Certainly the irreversibility of a mistake is a huge mark against the death penalty, relative to the alternative (life in prison without parole). But I could handle that as an "acceptable risk" if there were something weightier in its favor. For example, picture if most criminologists believed that the death penalty were an additional deterrent against crime, above the alternative of life in prison without parole. Let's say your typical criminologist projects that there will be 100 murders if we execute 50 murderers, but 150 murders if we merely sentence them to life in prison without possibility of parole. That would be 50 lives saved in exchange for 50 lives taken, and, even with the risk of false convictions, it seems reasonable to assume that the 50 people executed have, on average, less "right to life" than the 50 people saved by the existence of a death penalty. If those were the facts, I could be a death penalty supporter. But, instead, most criminologists believe there isn't an incremental increase in deterrence from the death penalty.
A person who commits the kind of crime that makes life in prison without parole the sentence in non-death-penalty states and countries didn't make that decision because he assessed the sentence and deemed it one worth paying. He did it either because he didn't think about the sentence at all ("heat of the moment") or because he thought he wouldn't be caught. If the death sentence had existed in the place where he committed the crime, he'd have been exactly as likely to commit the crime.
Have you ever looked at the way the Passer Rating (QB rating) is calculated in the NFL? The formula is done in a way that there are several different factors, and you can maximize or minimize a particular factor, such that better or worse performance in that factor no longer draws the rating up or down:
http://en.wikipedia.org/wiki/Passer_rating
So, for example, if you've already thrown enough touchdowns per attempt to maximize the TD/ATT factor in the rating, you can throw ten more touchdowns in a row and not increase that factor one more bit. The only way to push the Passer rating up further is by focusing on some other factor (like interception percent).
I mention this because I think that's how people's minds work when it comes to the sentences for crimes. At some point a particular factor in their mental calculus is "maxed out," and increasing it even more doesn't change the outcome of the formula. Obviously, increasing the severity of a sentence from something fairly mild (a week in jail) to something very severe (20 years in jail) makes a huge difference. But at some point it's only registering as "100% unacceptable" and further increases won't matter... the person is only going to commit the crime because he is thinking about the sentence at all (heat of passion), or because he thinks he'll get away with it. The only way to deter a person more, once you reach that point, is by working on other factors (e.g., the apparent likelihood of the person being caught, or the likely benefit to the person if he gets away with it, or the desirability of competing alternatives). Life in prison without parole is "100% unacceptable" as a sentence -- people only commit crimes that have that sentence because they aren't thinking about the consequence or they think they'll get away with it, and further increasing the severity of the punishment to death won't change the analysis for them.
That's all just my own theory about how the psychology works that results in the death penalty not being an incremental deterrent over a life sentence. But the fact it isn't an incremental deterrent is pretty widely accepted among criminologists, and it's supported by all sorts of statistical analysis. When you compare death-penalty states and countries to non-death-penalty states and countries you just don't see the data fingerprints that would be predicted if having the death penalty acted as a further deterrence to crime, above what you get with life sentences. And when you look what happens to crime in a particular place when the death penalty is instituted (using comparable places that don't implement it as controls), you don't see a change. It's simply not factoring into the potential criminal's decision-making, at least any more strongly than the life sentence would already have factored in.
It's that analysis that leads me to oppose the death sentence. If it's not deterrence, it's just revenge. It's just killing for the catharsis of killing. And even if it weren't for the very real risk of false convictions (which, essentially, are state homicides), I couldn't support that. I can approve of taking human lives to save human lives, but not merely to vent anger.
