Altruistic Systems

05/09/15

I've been thinking lately that we have pretty good evidence that how morally people act in society depends largely on the feedback systems in place. Look no further than the state of the police in the United States to see this in action. There are various systems in place that incentivize immoral behavior for the sake of passing the metrics that measure how good a job is being done, and also for the sake of actually funding the police.

Which makes me think: how could we better design systems such that people are not incentivized to act immorally?

I think first I should consider how important morality really is to most people, even me. Even trying to consider this question deeply, I find my thoughts wandering, "I should probably start cooking as soon as I get home." Morality is something that has intrinsic value, and I believe both short and long term emergent value, but how it helps the world is kinda invisible.

It's only something to notice when it's broken, or when in comparison to where it is broken.

But right now, things are broken, so I think morality can reveal a lot of value. I think the key is to make intelligently designed systems that incentivize morality, or at least no longer incentivize evil.

But how? Well, there are a few questions to ask there:

• Does the source of income rely on
  • abusing the system?
  • moving in the opposite direction of the goal?
• Are employees
  • encouraged to abuse the system?
  • encouraged to fix the system?
  • free to offer feedback on job difficulties?
• Are users
  • frustrated with the system?
  • expecting the system to do something it isn't?
  • in disagreement that the system should exist at all?
  • not being considered by the system?

I think long strides to morality can be made by making sure a system is aligned with its purpose and that that purpose is made transparent to all members affected by that system. Any piece of the system that is at conflicts with its transparent goal need to be assessed and mitigated.

But even necessary and good goals often have negative consequences. Automation is set to make many jobs obsolete, and none of the systems working towards putting automation in place has any kind of safety net setup for the displaced.

Systems that consider even this aspect and work to mitigate or erase their negative side-effects would then be altruistic systems. I suppose I could take it a step further and say that systems that try to handle problems outside of their specific goal and source of income, but meets its skillset and solves a real need, would also be altruistic systems, but there are lots of outliers there to consider.

I believe I have more thinking to do on this.