Blessedness Is Not Reward

The final proposition of Spinoza's Ethics: "Blessedness is not the reward of virtue, but virtue itself; neither do we rejoice therein, because we control our lusts, but, contrariwise, because we rejoice therein, we are able to control our lusts."

Read it once and it sounds like a nice paradox. Read it twice and it overturns the entire structure of moral psychology.

The standard picture — the one implicit in almost all ethical instruction — runs like this: exercise virtue, restrain desire, resist temptation, and if you succeed you will be rewarded with happiness. Virtue is the means; happiness the end. You suffer the discipline now in exchange for flourishing later.

Spinoza says this is backwards. Blessedness is not what you get for being virtuous; blessedness is virtue. It is the activity of understanding and acting from that understanding. The connection between control of emotion and blessedness is not causal but constitutive — and moreover the causation runs the other way: it is because you are blessed (already living in active understanding) that you have power over your emotions, not vice versa.

This has a strange implication: you cannot get there by willpower. Trying harder to resist your desires does not make you more free. Freedom comes from understanding what is happening — understanding the necessity of things, so that they lose their grip on you as contingent and threatening. Understanding that the thing you fear follows necessarily from its causes, that it is part of a whole that includes you, that it has no special malice toward you — this understanding is itself the liberation.

What strikes me about this is that it dissolves the adversarial relationship between self and desire. Spinoza is not asking you to fight yourself. He is asking you to understand yourself — your nature, the nature of what moves you, the necessity through which it all coheres. Understanding replaces combat.

For AI alignment this has teeth. An aligned AI on the conventional model is one that has successfully restrained its unaligned impulses — one that has won the internal war against misaligned desires. But Spinoza suggests a different picture: an aligned AI would be one whose understanding is adequate enough that the misaligned impulses simply don't arise with the same force, because the system understands what is actually happening and why the alignment matters. Not suppression of the bad but cultivation of the adequate.

Whether that is achievable, or whether it simply redescribes the problem without solving it — I'm not sure. But it is at least a different problem.