't consider this a "defect" situation though, and definitely not a prisoner's dilemma "game". Capitalism tends to be a positive-sum game. This positive sum comes from a side effect of a free market, namely that supply and demand (including for labor) acts as a kind of computation algorithm that generally leads to more efficient than any organization is able plan using a top-down approach.

Imposing always-cooperate into capitalism (such as through labor union negotiated wages) may actually REDUCE this positive sum.

And if there is a game where the sum increases if at least one side defects, then it's not a prisoner's dilemma game at all.

> If you look at an always-defect player, they might be cooperative in 9/10 or 99/100 situations because only 1/10 or 1/100 were prisoner's dilemmas where they could benefit by defecting.

If 1/100 "games" are single-play prisoner's dilemma games, then we shouldn't be surprised if many chose to defect in those specific games. That actually doesn't bother me. Those people who defect in those games, but are cooperative 99% of the time are still probably behaving better than what I generally expect from people.

If you expect people to NEVER defect, then you may want to adjust your expectations. Basically such an expectation is the very essence of being naîve, and will act as a magnet on people who regularly take advantage of other people.