# Claim It's likely that people in the future will be rewarded/penalized for the actions that they take today. For example, if today you take actions that increase probability that artificial intelligence will play out well, then you are likely to be rewarded for that in the future. Or if someone tries to concentrate power in their hands, they will be penalized for that. # Argument Firstly, if we want to talk about what people will do in the future, we need to have some framework for how to predict how people will behave in the future. I think that the standard game theory frameworks are not applicable here, because they don't take into account that humans are learning agents. For that reason, I'm going to assume [[Inductive decision theory]] - it's a descriptive decision theory that assumes that if there is some rule that explains players' past actions, then that rule is likely to continue to hold in the players' future actions. Additionally, the theory assumes that people are more likely to take actions that maximize their expected utility. Let's consider the following scenario. Let's suppose that there is John. John wants to do something good for the world that is not currently rewarded by the world - for example, doing some AI safety research and publishing it in a paper. If you write a paper about AI safety, you won't make money on it (at least not directly), but it is something useful for the world. Let's suppose that John writes the paper because he believes that he will be rewarded for that in the future. Later, at some point, company C creates and implements a new system of organization for the world. The system is better than the existing one. It can be created and implemented with the help of artificial intelligence or not. An important part of such system is to reward and penalize people for doing good/bad things. Because if people are rewarded/penalized for doing good/bad things, then they do / don't do good/bad things. So that new close-to-perfect system will be close-to-perfectly fair with regards to rewarding/penalizing people. Let's suppose that the creators of such system will decide that from the date X, where X is the time when the system was launched, people will be rewarded/penalized for the actions. That means that John will not get rewarded for their work because they did that before time X. However, the distribution of power can change in the world, since the date X. The people who become more powerful will be able to for example establish a new system and decide that the system will reward/penalize people since the date Y, where Y > X. By doing so, they can avoid accountability for their actions before Y, if it's in their interest. So, the only way that system can work is that people establish one date from which people are held accountable and then stick to it. But, in a sense, people already hold each other accountable for their actions, because there is already a legal system / economic system and many people also reciprocate good/bad actions of others. So, if we interpret "the system" in a broad way, then the system already has started. And we should interpret it in a broad way, because if we don't, then someone might create a new system, frame that system as something different and avoid accountability. For that reason, it is likely that people will be rewarded and penalized for the action that they take today. So, people have an incentive to behave well (in the interest of the entire collective) today. It's possible that it won't play out like in the described scenario, but to the extent to which the scenario is probable, to that extent there is an incentive to act in a way that is good for the collective.